N A S A TECHNICAL NOTE

N A S A TN 0-8384

C*!

__

*
Qo

0s)
n

0
t-

z
c

W
.I=

0
c3r

w
j

MINIMUM DISTORTION QUANTIZERS

Harry We Jones, Jre
Ames Resedrch Center

MofSett Field, Cali$

94035

N A T I O N A L AERONAUTICS A N D SPACE A D M I N I S T R A T I O N

WASHINGTON, D. C.

MARCH 1977

TECH LIBRARY KAFB, NM

I111lllllll l lllIIIlllIlIl

NASA TN ~ - 8 3 8 4
4. Title and Subtitle

2. Government Accession No.

7. Author(s1

~~

5. Report Date

March 1977
6. Performing Organization Code

8. Performing Organization Report No.

Harry W. Jones, Jr.*

~

3. Recipient's Catalog No.

i
II

MINIMUM DISTORTION QUANTIZERS

A-6714

I 10. Work

9. Performing Organization Name and Address

Unit No.

656-11-02-01

Ames Research Center

Moffett Field, California 94035

11. Contract or Grant No.

I

13. Type of Report and Period Covered

12. Sponsoring Agency Name and Address

National Aeronautics and Space Administration

Washington, D.C. 20546

Technical Note
14. Sponsoring Agency Code

I
15. Supplementary Notes

"National Research Council Postdoctoral Research Associate.

~~

16. Abstract

The well-known algorithm of Max is used to determine the minimum distortion quantizers for
normal, two-sided exponential, and specialized two-sided gamma input distributions and for meansquare, magnitude, and relative magnitude error distortion criteria; The optimum equally-spaced
and unequally-spaced quantizers are found, with the resulting quantizer distortion and entropy.
The quantizers, and the quantizers with entropy coding, are compared to the rate distortion bounds
for mean-square and magnitude error.

18. Distribution Statement

17. Key Words (Suggested by AuthorW)

Quantizers, Minimum distortion

Quantizers, Max
Distortion measures
Rate-distortion theory

Security Classif. (of this report)

Unclassified

Unlimited

STAR Category - 32
20; Security Classif. (of this page)

Unclassified
'For sale by the National Technical Information Service, Springfield, Virginia 22161

MINIMUM DISTORTION QUANTIZERS

Harry W. Jones, Jr.*
Ames Research Center
SUMMARY

The well-known algorithm of Max is used to determine the minimum distortion quantizers for normal, two-sided exponential, and specialized two-sided
gamma input distributions and for mean-square, magnitude, and relative magnitude error distortion criteria. The optimum equally-spaced and unequallyspaced quantizers are found, with the resulting quantizer distortion and
entropy. The quantizers, and the quantizers with entropy coding, are compared
to the rate distortion bounds for mean-square and magnitude error.

INTRODUCTION
The well-known optimum quantizers and optimum, equally-spaced level quantizers of Max (ref. 1) have minimum mean-square error distortion for a given
number of output levels, assuming a normal or Gaussian distribution of the
input parameter. Paez and Glisson (ref. 2) used the numerical algorithm of
Max to find optimum quantizers and optimum, equally-spaced level quantizers
for minimum mean-square error distortion, assuming either the two-sided exponential (Laplacian) distribution, or McDonald's special form of the g a m a distribution. Here, the work of Max and of Paez and Glisson has been repeated
and extended to higher and intermediate numbers of levels for the mean-square
error distortion and for the normal, exponential, and gamma distributions. In
addition, the optimum quantizers and the optimum equally-spaced level quantizers were found for these three distributions using the magnitude error distortion and the relative magnitude error distortion criteria suggested by
Andrews and Pratt (ref. 3 ) . The method of Max is reviewed, and the input distributions and distortion measures are defined. The quantizers are given and
discussed, and their performance described. The new results are listed in the
summary and conclusion section at the end of this report.
MAX ALGORITHM

If the input parameter distribution is p(x)

and the distortion for
input parameters x and representative value yi is d(x - yi), then for M
representative values for the parameter, the total distortion is

*National Research Council Postdoctoral Research Associate

rxi+l

where x i and
are t h e c u t p o i n t s d e t e r m i n i n g t h e r a n g e of x r e p r e s e n t e d by t h e v a l u e y i , and x1 and xM+1 are i n f i n i t e . The d i s t o r t i o n i s
minimized by d i f f e r e n t i a t i n g D w i t h r e s p e c t t o t h e x i s and y i s and
s e t t i n g t h e d e r i v a t i v e s e q u a l t o z e r o . For p ( x ) # 0 f o r a l l x and d ( x )
m o n o t o n i c a l l y i n c r e a s i n g w i t h x , Max ( r e f . 1, p. 268) shows t h a t s e t t i n g t h e
d e r i v a t i v e w i t h r e s p e c t t o x i equal t o zero r e q u i r e s t h a t

Thus, t h e c u t p o i n t x i i s halfway between t h e two r e p r e s e n t a t i v e v a l u e s y i

and ~ i - The
~ .d e r i v a t i v e w i t h r e s p e c t t o y i i s a l s o s e t t o z e r o , g i v i n g

SXi

d(x - yi)p(x)dx = 0

i = 1,

. . ., M

(3)

xi

S i n c e p ( x ) i s symmetrical a b o u t x = 0 , z e r o i s a c u t p o i n t f o r M-even
and
a r e p r e s e n t a t i v e v a l u e f o r M-odd, and t h e p o s i t i v e and n e g a t i v e r e p r e s e n t a t i v e l e v e l s and c u t p o i n t s a r e s y m m e t r i c a l , h a v i n g t h e s a m e magnitude and
o p p o s i t e s i g n s . The i n d e x i n g of t h e x i and y i i s t h e r e f o r e changed s o
t h a t x1 and y1 a r e t h e smallest n o n n e g a t i v e x i and y i , and x ~ / yfi12
~ ,
f o r M-even, o r X ( S - ~ ) / ~ ,
f o r M-odd
are t h e l a r g e s t p o s i t i v e noni n f i n i t e x i and y i .
For u n e q u a l - l e v e l s p a c i n g and M-even, x1 i s z e r o and y1 i s e s t i m a t e d .
Then, e q u a t i o n ( 3 ) i s s o l v e d f o r x2, e q u a t i o n (2) i s s o l v e d f o r y 2 , and s o
on, u n t i l e q u a t i o n ( 2 ) i s s o l v e d f o r yGl2. I f e q u a t i o n ( 3 ) i s n o t s a t i s f i e d
by t h i s y k / 2 , e s t i m a t e d y1 i s a d j u s t e d i n t h e s a m e d i r e c t i o n t h a t yfi/2
would be a d j u s t e d t o s a t i s f y e q u a t i o n ( 3 ) . For u n e q u a l - l e v e l s p a c i n g and
M-odd, y1 i s z e r o and x1 i s e s t i m a t e d . E q u a t i o n ( 2 ) i s s o l v e d f o r y 2 ,
e q u a t i o n ( 3 ) i s s o l v e d f o r x2, and s o on, u n t i l y fi+1)/2 i s found and
t e s t e d i n equation ( 3 ) . I f t h e i n t e g r a l i s n o t suf i c i e n t l y c l o s e t o zero,
t h e e s t i m a t e d x1 i s a d j u s t e d i n t h e same d i r e c t i o n t h a t
would b e
a d j u s t e d t o s a t i s f y e q u a t i o n ( 3 ) . I f t h e o u t p u t l e v e l s are e q u a l l y s p a c e d ,
t h e x i and y i are i n t e g r a l m u l t i p l e s of h a l f t h e l e v e l s p a c i n g , and t h e
d i s t o r t i o n i s minimized w i t h r e s p e c t t o a s i n g l e p a r a m e t e r , t h e l e v e l s p a c i n g .
More d e t a i l i s g i v e n by Max ( r e f . 1 ) . Two computer programs w e r e developed t o
f i n d t h e equal- and u n e q u a l - l e v e l spaced q u a n t i z e r s . The programs u s e t h e
d i f f e r e n t i n p u t d i s t r i b u t i o n s and d i s t o r t i o n f u n c t i o n s d e s c r i b e d below.

INPUT DISTRIBUTIONS
Three i n p u t d i s t r i b u t i o n s - t h e normal, t h e two-sided e x p o n e n t i a l , and a
s p e c i a l i z e d gamma d i s t r i b u t i o n - a r e c o n s i d e r e d . The d i s t r i b u t i o n s , which
have z e r o mean and u n i t v a r i a n c e , a r e p l o t t e d i n f i g u r e 1. The normal d i s t r i bution is
2

1 -x*/2
p w = -e

&

(4)

The two-sided exponential or Laplacian distribution is (refs. 2 and 4 )

p(x)

1 -filxI
-e

fi

(5)

McDonald's specialized gamma distribution (refs. 2 and 5) is

These distributions are referred to below simply as the normal, exponential,

and gamma distributions.
The normal distribution is used most frequently in rate distortion
theory, but Paez and Glisson (ref. 2) refer to McDonald's (ref. 5) evidence
that speech amplitude variations can be modeled by the gamma distribution,
which is approximated by the mathematically simple exponential density. There
is also evidence that two-dimensional Hadamard transform coefficients of image
data have the exponential distribution (ref. 6). Experimental Hadamard transform coefficients (ref. 7) often have a gradually decreasing exponential slope
as x increases, and can be modeled by the gamma distribution, or by an even
more highly peaked distribution.
DISTORTION MEASURES
Three distortion measures - the mean-square error, the magnitude error,
and the relative magnitude error - are considered. The distortion for representative values of 0 and 5 2 . 0 and of cut points of k1.0 is plotted in
figure 2. The mean-square error function, which defines distortion when substituted in equation (l), is
d(x - y.)
1

(x - y.)2
1

(7)

The magnitude error function is

The relative magnitude error function is

The mean-square error distortion is used most frequently in rate distortion theory and in image data compression. The magnitude distortion criterion
and the relative magnitude distortion criterion were suggested for image data
3

by Andrews and Pratt (ref. 3 ) . The mean-square and magnitude distortion are
monotonically increasing as I x - yi I increases , but relative distortion does
not have this property; the property is used in the derivation of equation ( 2 ) .
For example, consider a representative value of 2 , as shown in figure 2 ,
with input values of 1 and 4 . Substituting in equation ( 9 ) , we have
14 - 21 > 11 - 2 1 , but 14 - 2 ) / 1 4 1 < 11 - 2 1 / 1 1 1 . In fact, the limit, as x
becomes infinite, of Ix - 21/1x1 is 1. Another problem with relative distortion is that zero is a cut point for M-even, with the representative values
symmetrical about zero. The distortion for small x is
x - y1 / 1x1 which
becomes infinite as x approaches zero. It is shown in the appendix that the
relative distortion criteria require one representative value to be zero, so
that the only sets of cut points and representative values that are symmetrical about zero have an odd number of representative values. Only the levels
for odd numbers of M are computed, but M-even can be used. For example,
to use four levels, the optimum quantizer would have representative values of
0, of the one negative level for M = 3 , and of the two positive levels for
M = 5, or positive and negative levels could be interchanged. It is also
shown in the appendix that, for M-odd, equation ( 2 ) is correct even though
the monotonic distortion requirement is not satisfied.

MINIMUM DISTORTION QUANTIZERS

The results of running the Max algorithm programs for the above input
distributions and distortion measures are given in tables 2 through 17; these
tables are described in table 1. The number of levels increases from the
lowest to highest, as indicated in parentheses, by adding 1; by multiplying
by 2 ; or by multiplying by 2 and adding 1. The arrangement of the tables folAll
lows Max (ref. 1) directly, and differs from Paez and Glisson (ref. 2).
the numbers in the tables have been rounded to four significant digits. The
final infinite cut points have not been included in tables 9 through 17. The
letter a next to most of the values for M = 1 and M = 2 indicates a value
derived by direct computation, rather than by the Max algorithm programs.
The equal and unequal quantizers for normal input distribution and meansquare error for M = 1 (1+) 36 are given by Max (ref. 1). The corresponding
values given here in tables 2 and 9 are usually identical, differing at most
by 5 units in the fourth place. The equal and unequal quantizers for exponential and gamma input distributions and mean-square error for M = 2 ( 2 X ) 32
are given by Paez and Glisson (ref. 2 ) . The corresponding values given in
tables 3, 10, and 11 differ slightly in most cases, and differ significantly
for the unequal M = 1 6 and M = 32 gamma distribution quantizers. Since
the same programs that obtained good agreement with Max were used, and since
the mean-square error distortions obtained are in every case less than those
given by Paez and Glisson (except for gamma, M = 2 , where directly computed
values are used), it appears that the values given here are more optimal.
The results given in the tables are partially plotted in the figures
discussed below. Figure 3 shows the optimum quantizer level spacing for normal, exponential, and gamma input distribution and for mean-square error.
4

This figure, like many of the other figures included here, shows intermediate
values not listed in any table. The level spacings for the exponential and
gamma distributions are wider, reflecting their higher probabilities for large
x. This also appears in figure 4 , where the largest representative value is
higher for the exponential and gamma distributions.
Figures 3 and 4 show also the similarity of the exponential and gamma
quantizers, and indicate that the gamma quantizers have narrower spacing for
small M-even than for small M-odd. This is due to the infinite value of
the gamma density at x = 0 , which requires a zero or small positive representative value. This effect, much reduced, also occurs for the exponential
distribution.
Figure 5 shows mean-square distortion for the optimum equal quantizers,
and figure 6 shows the distortion for odd and even M, gamma input, equal
quantizers. The distortion is significantly less for the normal input distribution because of its narrower spread and lower peak.
Figure 6 shows that, for the gamma distribution and equal spacing,
2* (integer)-1 levels provide better performance than 2* (integer) levels.

Figure 5 also includes the distortion for the normal input, optimum unequal
spacing quantizer, which is less than the distortion for the equal level quantizer. Figure 7 shows the distortion for all the minimum mean square error
unequal quantizers. As in the normal input case, the distortions for the
exponential and gamma inputs are less with unequal level spacing quantizers.
Although the normal input distribution again has least distortion, the differences are smaller because the optimum unequal quantizers adjust to the input
distribution shape. Figure 8 gives the largest representative value for
unequal-spacing, minimum mean-square error quantizers. Compared to the
largest representative values for equal spacing quantizers given in figure 4 ,
the largest representative value for unequal spacing quantizers increases more
rapidly with M until a final value is approached. The largest representative
values for odd and even M do not have different curves, as they did for
equal spacing quantization. In all cases for unequal spacing quantizers,
increasing the number of levels decreases distortion.
Figure 9 shows the optimum equal spacing for magnitude distortion and
figure 11 shows the optimum equal spacing for relative distortion. The spacing is smaller for magnitude distortion, and smaller still for relative distortion, because of the reduced weighting of large magnitude errors, especially
for the representative value interval that extends to infinity. The gamma
distribution for magnitude distortion again has narrower spacing for an even
number of levels. A s discussed above and in the appendix, only odd numbers of
levels are used with relative distortion. As in the mean square error case,
the normal input quantizers have the smallest spacing and the g a m a input
quantizers have the largest spacing. Figure 10 gives the distortion for minimum magnitude error, equal-spacing quantizers. The distortion for the gamma
distribution for odd and even numbers of levels is given in figure 6. A s in
the mean-square error case, distortion is lower for odd numbers of levels.
The distortion for minimum relative error equal spacing quantizers is given in
figure 12. Again, as in the case of mean-square error, the equal quantizers
have least distortion for a normal input and have most distortion for a gamma
5

i n p u t . The optimum unequal q u a n t i z e r s f o r magnitude and r e l a t i v e magnitude

d i s t o r t i o n are n o t s e p a r a t e l y p l o t t e d , b u t a r e d i s c u s s e d i n t h e n e x t s e c t i o n .
Max ( r e f . 1) i n d i c a t e d t h a t , f o r M-large, i n c r e a s i n g t h e number of levels
t o 2M would, as a n a p p r o x i m a t i o n , c a u s e e a c h p r e v i o u s r e p r e s e n t a t i v e l e v e l t o
b e d i v i d e d i n t o two e q u a l i n t e r v a l s ; f i g u r e s 3, 9 , and 11 confirm t h i s . For
a p p r o x i m a t e l y c o n s t a n t p r o b a b i l i t y d e n s i t y i n each i n t e r v a l , t h e mean s q u a r e
e r r o r i s reduced by o n e - f o u r t h ; f i g u r e s 5 , 6 , and 7 c o n f i r m t h i s . S i m i l a r l y ,
d o u b l i n g a h i g h number of l e v e l s would r e d u c e t h e magnitude e r r o r by o n e - h a l f ,
as shown i n f i g u r e s 6 and 10. The r e l a t i v e d i s t o r t i o n ( f i g . 1 2 ) i s less w e l l
behaved, a l t h o u g h t h e d i s t r i b u t i o n s f o r normal and e x p o n e n t i a l i n p u t s n e a r l y
f o l l o w t h e one-half s l o p e .
Although t h e minimum d i s t o r t i o n q u a n t i z e r i s e x a c t l y d e f i n e d , t h e d i s t o r t i o n can approach t h e minimum f o r s i g n i f i c a n t l y d i f f e r e n t q u a n t i z e r s . For t h e
M = 38 and M = 40 q u a n t i z e r s i n t a b l e 9 , d i s t o r t i o n i s 10 p e r c e n t l a r g e r
f o r M = 38. I f a n M = 40 q u a n t i z e r i s d e s i g n e d u s i n g t h e M = 38 q u a n t i z e r
w i t h any two a d d i t i o n a l r e p r e s e n t a t i v e v a l u e s , t h e d i s t o r t i o n must b e less
t h a n t h e M = 38 d i s t o r t i o n , and w i t h i n 10 p e r c e n t of t h e minimum. T h i s
i m p l i e s t h a t t h e form of t h e minimum d i s t o r t i o n q u a n t i z e r i s l e s s a c c u r a t e l y
d e f i n e d t h a n i t s performance, and t h a t c o n v e n i e n t a p p r o x i m a t i o n s t o t h e quant i z e r s , o r even g r e a t l y d i f f e r i n g q u a n t i z e r s , w i l l o f t e n perform a c c e p t a b l y .

PERFORMANCE OF THE MINIMLTM DISTORTION QUANTIZERS

It i s w e l l known t h a t , f o r a g i v e n d i s t o r t i o n , t h e normal i n p u t d i s t r i b u t i o n r e q u i r e s a h i g h e r minimum t r a n s m i s s i o n r a t e t h a n any o t h e r z e r o mean,

u n i t v a r i a n c e d i s t r i b u t i o n ( r e f . 4 , pp. 101-102).
However, t h i s rate d i s t o r t i o n bound d e f i n e s o n l y t h e minimum rate f o r a l l p o s s i b l e t r a n s m i s s i o n methods.
Simply u s i n g a q u a n t i z e r w i t h a n i n t e g e r number of e q u a l l y spaced l e v e l s f o r
each sample i s n o t t h e b e s t t r a n s m i s s i o n method, and i t i s r e l a t i v e l y l e s s
e f f i c i e n t f o r t h e h i g h l y peaked e x p o n e n t i a l and gamma d i s t r i b u t i o n s . Huffman
coding ( r e f . 8, Ch. 2 ) can b e used t o r e d u c e t h e t r a n s m i s s i o n r a t e from
log, M t o t h e q u a n t i z e r e n t r o p y , a lower b u t v a r i a b l e r a t e . A p l o t o f t h e
e q u a l l e v e l s p a c i n g q u a n t i z e r d i s t o r t i o n v e r s u s e n t r o p y ( f i g . 1 3 ) shows t h a t
t h e normal d i s t r i b u t i o n does r e q u i r e h i g h e r r a t e ( e x c e p t a t s m a l l M) f o r a
t r a n s m i s s i o n system c o n s i s t i n g of an e q u a l l y spaced q u a n t i z e r and a Huffman
c o d e r . The optimum unequal q u a n t i z e r h a s l e s s d i s t o r t i o n t h a n t h e e q u a l quant i z e r . Although t h e unequal q u a n t i z e r f o r t h e normal i n p u t d i s t r i b u t i o n a g a i n
h a s l o w e s t d i s t o r t i o n , t h e d i f f e r e n c e i n d i s t o r t i o n ( f i g . 7 ) i s less t h a n f o r
t h e e q u a l q u a n t i z e r . A p l o t of t h e unequal l e v e l s p a c i n g q u a n t i z e r d i s t o r t i o n
v e r s u s e n t r o p y ( f i g . 1 4 ) shows t h a t t h e normal d i s t r i b u t i o n r e q u i r e s a t r a n s m i s s i o n r a t e e q u a l t o o r g r e a t e r t h a n t h e o t h e r d i s t r i b u t i o n s , a t medium and
l a r g e M.

The equally-spaced and unequally-spaced minimum d i s t o r t i o n q u a n t i z e r s can

each b e used w i t h e n t r o p y coding, g i v i n g f o u r p o s s i b l e systems f o r each comb i n a t i o n of i n p u t d i s t r i b u t i o n and d i s t o r t i o n measure. The performance of t h e
systems l i s t e d i n t h e t a b l e s i s shown i n f i g u r e s 1 5 through 23. The r a t e
d i s t o r t i o n bound, o r a lower bound on t h i s bound, i s g i v e n f o r q u a n t i z e r s

designed for mean-square and magnitude error (ref. 4 , pp. 92-102, 141). In
figure 15, the distortion is lower for entropy-coded, equal-spaced quantizers
than for entropy-coded, unequally-spaced quantizers. This result was found by
Wood (ref. 9), who used approximations for entropy and distortion based on the
work of Ma.x (ref. 1). Goblick and Holsinger (ref. lo) noted earlier that the
entropy-coded, equal-spaced quantizers were within 0.25 (here 0 . 3 ) bits of the
rate distortion bound.
Figures 16 and 17 indicate that the cases of exponential and gamma input
and mean-square error are similar. Entropy-coded, equal quantizers give the
best performance, and they approach the rate distortion bound. Although these
highly peaked distributions give poor performance with uncoded, equal-spaced
quantizers, the low probability of the higher representative levels allows
larger rate reduction with entropy coding. For the gamma input and mean-square
error, entropy-coded (variable rate), equal quantizers are 0.5 to 1.5 bits
better than unequal quantizers. The rate reduction is comparable to the
reduction obtained using variable rate adaptive Hadamard image compression
(ref. 11), and these two variable rate methods can be combined.
Figures 18, 19, and 20 give the rate distortion bounds and quantizer performance, with and without entropy coding, for the magnitude distortion measure. As in the case of mean-square distortion, the entropy-coded, equalspaced quantizers have lowest distortion, approaching the bound, and are
superior to uncoded, unequal quantizers, except at some small M.
Figures 21, 22, and 23 give the quantizer performance, with and without
entropy coding, for the relative distortion measure. The lower bound on the
rate distortion cannot be found using the method for difference distortion
measures, because the relative distortion is a function of the sample value as
well as the error (ref. 4 , p. 92). For the relative error, the unequallyspaced quantizers give significantly less distortion than entropy-coded, equal
quantizers.
For all three distortion measures, the performance differences increase
greatly for the exponential and gamma input distributions.
SUMMARY AND CONCLUSION

The results of the work include the structure of the new quantizers, the
performance of the quantizers, and certain properties of the quantizers due to
input distribution or distortion properties, as follows.

1. The minimum mean-square error, equal- and unequal-spaced quantizers

for normal, exponential, and gamma input distributions, with the distortion
and entropy, were found for new numbers of levels.
2. The magnitude error quantizers for normal, exponential, and gamma
input distributions, and the distortion and entropy, were found.

3 . The relative magnitude error quantizers for normal, exponential, and

gamma input distributions, and the distortion and entropy, were found.
4 . For exponential and gamma distributions and mean-square error, it was
shown that equally-spaced quantizers with entropy coding are far superior to
unequally-spaced quantizers with entropy coding, and that they approach the
rate distortion bound. (These results were shown for the normal input distribution and mean-square error by Wood (ref. 9) and by Goblick and Holsinger
(ref. l o ) , who used the work of Max (ref. 1)).

5. For magnitude error and small M y unequal-spacing quantizers with

entropy coding sometimes have slightly less distortion than entropy-coded,
equally-spaced quantizers. The entropy-coded, equal-spaced quantizers are
superior for medium and large M, and approach the rate distortion bound.
6. For relative magnitude error, unequal quantizers have significantly
less distortion than entropy-coded, equally-spaced quantizers.
7. The rate reduction for entropy-coded, equally-spaced quantizers is
significantly larger for exponential or gamma input distributions than for the
normal input distribution.

Equally-spaced quantizers with odd numbers of levels are superior to

8.
equally-spaced quantizers with even numbers of levels for the gamma input distribution. Gamma distribution quantizers usually have a representative level
either equal to, or very close to, zero.

9. The exponential input distribution has a similar but much smaller

superiority of the equally-spaced quantizers with odd numbers of levels.
10. The relative magnitude distortion criterion forces one representative
level to be zero.

Ames Research Center

National Aeronautics and Space Administration
Moffett Field, California 95035, August 15, 1976

APPENDIX
RELATIVE MAGNITUDE DISTORTION

Suppose t h a t M i s even; t h e c u t p o i n t x1 i s z e r o and t h e d i s t o r t i o n

f o r t h e f i r s t r e p r e s e n t a t i v e v a l u e , from e q u a t i o n (1) i s :

D1 = 2

p ( 4 dx

i s g r e a t e r t h a n y1 and i s i n f i n i t e f o r M = 2. To minimize t h e d i s t o r t i o n , t h e d e r i v a t i v e w i t h r e s p e c t t o y1 i s set t o z e r o as i n e q u a t i o n ( 3 )

h e r e , and i n e q u a t i o n s ( 2 ) , ( 4 ) , and ( 6 ) i n r e f e r e n c e (1
Removing t h e absol u t e v a l u e by u s i n g two i n t e g r a l s i n t h e r a n g e s 0 t o y1 and y1 t o x2 and
using L e i b n i t z s r u l e ,

x?

Suppose p ( x )
t i n u o u s a t x = 0.
dD1
-= 2
dy 1

i s m o n o t o n i c a l l y d e c r e a s i n g as
For s m a l l p o s i t i v e E,
E

x-p(o)dx

lY1

i n c r e a s e s and i s con-

- 2

x-p(x)dx

x-p(x)dx

Y1
X

XI,

2p(o)~n

x-p(x)dx

x-lp(x)dx

Y1

The Rn E i s a l a r g e f i n i t e n e g a t i v e number, t h e second i n t e g r a l i s p o s i t i v e ,

and t h e t h i r d i s bounded as f o l l o w s :

W e t h e r e f o r e have
1

- > 2p(o)[m]
dy 1

2p(o)Rn

(A5 1

Y1

S i n c e Rn E i s f i n i t e , t h e d e r i v a t i v e of D1 w i t h r e s p e c t t o y1 i s
i n f i n i t e f o r a l l nonzero y l .
D1 i s l a r g e f o r y1 n o t e q u a l t o z e r o and
i n c r e a s e s as y1 i n c r e a s e s , s o t h a t y1 i s f o r c e d t o z e r o by t h e b e h a v i o r of

dDl/dyl. For the case where M is odd, y1 is zero and the relative distortion is always 1 or less, since y1 = 0 may be used in equation ( 9 ) . The
other values of yi are used when they are closer to x and reduce the
distortion.
The assumption that p(x)
is continuous at zero is not satisfied for the
gamma distribution, but if we take p(x) = p'(x)/w,
is continuous
p'(x)
and, as in equation (A3)

1:

dD1

-= 2
dyl

+...

(o)dx

2p1(0)[-(2/5)x-2/51~l

(4/5)p1(0>[-e-2/5

Bounding the third term of equation (A3) by

derivative again forces y1 to zero.

+ ...

m]

(A6

...

y73/2, the infinite positive

Even for M-odd, the relative magnitude distortion is not a monotonically

increasing function of Ix - yil , as shown in figure 2 and mentioned in the
text above. In reference 1, Max uses the monotonically increasing property to
prove equation (2) above, which is used in the algorithm. Equation (2) can be
shown directly, in a manner parallel to that of Max (Max eqs. (1) and (5)).
From equations (1) and ( 9 )

M
D = C
Setting
give

dD/Dxi

rrxi+l Ix - Y-

I
(A7

equal to zero, by Leibnitz's rule, the

For p(xi) # 0 and ]xi1 # 0, which is true for M-odd

can equal the zero representative value,
Ixi - yi-i
Since ~ i - xi,
~ , and yi

= [Xi

i - 1 and

terms

because no cut point

Yil

are increasing positive values,

which is equation (2).

We have shown that the relative magnitude distortion requires a representative value of zero, and that for M-odd, which implies a zero representative
value, the relative magnitude distortion can be treated by the Max algorithm.

10

REFERENCES

1. Max, Joel: Quantizing for Minimum Distortion, IRE Trans. Inform. Th.,
vol. IT-6, March 1960, pp. 7-12.
2.

Paez, M. D.; and Glisson, T. H.: Minimum Mean-Squared-Error Quantization

in Speech PCM and DPCM Systems. IEEE Trans. Comm., vol. COM-20,
April 1972, pp. 225-230.

3.

Andrews, H. C.; and Pratt, W.: Image Transforms, Ch. 6 in Computer Techniques in Image Processing, H. C. Andress, ed., Academic Press, 1970.

4.

Berger, T.: Rate Distortion Theory:

pression. Prentice-Hall, 1971.

5.

McDonald, R. A.: Signal-to-Noise and Idle Channel Performance of Differential Pulse Code Modulation Systems - Particular Applications to
Voice Signals. Bell System Tech. J., v o l . 45, Sept. 1966,
pp. 1123-1151.

6.

Fukinuki, Takahiko; and Miyata, Masachika: Intraframe Image Coding by

Cascaded Hadamard Transforms. IEEE Trans. Comm., v o l . COM-21,
Mar. 1973, pp. 175-180.

7.

Cotton, M. C.: Image Processing Using Three-Dimensional Hadamard Transforms. Ph.D. Dissertation, Stanford Univ., 1977.

8. Ash, R.:

a Mathematical Basis for Data Com-

Information Theory. Interscience, New York, 1965.

9. Wood, Roger C.: On Optimum Quantization.

v o l . IT-15, Mar. 1969, pp. 248-252.

IEEE Trans. Inform. Th.,

10. Goblick, T. J., Jr.; and Holsinger, J. L.: Analog Source Digitization:
A Comparison of Theory and Practice. IEEE Trans. Inform. Th.,
v o l . IT-13, April 1967, pp. 323-326.
11. Wintz, Paul A . : Transform Picture Coding.
July 1972, pp. 809-820.

Proc. IEEE, v o l . 60,

11

TABLE 1.- LIST OF MINIMUM DISTORTION QUANTIZER TABLES

.

..

.~ . . .

Input
distribution

Distortion
criterion

Normal, exponential,
g-

Mean square

1(1+)40

Mean square

2 (2x) 2048

Magnitude

1(1+)40

Magnitude

2 (2x) 2048

Relative
magnitude

1(1+)40

Relative
magnitude

2 (2x) 2048

'ab1ea

Spacing

Number of
levels, M

Gamma

Mean square,
magnitude

1(Zx,1+)51:

Normal

Mean square

1(1+)40,
64(2x) 251

10

Exponent,a1

Mean square

11

Gamma

Mean square

1(1+)40,
64(2x)25(

12

Normal

Magnitude

2(2x)16

13

Exponential

Magnitude

2 (2x)16

14

Gamma

Magnitude

1(2x,1+)15

15

Normal

Relative
magnitud e

1(2x ,1+)31

16

Exponential

Relative
magnitude

1(2x,1+)31

Gamma

Relative
magnitude

1(2x ,1+)31

17

'

1(1+)40,
64(2x) 25(

a
Tables 2-8 contain the spacing, distortion, and entropy for each number
of representative values. Tables 9-17 contain the cut points and representative values, distortion, and entropy for each number of representative values.
The probability and distortion for each representative interval are also
given.

12

TABLE 2.- EQUIDISTANT QUANTIZATION LEVEL SPACING FOR MEAN SQUARE DISTORTION

Normal PDF

Exponential PDF

M
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

Gamma PDF

Spacing

Distortion

Entropy

Spacing

Distortion

Entropy

Spacing

Distortion

Entropy

0.0000
1.5960a

1.ooooa
.3634a
.1902
.1188
.08218
.06066
.04686
.03744
.03070
.02569
.02186
.01885
.01645
.01450
.01289
.01154
.01040
.009434
.008598
.007873
.007239
.006682
.006189
.005751
.005360
.005008
.004692
.004406
.004146
.003909
.003693
.003495
.003313
0.003146
.002991
.002848
.002715
.0.02592
.002477
.002370

0. 0ooa

0.0000
1.4140a

1.OOOOa

0.oooa
1.oooa
1.324
1.752
1.872
2.127
2.213
2.394
2.468
2.608
2.670
2.783
2.842
2.937
2.991
3.074
3.118
3.191
3.236
3.301
3.344
3.402
3.437
3.491
3.527
3.576
3.612
3.657
3.690
3.733
3.759
3.805
3.829
3.866
3.895
3.930
3.958
3.991
4.018
4.049

0.0000
1.1290a
1.8170
1.0790
1.3070
.9175
1.0420
.7926
.8757
.6990
.7593
.6245
.6726
.5649
.6052
.5318
.5509
.4837
.5062
.4565
.4687
.4192
.4366
.3939
.4089
.3716
.3847
.3519
.3633
.3342
.3442
.3182
.3271
0.3033
.3118
.2899
.2978
.2774
.2851
.2663

1.ooooa

0.oooa
1.oooa
1.337
1.754
1.846
2.058
2.133
2.268
2.339
2.433
2.504
2.571
2.642
2.697
2.763
2.779
2.870
2.890
2.967
2.964
3.055
3.063
3.137
3.139
3.214
3.211
3.285
3.278
3.352
3.341
3.416
3.402
3.476
3.461
3.533
3.517
3.588
3.558
3.640
3.617

1.2240
.9957
.8430
.7334
.6508
.5860
.5338
.4908
.4546
.4238
.3972
.3739
.3534
.3352
.3189
.3042
.2909
.2788
.2678
.2576
.2482
.2396
.2315
.2241
.2171
.2105
.2044
.1987
.1935
.1881
.1833
0.1788
.1744
.1703
.1664
.1627
.1591
.1557

1.oooa

1.536
1.904
2.183
2.409
2.598
2.761
2.904
3.032
3.148
3.253
3.350
3.440
3.524
3.602
3.676
3.746
3.811
3.874
3.933
3.990
4.045
4.097
4.146
4.194
4.241
4.285
4.328
4.370
4.408
4.449
' 4.487
4.524
4.559
, 4.594
4.628
4.661
4.693
4.724

1.4140
1.0850
1.0240
.8688
.8209
.7295
.6921
.6309
.6034
.5594
.5362
.5028
.4837
.4573
.4431
.4218
.4083
.3906
.3790
.3640
.3554
.3423
.3337
.3226
.3147
.3050
.2980
.2892
.2842
I .2751
.2708
0.2639
.2586
.2624
.2476
.2419
.2374
.2323

.5000a

.2634
.1931
.1322
.lo79
.08249
.07075
.05704
.05020
.04267
.03834
.03306
.03007
.02645
.02428
.02206
.02044
.01848
.01723
.01573
.01474
.01380
.01278
.01205
.01140
.01062
.01007
.00943
.008969
.008592
I .008040
'
.007741
0.007405
.007012
.006721
.006382
.006128
.005834
.005611

aValue derived by direct computation, not by the Max algorithm.

'

.6816a

.2761
.2772
.1444
.1673
.09196
.11160
.06481
.08033
.04861
.06022
.03804
.04687
.03071
.03855
.02538
.03127
.02138
.02662
.01828
.02315
.01583
.02002
.01386
.01749
.01224
.01542
.01090
.01370
.009769
.012260
.008811
0.01099
.007991
.009914
.007282
.008980
.006666
.008184

TABLE 3.- EQUIDISTANT QUANTIZATION LEVEL SPACING FOR MEAN SQUARE DISTORTION
Gamma PDF

Exponential PDF

Normal PDF
M
Spacing
2
4
8
16
32
64
128
256
512
LO24
!048

Distortion

1.5960a 0.3634a
.1188
,9957
.03744
.5860
.01154
.3352
.003495
.1881
.001040
.lo41
.0003043
.05687
.03076
.00008769
,00002492
,01650
.008785 .000006997
.004650 .000001944

Distortion

Entropy

Spacing

1.0ooa
1.904
2.761
3.602
4.449
5.309
6.182
7.069
7.968
8.878
9.796

1.4140a 0.5000a
1.0850
.1931
.07075
.7295
.02428
.4573
.2751
.008040
.002574
,1607
,09018
.0007638
,0002145
.04876
.02569
.00005714
.01323
.00001491
.006724 .000003808

Entropy

Spacing

1.0ooa

1.1290a 0.6816a
.2772
1.0790
.1116
.7926
.5318
.03855
.3182
.01226
.1851
.003574
.09923
.0009703
.05205
.0002476
.02662
.00006318
.00001586
.01349
.006794 .000003952

1.752
2.394
3.074
3.805
4.580
5.413
6.300
7.225
8.182
9.159

aValue derived by direct computation, not by the Max algorithm.

Distortion

Entropy

1.0ooa
1.754
2.268
2.779
3.402
4.085
4.897
5.763
6.682
7.628
8.593

TABLE 4.-

EQUIDISTANT QUANTIZATION LEVEL SPACING FOR MAGNITUDE DISTORTION

Normal PDF

Gamma PDF

Exponential PDF

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40

cn

Spacing

Distortion

Entropy

Spacing

Distortion

Entropy

Spacing

Distortion

Entropy

0.0000
1.3490a
1.0270
.8338
.7044
.6116
.5416
.4869
.4428
.4065
.3760
.3500
.3276
.3081
.2909
.2756
.2619
.2497
.2385
.2284
.2192
.2107
.2029
.1957
.1890
.1827
.1769
.1715
.1664
.1616
.1571
0.1529
.1489
.1451
.1415
.1381
.1349
.1318
.1289
.1261

0.7979a
.4754
.3406
.2683
.2217
.1900
.1662
.1482
.1337
.1221
.1123
.lo42
.09703
.09100
.08558
.08092
.07666
.07294
.06950
.06646
.06362
.06109
.05870
.05655
.05451
.05267
.05091
.04931
.04778
.04638
.04502
0.04378
.04258
.04148
.04040
.03941
.03845
.03755
.03668
.03587

0.0ooa
1.oooa
1.574
1.976
2.278
2.523
2.728
2.905
3.059
3.197
3.322
3.435
3.539
3.635
3.724
3.807
3.886
3.960
4.029
4.095
4.158
4.218
4.275
4.330
4.383
4.433
4.482
4.529
4.574
4.618
4.660
4.701
4.741
4.779
4.816
4.853
4.888
4.922
4.956
4.988

0.0000
9803a
.9776
.7579
.7070
.6030
.5643
.5029
.4743
.4333
.4115
.3821
.3649
.3426
.3287
.3112
.2997
.2857
.2760
.2643
.2560
.2462
.2391
.2306
.2244
.2171
.2116
.2053
.2004
.1947
.1903
0.1853
.1814
.1769
.1733
.1692
.1659
.1623
.1593
.1559

0.7071a
4910a
.3540
.2993
.2459
.2200
.1914
.1760
.1582
.1477
.1354
.1278
.1189
.1130
.lo62
.lo16
.09613
.09247
.08797
.08485
.08119
.07855
.07546
.07319
.07055
.06858
.06629
.06456
.06255
.06102
.05925
0.05788
.05630
.05507
.05361
.05250
.051.23
.05022
.04907
.04815

0.0ooa
1.0ooa
1.503
1.927
2.151
2.404
2.552
2.732
2.845
2.984
3.077
3.191
3.270
3.365
3.435
3.517
3.579
3.651
3.707
3.772
3.822
3.881
3.927
3.981
4.023
4.073
4.113
4.158
4.195
4.238
4.273
4.312
4.346
4.383
4.414
4.449
4.479
4.512
4.540
4.572

0.0000
5253a
1.1310
.6085
.8214
.5347
.6583
.4708
.5554
.4204
.4827
.3798
.4296
.3474
.3883
.3207
.3552
.2978
.3273
.2787
.3045
.2622
.2850
.2474
.2677
.2347
.2531
.2230
.2397
.2129
.2282
0.2035
.2175
.1953
.2083
.1875
.1995
.1806
.1917
.1740

0.5774a
.4818
.3141
.3300
.2295
.2520
.1841
.2063
.1554
.1757
.1353
.1528
.1198
.1363
.lo83
.1233
.09917
.1127
.09107
.lo35
.08482
.09627
.07948
.08959
.07443
.08428
.07044
.07920
.06653
.07513
.06343
0.07113
.06031
.06791
.05782
.06467
.05525
.06206
.05321
.05937

0.oooa
1.0ooa
1.449
1.923
2.023
2.313
2.363
2.576
2.608
2.777
2.803
2.942
2.964
3.082
3.102
3.204
3.222
3.313
3.331
3.411
3.428
3.499
3.516
3.582
3.599
3.657
3.674
3.729
3.746
3.795
3.811
3.858
3.874
3.916
3.932
3.972
3.989
4.025
4.042
4.076

aValue derived by direct computation, not by the Max algorithm.

TABLE 5.- EQUIDISTANT QUANTIZATION LEVEL SPACING FOR MAGNITUDE DISTORTION

Normal PDF

Exponential PDF

Distortion

Spacing

0.4754
1.3490a
.2683
,8338
.1482
8
.4869
,08092
,2756
16
.04378
,1529
32
.02351
.08361
64
128
.01254
.04521
.006646
.02422
256
.003505
.01288
512
1024 j
.006810 I .0018390
2048
,003581
.0009617
2
4

1.0ooa
1.973
2.905
3.807
4.701
5.596
6.496
7.404
8.320
9.242
10.170

Distortion

Spacing

Entropy

..

0. 4910a
0. 9803a
.2993
.7579
,1760
,5029
.lo16
.3112
.05788
.1853
,03253
.lo75
,01806
,06123
.009913
.03434
.005374
,01900
.010390
.002881
,005611
,001524

1
~

Gamma PDF
~~

Entropy

Spacing

Distort ion

Entropy

1.0ooa

0. 5253a

0.4818
.3300
.2063
.1233
,07113
.04028
.02220
.01200
.006343
.003289
.001683

1.0ooa
1.923
2.576
3.204
3.858
4.550
5.287
6.069
6.894
7.763
8.673

1.927
2.732
3.517
4.312
5.124
5.951
6.795
7.653
8.527
9.418

.6085
.4708
.3207
.2035
.1235
.0723
.04105
.02269
.012210
.006408

.
I

aValue derived by direct computation, not by the Max algorithm.

TABLE 6.-

EQUIDISTANT QUANTIZATION LEVEL SPACING FOR RELATIVE DISTORTION

Normal PDF
M

1
3
5
7
9
11
13
15
17
19
21
23

Exponential PDF

G a m a PDF

I'

Spacing

Distortion

Entropy

Spacing

Distortion

Entropy

Spacing

0.0000

1.OOOOa

0. 0ooa
1.557
2.282
2.763
3.121
3.405
3.641
3.843
4.019
4.174
4.314
4.441
4.557
4.664
4.763
4.855
4.942
5.023
5.099
5.172

0.0000
,4260
.3103
.2499
.2118
.la52
,1653
.1498
.1374
.1271
,1185
.1111
,1047
.0991
,09414
.08972
.08575
.08217
.07891
.07593

1.ooooa

0.0ooa

0.0000
,2989
.2182
.1762
.1497
.1312
.1174
.lo66
.09790
.09073
.08469
07952
.07504
.07110
,06762
.06452
.06172
,05920
.05690
,05480

.6203
.4338
.3391
.2809
.2410
.2118
.la94
,1717
.1572
.1452
.1350
.1262
.1186
.1119
.lo60

.loo8
.09602
.09174
.08786

.5569
.4128
.3362
.2873
.2529
.2270
.2068
,1904
.1768
.1653

.1555
.1469
.1394
.1328
.1268
.1215
.1166
.1122

.lo81

.6147
.4791
.4035
.3536
.3174
.2896
.2675
.2493
.2340
.2209
.2096
.1997
.1908
.la29
.1758
,1694
.1635

,1581
.1531

aValue derived 'by direct computation, not by the Max algorithm.

1.567
2.273
2.730
3.070
3.339
3.563
3.754
3.922
4.070
4.203
4.324
4.435
4.537
4.631
4.720
4.802
4.880
4.954
5.023

Distortion

.3518
.3424
.3339
.3261
.3189
.3122
.3061
.3003

Entropy

4.072
4.163
4.248
4.328
4.403
4.474
4.541
4.604

TABLE 7.-

EQUIDISTANT QUANTIZATION LEVEL SPACING FOR RELATIVE DISTORTION

Normal PDF

Exponential PDF

Spacing

Distortion

Entropy

Spacing

Distortion

Entropy

1
3
7
15
31
63
1 27
255
511
,023
:047

0.0000
.6203
,3391
.1894
.lo60
.05898
.03252
.01777
.009627
.005175
.002763

1.OOOOa

0. 0ooa
1.557
2.763
3.843
4.855
5.829
6.782
7.721
8.653
9.581
10.510

0.0000
.4260
.2499
.1498
C8972
.05321
.03117
.01802

1.OOOOa
.6147
.4035
.2675
.1758
,1138
.07243
,04530
.02788
.01690

0. 0ooa

.5569
.3362
,2068
,1268
.07698
.04612
.02726
.01591
.009160
,0052390

1.567
2.730
3.754
4.720
5.654
6.568
7.469
8.359
9.243
10.120

aValue derived by direct computation, not by the Max algorithm.

Gamma PDF

0.0000
.2989
.1762
.lo66
.06452
.03870
.02292
.01339
.00772
.004400
.002480

Distortion

Entropj

1.OOOOa
.7942
.5399
.4197
.3261
.2519
.1931
.1469
.1109
.08321
.06205

0. oooa
1.585
2.595
3.472
4.328
5.189
6.063
6.949
7.844
8.745
9.650

TABLE 8.- EQUIDISTANT QUANTIZATION LEVEL SPACING FOR GAMMA PDF

Magnitude d i s t o r t i o n

Mean s q u a r e e r r o r
M

1
3
7
15
31
63
127
255
511

Spacing

Distortion

Entropy

Spacing

Distortion

Entropy

0.0000
1.8170
1.0420
,6052
.3442
.1882
.lo14
.05289
.02678

1.OOOOa
.2761
.09196
.03071
.009769
.002955
.0008294
,0002221
.00005784

0.000
1.337
2.133
2.763
3.416
4.147
4.933
5.789
6.708

0.0000
1.1310
,6583
.3883
.2282
,1322
.07533
.04205
.02300

0. 5774a
.3141
.1841

0. oooa
1.449
2.363
3.102
3.811
4.535
5.288
6.078
6.907

.lo83
.06343
.03636
.02043
,01120
.006015

aValue d e r i v e d by d i r e c t computation, n o t by t h e Max a l g o r i t h m .

19

TABLE 9.- OPTIMTJM QUANTIZATION LEVEL SPACING FOR NORMAL PDF W I T H

MEAN SQUARE DISTORTION

a
Value derived by direct computation, not by the Max algorithm.

20

105500000

01631000

00328500

e3H23000
1024400110

Y ( I )
000000UO
07646000

001300000

10724000tl

PtI)
0297800Q
2 4 44000
0 1067000

00~42200
00144300
0 184300

.9816(300

5
X(1)

!I.

6
1
2

s
UISTBRTIBN
EhUTHOPY

rl

D(I)

X( I
Y ( I )
03177000
000000u0
06569Cli10
1~000000U
.8940c100
i . 4 4 7 r ~ ~i ~
~

P(I)

02450000

.iwm~o

00739600

D(I)
00087180
00088770
0 1 13900

e05798L)Cl
2,4630000

7
X ( I )

2
3
4

UISTUHTIBN =
ENIHBPY =

*2c)03000
e874AOOO

1.6110000
0000ROUO
0

Y ( I )
oOU00000

e5606000
1.188000U
200330000

P(I)

ntr,

02207000

00057200

iO87000
.i373000
00536200

00058560
00075430

0057410

U 4 4 0Q 0 f l

2.647OOOU

21

ENfeBPY

22

3.1250000

3.2540000

0000UUC~O

205650000

0 0134000

CIO 1 3 7 2 0

23

DIS'IURTIUI~=

M =

0 140600

14

X(1)
o0000000
o2935000

1
2

e5959000
e9180000
1e277WOUO

3
4

5
6
7

DlSlRHT1Cflu

1 0703UUOU
20282UUUO

o n

Y ( I )
lA57000
e44130013
e7505000
lo0860000
e

1.468000Q
1 e93900O0
2.6250000

P(1)
1154000
a 1089000
90963300

00784200
a0566000

00330300
00212500

. O O O A ~ Q ~

e0008360
00008553
00011140

122300

P(1)
laHPOO0
e lU62000
00982800
e0855100

U(I)
e0006790
000Gh791

00006795
oOC106805

00686Y140

e0006825

0049OQ00
00263000

00006873
00007033
a0009168

e0095480

24

D(I)
e0008263
e0008267
e0008278

.2stj2n~0

.388ooon

.65680UO

52240UU
07995000
1=O9QOQW
104370000
1 0843UOUC)

204f)lU000

S
6

.0QCl5650
00005654

09023000

.09742ao
.OAA7200
e 076 1600

102S600LlO

00604800

lo6180000
2-0690000
207320000

00427100
02445uu
oOO81810

nO005679
.~a05720
.o~as854
a0007639

.a005662

PlI)
.096700a
00948200

00632300
00801800
.06HC)700

a0535500
mO37ABUO

002 1 2 8 U U
00070660

25

9
UISTCrHTIBh
EhTRbPY

M =

e0005464

3,9280000

19
X ( I )
1092000

V(I)

00000000
03294000
02184000
05551000
44040110
m6698000
07907000
10C7420r100
9 117000
103180000
1o 1 7 3 0 0 0 0
1 0 6 3 4 0 0 ~ ~1 0 4 6 4 0 0 0 0
200170000
1m8Q30000
20550LI000
20 23100QO
208680000
0000C1000
0

2
3
4

5
6
7
A
9
10

1030000
3 1280OO

DiSTBHTIBh

D(1)

P(1)
e0869600
e0855900
e 0 8 1 5 100

e0748700
e0659000
e0549500
00425300
00293500

00564400
e0053890

000345 1

0603452
e0003453
e 0003454
00003457
e0003463
e0003474

oQ003499
e0003582

00004682

4,0030000

Y(1)

26

e0061510

eUU75930

ENThWPY

2082600C10

205U30000

052650110
e7485000
e9837000
1a 2 3 8 0 0 0 0
1o S 2 3 O O O O
10857000O
2.27900130
209080000

00062080

P(I)
e0824900
e01301300
.0754wo

00666800
e0599500
e0496200
00381500
e 0 2 6 17 0 0
01 4 5 7 0 0
e0047520

d(11

e0082978
e0002978
eOU02979
e0002941

00002983
e0002988
e0002997
00003020

.OUQ~OW
e0004043

M a

21

1
2
3

4
5
6

7
0

9
10

11

M =

X ( I )
e0991AUO

02Q89000
e5026030
e7137000
e9360000
le1750000
1.4390000
1.7430000
2e1150000
206340nOQ

ooooouo

Y(1)

rOOOOaU0
1984000

e3994000
a605900U
e 82 14000
1e0510000

1.3OOOOOU
1 e57900OO
1 e9070000
2e3230000
2 0 9460000

P(1)
e07900UO
oQ779700
00749100
.0699000

U(1)
eOQa2587
00002587
e0002588

e05467tYQ

e0002588
00002590
,0002592

e0449600

00002596

.a343700
a0234400
e0129900
e0042130

.0a026m

e0630800

eOOUZ'624

eOUU2687
e0003516

22

1
2
3
4

5
d

7
0

10

11

27

P(I)
e0723800
e 07 1 5 9 0 0
e0692300
e0653600
e0600600

00534800
cO45791)O

e03725130
a0281900
e0190500
or3204600

en033630

U(1)
QOOl 7 5 8
e0001758
e

0003759

0001 7 5 9
0 0 0 17 6 0
e CJQ01761
e
0

0001762
OQ01765
e 000 1771
c

e0001784
flu0 1 8 2 8
oOOU2397
a

EN'liidPY

28

4,3270000

Y(11
e0000000
1676000
e3368008
05093000
e6870000
m8722000
1e O 6 8 0 O O 0
1 e27900OO
1 c51ooooo

1~7720000
20013200ua
2 e 47900OO
3e0780OUO

ENTHCrPY

P4I)
00667900
06617013
e0643100
~0612500

e0570600
e 0 5 18200
DO456500
00387200
e0312200
e0234200
0 157000

oO085550
e0027290

D(1)
0001562
e0001562

00001562
0 0001563
00001563

aoois63
e0001564

e0001566
0000156Q
e0001574
00001584
000 1624
0 0 0 2 12 9

6,3840000

P(I)
oObd17UO

oO630600
D06086ClO
e0576100

o0533700
00482300
e0422900

o03572OO
o0286900
a 02 14500
*L)l433UO
.0077780

o002.4750

!I(
I )
0001394
.OUG1394
OClO 1394
000 1394
0

00001395
oO001305
o000 1396
r0001398
~ O ~ O l ~ O U
e0001405
00 0 1 4 1S
,UO01450
0 0 0 0 1YO 1

29

M =

27
Y

II

I)
e0620un0
e0615000
~0600100
rU57560Q
054 1900
e0499500
00449400
0392500

1
2
3
4

5
6
7
0

o0330200
.i)%64300

1 (1
11
12
13

.
a

0 196900
u 13 1 1110

e0070960
00022490

14

U 1S I UR T I

M =

28
1

2
3
4

5
6
7
t3
5)

1u
11
12

13

30

2.5780000

e12064920

0001249
e 000 1 2 4 9
e 000 1249
eclr301249

.0001250
00001250
e OU(31251
e

uoo 1 2 5 1
000 1253

00001255
OOCI 1 2 5 9
0 0001268
e

oou 1300

0001705

id

ENltirtPY

L ~ ~ O L I O ~ I3. ~.63r)oao

e0001534

4.S430000

Y ( I )

.oooooar~
.i45iocm
02913000
m4396000
05911000
*7474000
e9099000

1~0810000

1 e263000O
1.4610000
le67900OO

1 e928OOOC1
2.2250000
2.6080000
3e18900OO

Y =

0020510

Uti)

P(1)

e0578500
e0574400
.05623oa
e0542400
e0514900
.048omo
e0438900
e0391800
e0339800
e0283900

e0225800
e0167300
e0110700
e0059570
e 00 18760

OR01 015
e 0001015
00010 15
0 0001015
0 000 1015
e0001015
eO00t015

Q001016
000010 16
e 000101 8
e

0001019

.0001023
e0001031
oOOUlOS6
e0001386

30
X ( I )
OUnOUOO
I) 1 AllSCIOU
02H2UOOU

Y(1)

a70 I 600
a2109000
3531 000
04977000
o6459000

1
2
3

e4254090

eS718UQO

a7224000
e6767000

e7989000

7
H

1eO43UOOO

1 e 1270000

.95a5oor3

D( I )

P( 1 )

e0558900
~0552500

e000U919

.UOOU919

e0537000

135 15400
e0487100
e0452400
e0412000
e0366600

e0000919
.0000920

0 0 0 0 9 19

e0000920
eOOOr3920

e0000920

31

1
2
3
0

5
6
7
8

Y
10

11
12
13
14
15
i6

y =

32

32

riY131000
3 3 1 4000
04667000

06049000
07471000
08946000
1.0490000
1 0 2 12uooo

1.3860O~
1 0576Q00O
1e 7 8 7 0 0 0 0

2e0290000
2.3180000
2.6Y10000
3 026100UO

DISTBRTIWN =

.OSlt)900
.OS061100
00468980
e0465300
00436400
00402600
00364300

0000761
000076 i
00000761
00000762
00000762
00000762
00000762

e0322200
oO276900
on229600
.0181100
0 133200
e0087490
e0046720
00 1 4 6 1 0

,0000762
.OD00763
00000764

.oa007~s
a0800768
,,0a0~)774

oO000793
000 1 0 4 2

.0025050

33

M =

34
1
2
3
4

5
6

7
8
9

1u
11

12
13
I4
15
16
17

M =

34

35

a0340000
vO305300
o

026AOOO

e0228700
01 8 8 J O O
a0147700
a 0 9 07YUO
a0070440
a

.0037380
~OUllblO

35

37

M =

P(1)
,0456400

040d2000
0522t3000
06440000

058270013

076840UO
e8969000

07052000
08315000

e0407590
00386600

100310000
1 1 7 10000
1031900UQ

e9623000
100990000

X(1)

1
2
3
4

5
6

7
8

9
1u
11
12

le
19

00000000

14
15
16

57

02874000

1.47SUOLiO
1.6~OOOUO
1 .BIOOOUO
2005700L)O
2 e 3 i1 00(.10
206310OflO
3e9880000

13

EN'lQ6PY

00572300
0 17 19000

Y ( I 1

00000000
i i4sooc1
02294000
m3454000
04629000

.0454400
m0d48400
e0438600
,0424900

.oa00498

*0000498
oOQ00498
. O O O C J ~ ~ ~
00000498

2.8~10000

00362400
s 0335100
00305100
00272700
00238400
00202700
0 0166300
0 0 129900
001394550
0061510
e0032520

3036500r30

00010060

000a0498
00000498
00000498
e0000498
r0000498
e0000499
.oa00499
~ 0 ~ 0 0 ~ 0 0
00000500
,0000502
00000506
m0000519
rOOU0682

Y(1)
013557300
0 id74000
02798000

P(1)
00644200
00440500

oQOC10461
00000461

1024300O0
i.39500ao
105600000
3.7400000

1.9410090
201720000
2.4500080

38
X(1)
000u0u00
1 1 16 O O O

e223605)O
03566000
0 6510000
05675000
e 6 13 6 6 0 0 0
o80980E30
e9356000

e3934000

05087OOU

.0433200
00422300
e0407900
m0390100

06263000
07468000
0871100O

90369200

1e0600000

e0318600

00345200
00289400

36

D(I)
,0000498

D(I)
00000461

oOOOO461
*0000461
e QUO0461
.omo46i
00000461
oOQ00461
00000461

11

13
14
15
16
17
18
19

ll I S 1'8i3 T iU N
EN7ROPY

201980000

o0258208
e0225380
00191200
.01566i30
00 1 2 2 1 0 0
.OBt)tr)720

2e3360000

204740@0O

a0057620

2.6540000

2.8330000

3010900i10

3038400130

00304iU
.00U93&9

102060Cl00
103520000
105C)9000Q

12

1*2770000
5a429flfl00
1a5910000

1068QOU00

Po769UUOO

108690000

1*9690001)

200830000

0017920
4,9720000

39
P(I)
1

.043351)0

,0431H00

3
4

mO4.26700

o0406500
13 3 9 1 6 0 0

7
0

0418200

*03736C)O
oO352700

9
10
11
12
13
14
15
16
17

*0339100
*0303000

18
19

0005~060

20

o0274700
00244600
00% 13100
001 Y O 5 0 0
*[I147600
aUll49UQ
oOCJ83370
.0028480
al!OO8778

37

40

0053QOOO

m2660000

e0413000

e3738060

e4831000
05944000
*7082006

004037Q0
0391 200
00375900

r0000396

00357800

00000396

e0337000
0 0313800
OO288400
00261000
o023200C7
m0201800

e0000396

7
8

1592000

00422500
so419300

e8251000
rP4590OO

10
11
12
13

100710000
1 0 2u30000
1 m3420001)
1m 4 9 0 O O Q C l

14

1 065OOOOO

mU170700

15

10 8 2 5 O O O O

16
17

2.02 1 0 0 0 0

18
19

2 52Ot~OOU
2m8750000
30 421000(1

0 139300
0 108300
q01370450
-0fl50790
00267 20
00008219

20

38

D(I)

P(1)

Y ( I )

2o267000n

mOQOO396
m0000396

~0000396
*0000396
e0000396

e0000396
m0000396

0000397
00009397
mOOO(3397
00000397
mOOOOJY8
rOOU039B
0

mO000400

eOO0Q403
e

OQOO413

mOODO543

M =

64
1
2
3
4

5
6

7
6
9

10
11
12

13
14

15
16

17
18
J. 9

20
21

22
23
24
25
ah
27

28
29

3U
3 s.

52

ENTtiOPY

128

39

40

1.5670000

42
43
44
47

le6160000
1 A700000
1072400OL)
1 *7790000
1 o8360000

48

1 e89600OU

49

1.9570000
2. f2210000
2.0Bc)0000
2. 1 5 9 O U U O

45
46

50
51
52
53
s4
55

2.2330000
2-31 10000
203950000
204640000

56
57
58
59
60
61

2.5860000

2.6860000
2.8030000
2.9330000
5.0830000
3 26 10000
3.4600000
3.77700UO

62
63

54

ENTROPY

e0058380
00055330

e0000012

e0052260

* 0000012

e00491 80
oOO46100

.a000012

o0043030

.00000J2
eo000012
.0000012
.OU00012

a0039970
e0036930
a0033920

.0000012
.0000012

e0030940
.0028010
00251 30
00022310

.0000012

e0019570
00016920
o0014370
e0011940
o0009650
oOOO7515

O O O O O 12
e0000012
.OO00012
~0000012
e0000012
00000012
e 0 0 0 0 0 12
.0000012
e 00000 12
.C)000090

e0005562
e0003821
oOOO2328
oOOO1424

e0000012

O O O O O 12

.0000012

.ooo 17 13

6.7200000

256
X ( I )

1
2
3
4
5

6
7

e0000000
~0165900
e0331800
e0497800
00663800

t(1)
.00000rl2

00248900

.ouoooo2

e0829900

e0414600
.0580AU@
.a746800
00912900

e0996000

lfl79000

eo000002
.0000002
ac1000002
0000002
e c1000002

1162000
1329000
1495000

1245000

.00000L32

014120QO

.0000002
r0000002

10

Y C I )

00082950

1578000

41

11
12
13

0 1662000
01828000
01995000

14

02162000
02330060

15
16
17

18
19
20
21
22
23
24

25
26
27
28
29
30

31
32

02665000
02833000
03001000
0 3 170000

e3339000
e3508060
e3677000
038470UO
o4018000
04188000

m4359000
04531000

04703000
04875009

05048OQO

33

05221000
e5395000

34
35
36

05569000
05744000
05920090

37
38
39

40
41
42
43
44
45

46
47
48

49
50
St

52

42

02497000

06096090
06273000
06451000
06629000
06808000

06987000
0 71 6 8 O O O
07349000
07531000
077 1 3 O O O
07897000
e8882000
08267000
08453000
08641000
08829000

I1 I. 111111 111 l 1 1 . 1

11111 I

53
54
55
56
57
58
59
68

61
62
63
64
65
46
67
68

69
70
71

72
73
74
75
76
77
78
79

80
81
82

83
84
85
86
87
88
89

90
91
92
93
94

o9019000

o9209000
o9401000

o9594000
.97880(30

.9983000
I 0 l8@00(3
1oO380OUO
1o O 5 8 O O O C )
1007ROO@0

1 mQ980000
1 11800DO
1 13900OO
1 15900OO
J. 1 8 O O O Q O
1.20ia000

i .2230000
1 .2d40OQCl
1 o2650000
1 e2870OQO
1r 3 6 9 O O O O
1o33100OO

i .3540000
1 d76DOOO
I o399OOOO
0.422OatIr3
1o 4 4 6 0 0 O O
1 046900OO
1 o493OOOO

1.5180000
1 o542OOOO
1e5670000
1 oS92OQOO
1.62AOOQO
1 o643OOQO
1 o670OOQO
1 06968000
M ~ ~ Q Q O

1.75 10000
P o77800QCl

1.887QQ00
108360R00

43

95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128

ENTRUPY

44

1.8650000
1.8950000
1.9250000
1.9560000
1o9880000
2.0210000
2.0540000
2.0880080
2. 1220000
2.1580000
2,1950000
2.2320000
2.2710000
2.3100000
2.3520000
2.3940000
2.4380090
2.4830000
2 0 5 310000
2.5800000

10 8800000
1.9 I oc)ooo
1.9410000
1.Q72OOOQ
2e00400QO
2.0370000

2. a7 io000
2.1050060
2 14 O O O O O
2017600Q0
2 0 21300OO
2.2510000

2.29@0600
2.3310000

2.3720000
2.415QOO0
2.4000000
2.5070000

2.5550000
2 a 605OOOO

2r6310000

2.6580000

2.6850000
2.7420000
2.8020000
2.8650000
2.9320000

2.71 30000
2 77 1OO(30

3.0040000
3.0820000
3.1660000
3.25900[30

303630000
3.4790000
3*6158OOO
3.7770000

7.7200000

2.8320000
2a8970000
20967GQOU

3.~42ooao
3. 1220000
3.2110080

3.3~380000
3.417aom
3 5420OOO
3 6 m m o
3.9570000

TABLE 10.- OPTIMUM QUANTIZATION LEVEL SPACING FOR EXPONENTIAL PDF

W I T H MEAN SQUARE ERROR

M =

DISTBRTIBN =
ENTRBPY =

D(I)
.5000a

1 .OOoa
0

0000030a

X(1)
.OOOOOOOa

1
DISTBRTIBN =

EWTRBPY =

?I=

Y ( I )

.7071a

P(I )
.5O0000Oa

D(I)
.2500a

5000a

1.0000030

3
X(1)
.7062000

1
2
DISTBRTIBN
ENTRBPY =

P(1)
1. O O O O O O O a

M =

Y(1)
,OoOOOOOa

X ( I )
.OOOOOOOa

.ooooooo
=

Y ( I )
.0000000
1.4120000

P ( I )

.6279000
.1860000

D(I)
.0800R00
,0910800

-2622030
1.3240030

X(I)
1
2

.onooooo
1.125OOUO

P(I)

Y ( I )

.4193000
1.8310000

.3981000
1019000
0

.D( I )

0 3 7 1 703
-0500800

a
Value derived by direct computation, n o t by the Max algorithm.

45

rlISTBHTIOIL =
ENTROPY

1.72Y0030

DISTBRTIBN =

'*1=

ENTRBPY =

.8379000
2-2480000

.2200000
.OS70800

,0205000
,0274800

P(1)
3 188000
.1441000
.0370700

0 13 0 10 0
.0133900
*0178800

1184030

1.9520030

.ooooooo
o7177000
1.8400300

Y(1)
,2992000
1-1360000

205430000

D(I)

2.2100030

X(1)

1
2
3
4

DISTBHTIBN =
ENTRBPY =

46

D( I )
*0223800

X(1)

M =

P(1)
.4457000

1
2

Y =

.oooooon

Y(I)

X ( I )
*4190300
1.5430000
.OOOOOOO

1
2
3

ENTRBPY =

,1745030

.2989000
1-0160000
2.136OOOO
.OOOOOOO

,0663230

2.3800030

Y(1)

oooooc)o
.5979000
1.4340003
208390000

P(I)
.3441000
.2088000
-0944700
.0247000

D(I)
.no92040
-0085040
oOO9756O
,0116000

DISTBHTIBN =
ENTRUPY =

M =

.ooooooo
.5311000
1,247OOOO
2.3640000

,2326000
,8296000
1.6640000
3,0640000

2,5700050

,232Si300
7632000
1,4780000
2.5930000

.0000000

DISTBHTIBN =
ENTRBPY =

P(I)

Y ( I )

,0000000
,4650000
1.0610000
1,8950000
3,2920000

,2798000
1900000
,1061000
,0490400
,0129700
0

D(I)
,0046360
,0043370
,0043790
,0045070
,0058810

,0420530
2.7080030

10

1
2
3
4
5

DISTBRTIRN =
ENTRBPY =

M =

D(1)
,0060370
,0060960
,0062750
,0062600

,0533330

X ( I )

,2641000
1502000
,0680700
,0176700

1
2
3

P(I)

Y(I)

X ( I )

1
2
3
4

X(I)
,0000000
-4225000
,9523300
1,6660000

2.7770300

Y(I)
,1903000
,6546000
1,2500000
2.0810000
3.4720000

P(1)

.2249000
,1451000
,0826200
,0375600
*0090540

D(I)
,0032860
.oo33oro
,0033320
,0034290
.0044550

,0356130
2.5580030

11
1
2

X ( I )
.1901000
c 6 121000

Y(I)

.ooooooo
,3803000

P(I)

,2355000
,1717000

D(I)
,0026520
,0025060

47

1.1410000
1 .R530000
2.9610000

3
4
5
6

.OOOOOOO

DISTBRTIUN =

2.9740030

X ( I )

,0000000
.3510000
.7725000
1.3010300
2.0120000
3,1160000

DISTBRTI3N =

1,0040000
1,5980000
2.4250000
3.8060000

P ( I )

d(11

.1956000
1367000
,0882600
.0503500
.0229700
,0061 020

,0019843
oOfl19890
00 19970

.0020740
,0026640

P( I )
.2033000
1557000
1088000
.0703200
.0401700
.0183700
0050 190

.0016560
.0015750
.0015780
001 5850
.0016000
.0016460
.0020860

.0020160

3,0960030

13

X ( I )

1
2
3
4
5
6

3608000
,51 14000
,9323300
1.4600000
2.1680300
3.2670000

.0000000

DISTBHTIBN =
ENTRBPY =

Y =

Y ( I )

..5409000
1610000

.0254530

ENTRFIPY =

.OG25150
.0025390
.0026120
.0033590

12

1
2
3
4
5

1108000
.0631800
.0287700
.0077320

,0297130

ENTRBPY =

M =

.8440000
1.4380000
2.2680000
3.5530000

Y(1)
.O000000

.3217000
,7011000
1 lh30000
1 7b600UO
2.5810000
3.9S400O0

..

D(1)

.021R030
3.1990030

14

D( I )

48

1
2
3
4

5
6
7

!7ISTbHTI4N =
ENTRBPY =

3031300
,65020130
t.n700C)00
1 .T9700O0
2.3040300
3.3930000

1395000

.4608000

.8397000
1.3010000
1,892i)OUO
2.71SOOOO
6,0810000

,1729000
1277000
,0893200

,0577600
,0330300

,01 51 4 0 0
,0040920

-00 12873
o0012890
- 0 0 12920
oOO12970
~0013090
oOO13460
,0017050

,0 1 9 0 5 3 0
3.3050030

I5
1
2
3
4

5
6
7

Y
DISTBHTIYr\: =

ENTROPY

.30300oc
0

X(1)
.I394000
,4394300
,7992200
1.20903UO
1.7350;300
2.44103UO
3.5330000
.0000CIOO

Y ( I )
.0000000
-2788000
,5999000
.975souo
1,4430000
2,0300000
2.8510000
4.214000n

P(I)
1788000
,1 4 1Y O 0 0
1048000
,0735300
.0474400
,0 2 7 1 5 0 0
.0124600
,0034560

D(I)
0 0 1 1030
,0010550
.DO10560
.0(310590

,0010630
,0010730
.01311030
- 0 0 13840

,0 1 6 6 9 3 0
'3.3930030

? I =1 6

X ( I )

1
2

3
4

9
6
7
9

TlIST8HTI3N =

,0030000
..2623200
,5 6 1 8 3 0 0
.9111300
1.33O0300
1.9550300
2.5580000
3.6430000

Y(1)
.1%31000
,4 0 1 3 0 0 0
.7221000
1.1000000
1 ,5 h 0 0 0 0 0
2.1490000
2,9670000
4.3200000

P ( I )

,1 5 4 9 0 0 0
o1192000

,0880500
06 1 6 3 0 0
,0 3 9 9 1 0 0
,0228800
,0105300
.0028920

D(I)
,0008821
,0038830

,0038842
,0008863
,0006901
.0008984
.0009231
,00115YO

0 1 4 8 1 30

49

ENTR4PY =

M =

3,4850000

17
Y ( I )

X ( I )

1
2
3
4
5
6
7
8

DISTDRTISN =
ENTROPY

v =

,0000000
,2457000
,5237000
,8437000
102210000
1,5800000
2,2660000
3.0800000
4.4210000

1229000
.3847000
o6837000
1.0320000
1 *4500000
1,9730000
2.6730000
3,7510000
.0000000

9
9

DISTBRTI4N =

FNTRSPY =

3,5650030

,0000300
,2327000
.A343300
.7929300
1.1410000
1.5580000
2.0800000
2.7780300

Y(I)
1100000
,3355000
,6 3 3 1 000
,9527000
1,3290000
1,7870003
2.3720000
3.18300GO
4.5170000

.0165700
,0076800
.0021590

D( I )
,0006295
,0006299
,0006305
,0006313
,0006328
.0006355
,0006413
.0006584
,0008041

3.8500300

Y(1)
,0000000
021 98000

P(I)
,1439000
1199000

r;( I )
,0005571
,0005375

P ( I )

1402000
1 1 12000
,0856100
,0633200
.0443900
0

,0268000

,0117930
3.6470030

19
X(1)

1
2

50

D(I)
,0007679
,0007380
,0007387
,0007398
.0007415
,0007446
.0007515
,00377 18
.0009507

.01 3 1200

X ( I )

1594000
.1301000
.1001000
,0739800
,0518300
*0336100
-0193000
,0089240
,0025420

18
1
2
3
4
5
6
7

P(1)

,1009300
03424300

.6037500
.9020000
1,2490000
1,6660000
2,1860000
2.8830500
3.9500000
,0000000

3
4

5
6

7
8
9

10
DISTBRTIBN =

X ( I )

,noooooo
,2093000
-4415000
.7025300
1,0000000
1.3470000

5
6
7

1.7630300

2.2520000
2.9760000
4.0370300

8
9

10
DISTr3RTIJN

FNTRBPY =

.0005379
,0005383
.0005391

.0005403
,0005426
,0005475
.0005620
.0036777

,0 106030

20

1
2
3

.0951800
a0732700
.0542200
.0380200
,0246900
,0142200
.0066080
.00 19200

.3 ,7 1800 30

ENTRBPY =

M r :

,4650000
,7424000
1.0620000
1.4370000
f ,8950000
2.4780000
3.2870000
4.6130000

Y(I)
.0994800
,3190000
.5640000
94 10000
1,1600000
1.5350000
1,9910000
2.5730000
3 ,3780000
4,6960000

P(I)

mr)

,1281000
,1041000
,0826400
.0636400
-0471200
,0330700
,0214900
0 124000
,0057790
.0016570

.0004653
,0004655
.0004658
,0004663
.0004669
,0004680
.0004699
.0004741
.0004865
,0005994

P( I )
.1311000
.1112000
.0904300
007 17900
,0552900
.0409500
.0287500
,0 197000

,0004 169
,0004035
.0004037
.0004039
,0004043
.0004049
.0004058
.0004075

,0396550
3.7920000

21

I )
.,3085000
0994 100
X(

1
2
3
4
5

6
7
8

,5406000
.8013000
1,0990000
1.4450000
1,8600000
2.3790000

Y(I)

.ooooooo

..41 9182000
88UUU
.hh30000
.9397000
1,2580000
1.6330000
2.oa8oooo

D(I)

51

9
10
11

3.0710000
4.1280000
,0000300

IIIST9RTIBN =
FNTRBPY

M =

22

.ooooooo
.1899000
.3986000
.6301300
o8902000
l.lR70i300
1.5320000
1.9450000
2.0610000
3.1480000
4.1930000

8
9

10
11

nISTBRTI4N =

Y ( I )

.
.

.0907200
2891 000
.5081000
7522000
1.02800OO
1.345OOOO
1.71900OO
2.1720000
2,7500000
3.5460000
4.8400000

P(1)

..0976800
1178000
.0794500
.0631100
.0486500
.0360600
.0253500
-0165200
-0095720
.0044990
*0013290

D(I)
.0003528
.0003529
.0003531
.0003533
oOO33536
000354 1
.0003549
oOOO3563
,0003594
.0003685
.0004493

.0080150

ENTRBPY =

3.9250050

23
X ( I )

1
2
3
4
5
6
7
R

9
10
11
12

52

0004 1 1 1
.0004217
*0005103

3.8560030

X ( I )

.0108000
oOO50440
,0014940

,0087730

1
2
3
4
5
6
7

2.6690000
3.4720000
4.7840000

.0906500
.2804000
.A889000
.7202300
.9800000

1,2760000
1.6210000
2.0340000
2.5480000
3.2330000
4.2740000
0000000

Y ( I )

.ooooouo
.1813OOO
.3796000
.5983000
.8422000
1.1180000
1.A350000
1.8070000
2.2600000
2.5360000
3.6300000
4.9170000

P(1)

1203000
1035000
.0858800
.0698700
.0555100
.042C)OOO
0031 7400
e0223300
0145600
e0084470
.0039800
.0012170

D(I)
00031 91
.0003096

.0003097
.0003099
.000310 1
*0003103
0003108
,0003115
.0003127
0003155
.0003233
oOOO3859

nISTOHTIDN =
EYTRRPY =

M =

.00733!30
3.9840030

24
X ( I )

1
2
3
4
5
6
7
8
9
10
11
12
DISTORTIdN =

ENTRRPY =

M =

.0000000

.1740000
.3636000
057 19300
.8030000

1.0620300
1.3580000
1.7030000
2.1140000
2.6270000
3.3100000
4.3450000

Y(1)
.08344UO
.2646000
.A627000
.h812000
.9248000
1.2000000
1.5160000
1.8-890000
2.3400000
2.9150000
3.7060000
4.9840000

P(I)
..0919500
1091000
,0762800
.0620800
.0493300
.0380500
.0282300
.0198700
.0129700
.0075360
,0035610
00 10720

D(1)
.0002744
,0002745
.0002745
.0002747
.0002749
,0002751
,0002755
.0002761
,0002772
.0002796
.0002865
.0003489

.0067840
A.0450030

25
X ( I )

1
2
3
4
5
6

7
8
9
10
11
12
13

.OR33800
.2573000
.4467000
,6548000
.8857000
1.1450300
1.4400000
1.78400OO
2.1950000
2.7070000
3.3880000
4.4180000
.OOOOOOO

Y(I)

.ooooooo

..3478000
1668000
.5457000
.7640000
1.0070000
1.2820000
1.5980000
1.9700000
2.4210000
2.9940000
3.7820000
5.0540000

P(I)
..0968800
12000
11

081 6800
oO677700
.OS51600
,0438500
.033&300
.0251100
- 0 176800
.0115500
00672 10
-0031840
.0009935

D(1)
.0002502
.0002434
.0002434
.0002435
.0002436
.0002438
.0002440
.0002443
.0002449
.0002459
.0002480
.0002540
.0003017

53

nISTURTIBN =

ENTRBPY =

M =

.0062510
4.1000000

26
X ( I )

1
2
3
4
5
6
7
8
9

10
11
12
13

FNTR6PY =

M =

.ooooooo

..3337000
1503000
.5227300
.7301300
.9602000
1.2180000
1.51 30000
1.855OCJOO
2.2630000
2.77 10000
3,4450000
4.4580300

Y(1)

..243SOOO
077 1300
.4240000
.~213000
.R389000

1.0810000
1.3550000
1.670000O

2.0390000
2.4870000
3.0550000
3.8340000
5.0d20000

D(I)
,0002166
,0002167
.00021 6 7
0002 168
,0002169
.0002 170
00032172
.0002175
.0002180
.0002188
.0002207
.0002258
,0002722

P(I)
.1034000
0910000
.0777800
.0655900
.0544400
.044.3300

D(I)
..0001947
0001999
.0001?47
..0001945
0001

4.1590030

27
X ( I )

Y ( I )

.ooooooo

1
2
3

.0771800
.2376000
4 112000

.6003000
oR080000
1.03801)00

.1544000
.3209000
.5015000
.h991000
.9169000

1.2970OOO
6.5910000
1.934OOOO
2.3430000
2.8530000
3.5280000
4.5470000
.0000000

1 160O003
1.43AOOOO
1 7490000
2.1190000
2.5680000
3.1370000
3 9190000
5.1740000

5
5
7
8
9
10

11
12
13
14

54

P(I)

~1014000
,0867000
073 1300
.0607 100
.0494500
.0395400
.0303800
,0225800
.0159300
- 0 104400
006 1010
.0029160
.0009 139

.0352600

.0272200
0202200
.0142600
.0093370
.0054480
.0025970
.0008283

948

.0001949
.0001950
.0001952
.0001955
.0001959
,0001967
oOOO1983
00002030
.0002400

DISTBRTIBN =

ENTROPY =

M =

,0053850
6.2080030

28
X( I )

1
2
3

4
5
6

7
8
9

10
11
12
13
14

DISTBRTIBN =
EhTROPY =

.ooooooo
-1487OOr3
-3087000
48 18000
-6703000
-8772000
1.1070000
1.3640000
1.6570000
1.9980000
2.4040000
2.9090000
3.5770000
4.5760000

Y ( I )

-0717700
-2257000
39 17000
,5719000
.7687000
.9857000
1.2270000
1.5000000
1.8140000
2.1810000
2.6270000
3.1910000
3.9630000
5.19oooon

P(I)

.0948500
,0820400
-0701500
.0591900
-0491600
.0400600
.0318900
.0246500
-0183400
01 29700
.0085150
-0049950
.0024050
.0007731

D(I)
0001 745
-0001745
-0001746
0001 746
-0001747
-0001747
-0001749
.0001750
-0001752
.0001756
-0001763
.0001777
.0001817
-0002190

.0350050

4.2620030

29

.-2205000
07 17800
X ( I )

1
2
3
4

5
6
7
8

9
10
11
12

*3805000
*5536000

.7422000
.9491000
1.1790000
1.4.360000
1.7290000
2.0700000
2.4760000
2.9810000

Y ( I )

.oooooon

1436000
-2975000
.4636000
,6437000
.84060i10
1.0580000
1,2990000
1.5720000
1.8860000
2.2540000
2.6990000

P(1)

-0964900
.0857100
-0741200
.0633800
.0534800
.0444200
.036190G
-0288100
.0222700
01 65700
.0117100
.0076900

D(1)
-0UOl616
-0001577
- 0 0 0 1577
.0001578
000 1578
0001579
,0001579
-0001580
.0001582
.0001584
.0001587
000 1593

55

14

13

.6500000
4.6500000

IS

.OOOOOOO

DISTBRTISN =

ENTRBPY =

3.2640000
4,.0360000
5,2640000

.0046630

30
X( I )

.ooooooo
.1387000
..4468000
2871300

,6294000
.R074000
1,0140000
1.2420000
1.4990000
1.7910000
2.1300300
2.5340000
3.0350300
3.6960000
4.6810000

7
8
9

10
11
12
13
14
15
nISTBRTI8N =

ENTRBPY =

Y(I)
,0670900
2 103000
,3640000
.5296000
,7093000
.9056000
1.1220000
1.3630000
1.6350000
1.Y470000
2.3120000
2.7350000
3 3 150000
4.0780000
5.2840000

P(I)

,0890600
.0778ioa
oO673200
.0575800
.O4861OO
,0403900
oO329400
,0262600
02031 00
0 151300
- 0 107100
,0070540
.0041560
.0020170
,0006669

nc 1 1
.0001425
.0001426
.0001426
0001 426
.0001426
*0001427
-0001427
,0001428
oOOO1430
*0001431
0001 434
0091 440
0001451
.0001483
,0001787

,0843740
4.3600030

31
X ( I )

1
2
3
4
5

56

.0001606
,0001643
.0001866

4.3050030

1
2
3
4

?l=

,0045100
.0021700
,0007160

,067 1000
.2058000
,3543300
.5139000
,6866000

.E3746309

7
P

1.3810000
1,3100000

Y ( I )

.ooooooo
.1342000
,2775000
.4311000
-5968000
.7765000
,9728000
1 1890000

P ( I )

.0905000
= 08 10100
.0707700
06 12300
,0523700
,04421 00
.0367400
.0299500

D( I )
.0001326
*0001297
,0001297
,0091297
.0001297
.0001298
.0001299
,0001 299

. I

9
10
11
12

13
14
15
16

1.5660000
1.8580000
2.1970000
2.6010000
3.1030000
3.7650300
4.7500000
.OOOOOOO

DISTBRTIBN =

.OOO 1 2 9 9
,0001301
.0001302
.OOOl 3 0 5
.0001310
. O O O l 320
,0001349
.0001524

4.4030000

32
X(1)

1
2
3
4

5
6
7
8
9
l G

11
12
13
14
15
16

nISTORTIrjN =
ENTROPY =

!q=

.023a600
.0184600
01 3 7 6 0 0
.0097390
oOO64120
.0037770
0018320
-0006216

. 0 0 4 0 9 10

ENTROPY =

1.4300000
1.7020000
2.0140000
2.3800000
2.8230000
3.3830000
4.1460000
5.3540000

Y ( I )

,0629700
1969000
.3398000
,4930000

.0O00000
1299000
,2683300
.4164000
.5756000
.7478000
,9352000
1.1410000
1.3690000
1.6240003
1.9150000
2.2520000
2.6530000
3.1500000
3.8050000
4.7730000

.6582000

.a374000
1.033001)O
1.249000O
1 ,48900GO
1.75900OO
2.0700000
2.4330000
2.8730000
3.4280000
4.1810000
5.3640000

P ( I )

.0839200
-0739700
.0646400
e0559400
,0478700
-0404300
.0336200
oO274300
.0218800
.0169500
.0126400
.0089700
,0059250
.0035070
oOO17170
a0005856

D( I )
.0001178
.0001179
.0001179
.0001179
.0001179
.0001179
.0001180
0001180
.0001181
.0001182
.0001183
.0001186
.0001190
.0001200
000 1225
.0001477

- 0 0 3 8 5 10
4.4510000

53 .
1
2

X( I)
.01530100
1930000

Y ( I )

.ooooooo

P(I)
.0852300

.1260000

,0768100

D(I)
.0001103
,0 0 0 1 0 8 0

57

3
4

.3515000
.4797000
6 3 9 1 000
.8114000
.9990000
1.2050000
1.4330000
1.6890000
1.9800000
2.3170000
2.7190000
3.2180000
3.8740000
4.8480000

5
6
7
8
9
10
11
12
13
14

15
16
17

.ooooooo

DIST8RTIBN =

.0001080
.0001080
0001080
-000108 1
.0001081
.0001081
,0 0 0 1 0 8 2
0 0 0 10 8 2
.0031083
.0001085
000 1087
0001091
0 0 0 10 9 9
.0001123
,0001328

.
..

4.4910030

34
X ( I )

1
2
3
4

5
6

7
8
9
10
11
12
13
14
15
16
17

DISTeHTI4r\i =

58

.0677000
.0591500
0 5 11 9 0 0
.0438000
.0369900
.0307500
,0250800
~0200000
.0154900
.0115500
.0081880
.0054030
,0031 9 4 0
.0015600
,00054 1 7

.0036350

FNTHBPY =

,2600000
.4030000
.5564000
7217000
90 10000
1.0970000
1.3130000
1.5ii30OUO
1.R240000
2.1350000
2.4990000
2.9400000
3.4960000
4.2530000
5.4420000

.ooooooo
.1221000
2518500
.3898000
.5375000
.6963000
.8679000
1.0550000
1.2600000
1.48600017
1.7410000
2.0300000
2,3650000
2,7630000
3.2560000
3.9020000
4.8530000

.0034150

Y ( I )
.Ob93200
.1850000
.3185000
04611000
.5139000
,7786000
.9572000
1. 1 5 2 0 0 0 0
1,3670000
1.6060008
1.8750000
2,1840000
2.5450000
2 9.8 1 0 0 0 0
3.5310000
4.2740000
5 4310000

?(I)
.0793200
.0704600
,0621200
0543100
*0470200
.0402500
.0340200

*0283000
.0231100
- 0 184500
.0143100
0 106900
*0076030
.0050380
.0029980
.0014820

-0005229

D( 1 )
.0000985
.0000985
,0000985
.0000985
.0000985
.0000985
.OOCJO986
.0000986
.0000986
,0030987
.0000988
.0000989
.0000991
.0000994
0001002
0001022
.0001234

..

ENTRUPY

4 ,5 3 7 0 0 3 0

35
1
2
3
4

5
6
7
8
9
10
11
12
13
14
15
16
17
18

X(1)
.0593500
.1816000
3 1P 3 0 0 0
.4494000
,5972000
,7561000
.9279000
1,1150000
1.3200000
1.5470000
1.8020000
2,0910000
2.4270000
'2.8260000
3.3200000
3.9680300
4,9240000

.ooooooo

D I S T B R T I BIN =

1187000
,2444000
,3781000
,5207000
.6737000
.8385000
1.0170000
1.2120000
1.4270000
1,6670000
1.9360000
2.2460000
2,6080000
3.0440000
3.5950000
4.3420000
5.5060000

P ( I )

.0804800
,0729700
-0648200
.0571400
,0499500
.0432400
oO370200
0312700
,0260200
- 0 21 2 4 0 0
0169500
,0 1 3 1 4 0 0
,0098100
,0069770
.0046190
,0027450
.0013540
.0004864

D(I)
,0000925
,0000907
,0000907
,0000907
,0000907
,0000907
.0000908
,0000908
,0000908
,0000908
,0000909
,0000910
.0000911
.0000913
,000091 6
.0000923
,0000942
.0001116

,0032340

ENTRBPY =

Y ( I )

.ooooooo

4.5750030

36

1
2
3
4
5
6
7
8
9

10
11

X( I )
OQOO0OO
1154300
-2374000
.3669000
,5048CIOO
,6523000
8109000

.9823000

1.1690000
1.3730000
1.5000000

Y(I)
,0561200
1746000
,3002000
.4336000
,5760000
,7287000
.8932000
1,0710000
1.2hh0000
1.4810000
1,7190000

P(I)
.0752800
.0675200
,0598000
.0527300
,0461100
.a399300
oO341900
.0289000

.0240500
,0196500
00 1 5 6 9 0 0

D( I )
.0000834
.0000834
.OOC)O834
,0000834
.0000834
.0000834
.0000835
,0000835
,0000835
.0000835
.0000836

59

12
13
14
15
15
17
18

1,5540000
2.1420000
2.4760000
2.8730000
3.3640000
4.0070000
4.Q500000

DISTBRTIBN =
ENTROPY

M =

.01 2 1700

.0091 050

.0064800

,0043000
-0025640
00 12730
00004556

.0000837
.0000838
.0000939
.0000842
.0000849
,0000865
.0000999

.0030530

4-6170030

37
X ( I )

1
2
3
4
5
6

.0560900
.1714000
,2933000
.4227ClOO
.5605000

16

,7079300
.5663300
l.r)380000
1.2240000
1.4280300
1.6540000
1.90800i)O
2.1950300
2.5290000
2.9250000
3.4150000
4.0550000
4.9920000

19

.ooooooo

7
8

9
10
11
12
13
14

15
16
17

DISTORTISN =
ENTRBPY =

60

1.9880000
2.2960000
2.6560000
3.0910001)
3.6380000
4.3770000
5,5240000

.0328940
4,6550000

YII)

.ooooooo
1122000
.2306000
,3560000
,4894000
.6316000
,7842000
.9485000
1 1270000
1.3210000
1.5350000
1.7730000
2.0420000
2.3490000
2.7090000
3.1420000
3.6870000
4.4230000
5.5620000

P(I)
oO7623OG
00694900
062 1500
-0552200
,0486900
.0425800
.0368800
,0315800
oO267000
-0222300
*0181600
oO145100
.0112600
.0084270
oOO60020
.0039870
,0023820
.0011860
.0004411

D(I)
,0000784
.Q000769
.0000769
.0000769
.0000769

.0000769
.0000769
.0030769
.0000770
.0000770
.0000770

.0000771
.0000771
,0000772
,0000774
.0000776
.0000782
.0000797
.0000941

38
X ( I )

.O ~ I O O O O D

1
2
3
4
5
6
7
3
9
10
li
12
13
14
15
16
17

10~2000

.7244300
.3462900
.A75A300
6 130300

.76iI2200
- 9 183000
1 08903UO
1.2750300
1.47900OO
1.7O50C10O
1 9570000
2.3440000
2. Fj77OOOO
2.57 10000
J.d580000
4.0940000
5.0220i100

19

13

DISTSKTI3N

.1652000
.2d35000
.4088000
542uooo
.h840000
.836301jO
1. 0 0 0 0 0 0 0
1.1780000
1.3720000
1.58600OO
1.8230000
2 ,0 9 i 0 0 0 0
2 . 3 3 7 00013
2.756000G
3.1870000
3 7290000
4.45900~0
5.5840000

P( I )
.0715700
.0643900
.0576000
.0511800
.i1451400
,0394800
,0342000
.0293000
,0247800
,0206300
-0169700
.0134800
- 01 0 4 7 0 0
,0078440
,0055940
.0037230
,0022310
,0011170
,0004118

D(I)
,0000711
00007 1 1
.0000711
,003071i
.0090711
oOO30711
,0 0 0 0 7 11
.on00711
.0000711
,003072 2
,0000712
,0000722
oOi300713
.0000714
.0000715

oOOOO71 R
,0000723
.0000736
oO030904

.11037570

EVTRBPY =

Y =

Y ( I )

0 5 3 2 1 OCJ

4.5940030

39

z
3
4

X ( I )
.L75317UO
1B23i)OO

.?77ALldO
,3 9 9 1 3 0 0

y ( I )
.0001)000

.. 210630130
153000

?(I)
.0724200
.056J400
o0596300

!I
I(
)
*Or300670
.0000658
.0030658

.356s003

oO534000

.0000658

61

.5282300

.66;570013
.5127000
.9707000
1.1410000
1.3270000
1.5310000
1.7560000
2.iln80300
2,2950000
2.6260200
3.0200300
3.5050300
4.m1030n

7
8

9
10
11
12
13
14
15
16
17
18
19
20

5.06303GO
~00000130

DISTURTIBN =

.0000659
,0000655
,0000658
,0000658
,0000658
.0000658
.COO0659
.0000659
,0000659
,0000660
.0000661

,0000662
.OO00664
,0000669
003068 1
.OD00789

4.7300030

40

X ( I )

1
2
3
A

5
6
7
3

9
10

11
12
13
14
15
16
17
15

19
20

rlISTSRTIBN =

62

.OA74500
0 4 18600
-0366100
0317200
.0271800
.0329900
- 0 191500
- 0 156600
0 125200
.0097300
.0072920
.0052040
,0034680
,002081 0
.001O460
.0004007

,0026040

ENTRDPY =

-4617000
.ti947009
,7366000
.BR87000
1.0530000
1.2300000
1,4240000
1,637000O
1.87500OO
2.1420000
2.44800130
2.8050000
3.2350000
3.7750000
4.5020OO0
5.6180000

.0000000
1036000
- 2 1 26000
.3274500
.4489000
.5777300
- 71 4 8 0 u O
8 6 15000
1.0190000

l.lR90000

1.3740000
1.5770300
1.802C)300
2.17530300
2.3390000
2.6680300
3.0590300
S.5430000
4.1560000
5.0730000

.0224810

Y ( I )
.OS05500

1567000
-2684000
,3864000
.5113000
644 1000
.7856000
.9374000

1.10100(30

1.2780000
1.4710000
1.5840000
1.9200000
2.1860000
2 4 9Oc)OOO
2.8450000
3.2720000
3.8080000
4.52500i)O
5 6210000

P( I )
.0681600

.06 16700
.0555300

.0496600
.0441400
.0389500
.0340800
.0295400

,0253200
02 14200
-0178600
,0146100
.0116900
,0090980

.006a280

,0048830
,0032630
oOOl9680
.0009973
.0003830

D(1)
,0000609
.0000609
.0000609
,0000609
.0030609
.0030609
.0000609

..

000061 0
0000610
.0000610
,0000613
00006 10
.0000611
.0000611
.0000612
0000613

.Oc)OO615

.0000619
.0000630
.0000776

1
2
3
4
5

6
7
8
Y

111

fl
12
13
14

15
Ib

17
IS

1 (2
23
21
2%
23
24

25
2h
27
4d
h9

33

31
J2

63

128
X(1)

03 1851)O

e 0 6 4 1900

00978300

6
7
8

e0000019

e0207200
en200600

c oc300019
00000019
0 00000 19
00000019
0 0~3000
19
.0000019
0 oocloo 19
e0000019

13040119
1643080

oG594600

1987000

oOlA2300

02337000

00176400
CJ 1.7O5UO

0188400

03055000

11
12
13
la
15

00164700
C) 1 59 1 00
.Ol53500
e0148000
0 I42680

03423@00

03799000
04180000
04566UOO
e496AOL30

5ci)

00220200
0 0 2 13700

e2693000

9
10

.000Q010

.000001P

00aoooi9

,ooano19
D O G 0 0 19

00137400

013000919
00000019
00000019
e0000019
* 0uOno 19
00[300019
0 000ou 19

r5567000
057780tJO
e6146000
e6625600
0 7f162UCI0

00132200
0 (3 12 7 10 0
e0122200
e0117300
001 12500

07!308000
r7863CJUO

e0107800
0 1 03200
oC11396770

00094390
0110901 10
00085930
00082850

roa00019
e ouooo 19
0ciD00019
0 O O G O O 19

,0077870
00073980
00070200

000000J 9
rOrlQ0019
0 00000 1 0

e0066520
00062930
.0059458

1.4xooao

oOUfi6060

1 499QUUO
105640OOO
1o6310OOG

00052770
00049590
00046500
.on43510
00040620
.ou37~30
00035140

eo000019
0 a O O O 19
e 0 0 G 0 (J 19
.00000J 0
OOUOcJ019

$6
f7
$8
19

20
21
22
23

64

P(I)

.oaoooau

08428000
08904000

24
25
26
27

28

100920QUO

29
30
31
32
33
34
35
36
37
38
39
40
41

I 1460OOO
l"2QlLrOljG

09390000
e9886OClR

10257r3c~!20
J03150000

10375OOO0

lo7010005

lr7720000
i*84600WO

1e923OOUO

UOOClOl9
oOb00019

oaaooo19
0

QOOOOl9
000001 Y
0000cl19

00000019
00000019

42
43

44
45

46
47
40
49

50
51
52

53
54
55
56

57
58

59
60
61
62
63
64

200030000
2.0860000
2e1720000
2.2620000

2.3560000
2.454UflU0
2.5570U01)

2.6663000
2.700UUUO

2.9uouooa
3r02800BO
3.165U000

~.w
3.4660000
3*635001)0
3r8180000
40019GQOG
dr241000Q

4o4680000
4.769U000
5.0930000
5.475UOOO
5.9440000

256
1
2

3
4

5
6
7
8
9
10

65

11
12
13
14

15
16
17

IS
19
20
21
22

23
24

25
26
27
26
29
40
31
32

53
34

35
36
37
38
33
40
41

42

l6430UU
1 8 1 .clGOCI

1987000
2 16 1i l U 0
,23370L30
e 25 1 4 0 8 6
e

.2693OUl?
e2873000
eJr35SU00
e3238000
e3423000

.361!JOn0
3 7 5 8 C! 8 0
,3988OC10
4 14GUQO
e

4d73021U

e4Yh9000

e4765UBO
4964OQO
e5165!lL:O
e536700C
e5572UUU
e5778OUO
e 5 9 I! 7 U fi 0

e61970OcI
6 4 lOU00

efi6250(30
e6d42fl00

,70620~0
e7283000
e7507000
~
7
3

45

e79630QO
e 61 94000
eAd29OOQ

46
47
dB

e8664000
eP8113U00
9 14501JO

49
50
51
32

,93913000
e9637OUt.l
e9888G00
le01400UU

43
44

66

53

54
55
56
57
58
59
60
61
62

63
64

65
66
67
68

69
70
71
72
73
74
75
76

lrO4OOOOO
1.0660000
1.062000O
1.1i90000
1.1A60000
1.173Q000
1.2010000
1,2290000
1.2570000
1.2860000
1~3250000
1.345000CI
1.3750000
1.4050000
1.4360000

1.4670000
lr49900UO
1.5310000
1.564UOUT)
1.5970300
1.6310000
1.6660000
1.7010000
1.73600C0

77
76

1*7720000

79
BO

1.9460000

81
d2
63
84

1.HL1900QO
1.98AClOC1C

1.923C'~OO
1.9630000
2.0030000
2.0440000

85
136

2.046UOCCI

87
86

2.1290000

69

2*1720OCO
2*217L'OC;O
2 .262~?ClG0

qCj

2 3Cldiu05

Y1

$43

2 356rlcll~c~
2 A i j 4 (liC I cj I:
2. 4 5 L i . J s l ) L l

QLi

2, 5 l . l ~ l - l L ~I:i .

$2

67

95
96

97
Y8
99

100

io1
102
103
104
105
106
in7
1D H

lU9
110

111
112
113
124
11s
116
117
118

IS9
120

121

122
123
124
125
126
127
128

68

TABLE 11.- OPTIMUM QUANTIZATION LEVEL SPACING FOR GAMMA PDF

W I T H MEAN SQUARE ERROR

UISTURTIBi~

1 .OOoa

a
Value derived by direct computation, not by the Max algorithm.

69

70

71

D(1)
tOU33780
00026520

mOO26800
oQ027050
001327660

moo31260

D(I)
eOU21140
eCJ022750

00022920
00025120
e0023620
00026690

72

73

PtI)
.2213000
0 1rl53clQfJ
e0668100

-0432900
e0270500

+ 0 1 55400
e0076110
00 131300

74

D(1)
-0009067

00009766
OUO9816
e0009844

000098C)l
00009949
e 0010 120
oOO11330

DXSTWRTIOY
ENTFBPY

*0135500
3.39Yf.1000

75

76

77

P( I )
*25&40U0
11O 7 O m
.073aooo

.OS20800

B(1)
0005202
00004143

00004178
0004 1 8 7
0004 1 9 2

~0374600

~0269000
.O 166100
* 0122200
oOO73780
*OU36660
0 028JJL10

0004197

qOOO4205
*0004227
e0004241
mOOQ4297
e0004702

79

80

81

M =

27
1

2
3
A

5
6

7
H

9
1il
11

12
1J
14

a2

83

D(
OOCI I 39 1
-0001502

O O G 1509
00001511
9001512
O O G 1513
c 000 15 13
@OO 15 14
c

. [ ~ a o i 5i 5
L

0001516
000 1518
000 1 5 2 2

*(3001528
000 1 5 4 1
e UOG 1 7 3 4
a

84

M =

31
1

2
3
4

5
6
7

8
9
1Li
11

12
13
14

15
16

85

86

87

88

M =

35
Y(

2
3
4
5

o3689000

h
7

o5298OtlCl
7U8 1000

e9047000
1.12200@0
1.36100013
1.627C)OXl

8
9

10
11

1.9230000

12

2.2570000
2oh370000
3.07SiJOOO
3 5920000
4.2 I 6 O O O O
5.0830000
6.0600000

13
1.4

15
16
17
i-5

u1sr

Fi7

.oflo0000
00999800
-2251 OOfl

f3N

UI 1
e@OO 1138
o0000912
o0000920
a0000922
oOUOO922
Nl00922

a0000923
00000923
.a000923
00000924
00000924
o0800924
e0000925
oOOOO926
0000928
r0000931

e0000937
.01303280

iJ0.33370

89

M E

36
1
2
3
4

5
6

7
8
9
10
11

12
13
14

15
16
17

ia

90

91

92

M =

39
1
2
3
A

5
6

7
13
9
10
11

12
13
14

15
16

17
it3

19

20

93

2
3
4

5
6

7
t3
9

10

I1
1%
13
14

15
16
1. 7
18

19
2 (1

94

M =

64

Y(1)
.0139400
00669500
o 1293000

e0555500
.a429900
mO358060

o2736000
0 3 S 4 1 0 0 13

oO308000
o0269800

7
8

.439~0oa
0530E)OQn

.023wmo
00212900

3
4

i9~oao

5
6

06270000

.a i90500

10

07287aao

11
12
13

08360000

eOf70800
e0153400
eo137700
e0123600
e 0 1 107OO
00098910
Clct88 150
00078270

09492000

17

1 oQS9000O
f e lO5OOOO
1e 3 2 d O O O O
10470OOQQ
i .twooao

19

10779000O
lo4480000

i A

15
14

ia

2.1290000
2a3239000

211
21
22
23

205310000
2 756OOOO

24

wmooooo

25
26

3.2660000
30559000Q
3 e 8H4OCIOf.l
4.24PL1000
4 64300OO
5.1410000
5.7060000
6a3950000

27

28
29
30
31
32

D(I)

P(I)
.0999900

.013369210
mOO60890
oUO53240
a0046250
.OU39850
e0034020

oO028720
00 23 93 il
oO019640
a m 5 8 1o
000 12430
OOO09489
a0006968
00004ti54
o0049820

00001 4 3

ro0~oi54
r0000155
00000155
00000155

aoao 155
e0000155
00000155
0 0000155
*0000155

rOO00155
e0000155
e0000155
r0000155
0 0000 155

.oona

I 55

000n155
e 0000 155
0

.ooaoim

0000155

Ocl00155

e0000155
eOU00155
e0000155
e0000155

o0ROOS55
r0000156

cOOOOi56
0 oOnOl56
e0000556
00000156
00000229

UO 1 0 0 7 0

95

128

1
2
3
A

5
6
7

a
9
10

11
12
13
14
15
16

17

ia
19
20
21
22
23
24
25
26
27
29
29

30
31
32
33

96

X( 1)
.c1000000

*c)175500
.0~237no
.070360C
1007000
1329GCO
1669COO
.202&0CG

Y l I )

.cc60eoo
.029020c
.C5572111;
.0850000
1l6400C

D ( I )
.@5632@0
oO375600
e0297300

.a2swoo
.0224500
oO2026GO
00 185200
0 170800
.fl1586CO
01 47900
o0138500
.c1130000
01 22400
.nii5~00
.0109900

04A47000
a49G7000

id95000
lP43000
o22050b0
.3582000
e2473000
m3377OOC
o3795000
.A225000
.4+69000
*5125@00

.536@00@

*5595000

.58;360130
.6326000

.607!30170
6575@00
.7085000
.7609030
.HldA000
e701 O G O

.0097A20
.UG92220
.00873AC
.0082740
.0078AOC

*9770000

.OO70390
,0066680
.006316C;
.0@59790

23CdOCiO

.277d000
o3175000
o35860ijO
40 1 c o o n

.6@30000

.73A7OOC
.7waooo
.6621000
.Eiq85000

e9562000
1.0150000
1 *07600OO
1. 1390300
1 o2O3OOOO
1 *2690000
1.3370OQO
1.4O70OOO
1 *47900@0

*98S4000

1 .Od5@0C)O
1.1070000
1 170OOOO
1J36000C
1.303OOOO
1 ,372OOOO
1 .4420000

c! 1030[30

.007A290

.00565ea

DII)
00000 18
c)oooo19
.0000019
.OOG0010
.000@019
.nOGOOJ9
.0000019
O O G O C 19
.oocuo19
.0000@19
0000019
.flO0@029
@ O O O O 19
m 00000 19

.nooooi9
.OO@@C19
.on00019

.oaooo19
00000 J 9
O@OC1O 1 0

.GOO0019
oOOOOG19
O G O O G 19

.eocooig

.nocIoo I

.COC:OG19
9

oU053520

.OOGOCl9
ooooc 19

.0050590

.OCOQCl9

.OOA7790

.OOOOGl9
mOOOOO19
OOOOC? 19

.OOAEllC)

1.51500Q!o

,c10d25dO

1.5"l

1~5Y1flGOG

acj0AOOBn

34
35

1.62C"O

1.6660000

1 *7080000

36

1. . 7 9 9 0 0 0 @

1 *7CR000@
1.8J10000

,0037720
,0035460
oCG33300

G O O O @1 9
oC!GCi0019
OOr2c1019

*C!OCc)Ol9

37
3s

39
40
41
42
A3
A4
~5
46
47
48

49
50
51
52
53
54
55
56
57
58
59
550

61
62

63
66

108730000
1.9600000
2-2370000

1.9 l d C O O @
2oC'OAOOOO
2m0950000
2.1Y9000O
2-2860000

2-5370000

2.38700GO

2-4390000

2oA920000
2-60ioonn
2.7 I 4noa0
2.e320000
209540000

2mO~90000

20142GOCI0

2-5460000
2.~?570000
207730000
2mR9300CO
3m0180000
3.15ouoco
302870000

3m43100C0
305830000
3o7d20000
309110G@O
dmOAQDCfi0
Am2790000

3m0830000

3.317OOCO
3.3570000

6.4620OOil

-0023770

0022100
002051 0
m0018990

*0017540

00 16150
0001 4840
00013580
-0012390

3-5050000
3-6600000
3082d0000

-0011270
00@102@c)
. 0 0 0 9 189
oO00823!?

30998(30ofl
A.181COOO

00007345
-00065137

4.3770000
4.5~600on
4-PiOOCOO

.@GO5724
mOOOA995
-000AJ19

6-4810000
A-6980000
~m~31ilOOO S.OE;~~OGO
5.18300UO
5m45700GO
5m75POOOO
6 nYOOOOc

-0031220
00029240
00027330
0 00255 10

5.3'140000
5.h000000
5 mW5OflOG

6.2h5000C
hm659000C

00000019
onooo i 9
mOGOOO19

00000019
00000019
00000 19
.0000019

00000019
0 O O O D O 19
0 00000 19
.ouooo19
00000 19
@ 0 0 0 01.9
oQ0000l.9
*cofl0019
.0000019
-0000019
oG000019

.onono

i9

.0oooc19
.0000G19

.O@CC010

.0005695

O E O O C 19

-0Cfl3123

.c)c00c19
0 00000 1 rj

LlCO260 1
.coo2129
.00017@7
oOC32950

*@oofl020
mOGOO020
0

OCOOO33

97

0251000

.020(36@0

!J( I )
sCG00002
-GO00002
0 r3000002

.0173100

.oooooc;2

5
6

.015A800

iJ( I

7
8
3
10

11
12

98

oOA39500

0000002

* 0 14 1 A O O
0130900
.@122300
-0115100

.0000002
00000002

oC!06913[?

-0103500

.0000002
GOD0002

a03987113

mCO00002

.00@0002

oOGG0002

.0000007
.0000002
OL!O0002

.OOO00@2
.0000002
.onti0002

T2000002
.08000@2
0000003
0

.0000002

.OQ~OOO~
.0COOO02
.0000002
.OGO@OD2

@000002
.0000002

.ooGnoo2
.O000002
.Or300@fl2

.O@OOOfl2
0000002
.0r;O@002
.0OG13002

.. @ o o o o c 2

.i)c,oao02
OQC~002
.1)001?002
.OQ00002
.OC0@002

.00000G2
*000@GO2

.nnoooc2

.00@Qn02

.OUI?I-ICQ~
.00@03@2

.c000002
.Cr300002
.0@000O2
.00000@2

.000GO@2
.OBOO@r2

.cacnorj2

99

55

SB
57
s0

59
60
61

62
63
6 A

65
66
67
68

69

70
71

72
73
74
75
76
77
78
79

YO
81
H2
83
ClA

k! 5
'36

87
a8
59
90

91
92
a3
9A

95
96

100

00370 1 0

.GO26260
00035530
00248 IO
0

.0024110
e0023420
oOO22750
0022 100
0021450
o0020830
.0020210
00019610
.Q019020
00 18440
*0017880
oC017320
oOO16780
00016250
.0015730
oOO15230
orj014730
.OC 1 4 2 6 0
00 13770
GO13300

o0012850
.0012400
.OO11970
.0011SAO
*GG11130
*@010720
o001032Q
.0@09935
oOOO9555
.0009185

.OOO8823

.OGOOO02
00000002
-.09@c1002
00000002
oG00fl002
00000002
00000002
00000002
.0000002
.000@002
0c000002
00000002
.OOOfl002
oG000002
0 0001?002
000fl0002
000c0002

.ouoooo2
.nOoooo2
000Q0002
.OflO00@2

00000002
00000062

.OOOOOC2.
00000002
00000002
.OOCiO002
.0000002
.0O00002

00000002
0~@00002
.@0000@2
.0000002

.@OOC002
*OOGCOO2

00008469
oOC108124
e0007788
oOOO7459
0007 139
.0000827

00000002

.0OflO0~2
.000O002
00000002

.COO6523

0n000002

.0000002

000(30002

97

-0006227

98

.0005939

09

aG005658

100

a0005395

IC1

.oan5izc)

1G2
1G3
104
105
106
107
108
109
110
111
112
113
114
1 J. 5
115
117
118
119
123
121
122
123
124
125
126
127
128

so004862
a0004611
.000434A
.ClO04132
,0003904
aOD036r?2

a0'303458
a0003260
.oon3060
m0002866
-0002679
00002499
.0602326?
.0002159
.005199?
ofl00lR~6
00016 9 9
.@O91559
a 0 0 0 14 2 4
.Ob01295
*@001174
.oc161059

s0@00950
00000847
-0On0751
00000660

.0032081!

DIST8RTIbY
EQTRBPY

7.G650300

101

TABLE 12.- OPTIMUM QUANTIZATION LEVEL SPACING FOR NORMAL PDF

WITH MAGNITUDE DISTORTION

X(I)

Y(I)

0.0000

0.6745a

I 0.5000a I

.3371
1.247

.2859
.2141

.05511
.07733

,0460
.1599
,3136
.5210
.8010
1.1790
1.6890
2.3780

.04099
.05257
.06820
.08393
.09321
.08531
.05478
.02100

.001062
.001768
.003086
,005097
.007563
.009181
.007765
.005289

D(I)
0.2377a

__

0000
.7921

1
2

L j:
-

0000
.lo29
.2367
.4173
.6610
,9900
1.4340
2.0330

!
8

Distortion = .08162
Entropy
= 3.8900

aValue derived by direct computation, not by the Max

algorithm.

102

TABLE 13.- OPTIMUM QUANTIZATION LEVEL SPACING FOR EXPONENTIAL PDF

WITH MAGNITUDE DISTORTION
~~~

X(I)

YO)

.3083
Distortion
Entropy

.3563
=
=

.oooo
8

,3996
.9390
1.2490
1.6670
2.0850
Distortion
Entropy

.oooo

16

.1272
.2984
.5295
.8414
1.2620
1.8310
2.5980

=
=

.0542
.2001
.3968
.6622
1.0200
1.5040
2.1570
3.0390

Distortion
Entropy

D (1)

,07441

.2907
1.8650
.2159
,1516
.08519
.04732
.1573
2.8020

.02096
.01969
.Oi478
.02322

.08229
.08987
.09139
.08433
.06825
.04633
.02487
.01269
.08401
3.8150

,002602
.003813
.005201
.006433
.006970
.006326
.004538
.006124

aValue derived by direct computation, not by

the Max algorithm.

103

TABLE 14.-

OPTIMUM QUANTIZATION LEVEL SPACING FOR GAMMA PDF

___

XU)

0.0000

..

Y -O )

0.0000

1.1430

--

P(I)

D (1)

l.OOOOa

0 . 2887a

.2265

.1134
.09965

,
7

.oooo

.2586
.8664
2.0990

.5172
1.2150
2,9820

,4077
.1414
.0820
,07278

.04032
.02073
.02309
,01913

Distortion = .1662
Entropy
= 2.4860

.oooo

15

.1863
.4377
.7771
1.2350
1.8540
2.6880
3.8150

1.5440
2.2710
3.2520

.2589
.1128
.07862
.05912
.04243
.02730
.01485
.03540

.009482
.006029
.005663
.005734
.005528
.004770
.003475
.005371

~~

Distortion = .08262
Entropy
= 3.4760
aValue derived by direct computation, not by
the Max algorithm.

104

TABLE 15.- OPTIMUM QUANTIZATION LEVEL SPACING FOR NORMAL PDF

WITH RELATIVE DISTORTION
-.

1
1
'!

1
3

___~

q-x(1)2
--

~~

p (1)

Distortion
Entropy

.3160

--_

1- 1

.OOOO

.6320
-

0 . 5000a

1. OOOOa

0.0000

0.0000

=
=

l.OOOoa
.0000a

.2480
.3760

'1

.2480
.1541

Distortion = .5562
Entropy
= 1.5600

1
2
3
4

- _.

.07174
J
:
:
:

.OOOO

.1435
.5064
.9964

.05719
.0988
.1464
.2262

.05719
.03522
.03000
.05763

- ~-

Distortion = .3029
Entropy
= 2.6780
-.

15

.01912

1
2
3
4
5
6
7
8

.2510
.4433
.7029

.1692
.3329

1.7980

.OS588
.07213
.08771
.09495
.08263
.06347

.01912
.01203
,ELL141
.01003
.009887
.009369
.007438
.008352

Distortion = .1562
Entropy
= 3.8110
-

31

1
2
3
4
5
6
7
8
9
10

11
12
13
14
15
16
-_.
-.

.00237
.01065
.02455
.04333
.06866
.lo28
.1489
.2110
.2947
.4077
.5601
,7659
1.0440
1.4190
1.9250

.oooo
.00474
-01656
.03254
.05411
.0832
.1224
.1753
.2466
.3428
.4726
.6477
.8841
1.2030
1.6340
2.2160

.00189
.00330
.00554
.00748
.01009
.01358
.01823
.02437
.03235
.04235
.05406
.06583
.07354
.07031
.05089
.02713

.001891
.001176
.001128
.001040
.001140
.001349
.001662
.002095
.002666
.003388
.004229
.005060
.005569
,005250
.003753
_ _.002761

Distortion = .08642
Entropy
= 4.5250
aValue derived by direct computation, not by
the Max algorithm.

105

TABLE 16.- OPTIMUM QUANTIZATION LEVEL SPACING FOR EXPONENTIAL PDF

VITH RELATIVE DISTORTION
~

D (I)
0 . 5000a

1
.2632
.1754
3
.

..

.05709
.2586
.7684

1
2
3

.0199
.0901
.2086
.3689
.5854
.8775
1.2720

15
8

.07757
.04050
.04570
.04720

.07754
.1144

.1782
.1687

_-

.oooo
.0398
,1404
.2768
.4610
.7097
1.0450
1.4990

.02775
.01636
.01397
.01053
.008835
.007317
.005597
.01807

.02775
.04594
.06792
.07551
.07825
.07396

.06180
.08274
.-

.- .

Distortion = .1891
Entropy
= 3.8640

31

.OOOO

.1142
.4030
1.1340

.0000
.00528
.01860
.03662
.06096
.09379
.1381
,1979
.2785
.3874
.5343
.7327
1.0000
1.3620
1.8500
2.5080

10
11
12

16

.-

.003727
.006507
.01077
.01419
.01849
.02379
.03011
.03727
.04478
.05167
.05644
.05731
.05272
.04248
.02867
.02294
. . -.

Distortion = .09609
Entropy
= 4.7300

__

.003727
.002324
.002202
.001976
.002092
.002364
.002744
.003201
.003686
.004123
.004402
.004391
.003984
.003176
.002128
.003388
_ _
__

aValue derived by direct computation, not by

the Max algorithm.

106

~-

TABLE 17.-

OPTIMUM QUANTIZATION LEVEL SPACING FOR GAL4MA P D F

WITH RELATIVE DISTORTION

1
2
3
4

.oooo

.0220
.1592
.5660

.0440
.2744
.8575

.1290
.1229
.1387
.1740

.1483
.0549
.04134
.05744

2.7930

.oooo

.0020
.00906
.02692
.0721
.1864
.4756
1.2070

.0040
.01412
.03972
.lo45
.2683
,6828
1.7310

.03916
.02638
.03564
.05262
.07695
.lo29
.lo80
.07798

.04499
.009456
.009333
.01256
.01769
.02307
.02342
.01910

Distortion = ,2743
Entropy
= 3.7620
1
2
3

4
5
6
7
8
9
10
11
12
13
14
15
16

.000522
.002640
.008263
.02241
.04648
.07929
.1236
.1834
.2642
.3732
.5203
.7189
.9870
1.3490
1.8370

.oooo
.001044
.004236
.01229
.03253
.06044
.09815
.1491
.2178
.3105
.4358
.6048
.8330
1.1410
1.5570
2.1180

.02002
.01496
.02066
.03048
.03360
.03283
.03368
.03529
.03711
.03867
.03948
.03899
,03671
.03233
.02598
.03920

.02299
.00579
.005652
.007378
.006070
.004304
.003666
.003416
.003320
.003275
.003216
.003087
.002846
.002466
.001958
.007932

Distortion = .1517
Entropy
= 4.9140
aValue derived by direct computation, not by
the Max algorithm.

107

-X

0
X,o=l

Figure 1.- Normal, exponential, and gamma probability distributions.

Magnitude of quantizer input

Figure 2.-

The mean square, magnitude, and relative magnitude distortion

measures.

109

6
0)

=
l

>
0
W

g5
0
c

In
W

?p.
!
W

L4

+
In

e
4
3-

2-

0 Normal

0 Exponential

0 Gamma, M
Gamma,

even

M odd
I

5
6
7
Rate, log2 M

I
8

IO

I
II

Figure 3.- Optimum equal spacing for the meansquare error quantizers.

5 6 7 8
Rate, log2 M

Figure 4.- Largest representative values for the

equal spacing, minimum mean-square error
quantizers.

1.0

.I

.o

p!

In

c
0

E.

.oo

0
.e
L

In
.-

,000

.oooo

.ooooc

Figure 5.- Distortion for the equal spacing,

minimum mean-square error quantizers.

111

1.0

.I

.a1

in
U
C
0

E
c

0
.c
b
c
.-

0
c
.b .oiu)
._

001

,001

Magnitude
distortion

.0001
0 Norm(
0 Exponential

0 Gamma, M even
Mean square
distortion

.0001

I
I
I
5
6
7
Rate, log M

1011

Figure 6.- Mean square and magnitude distortions

for the gamma input equal-spacing quantizers.

Gamma, M odd

IO

1.0

8-

7-

6-

0)

50)

.e

14-

ea
e

0 Normal
0 Exponential

0 Gamma,

M even
Gommo, M odd

.OOl

Figure 8.- Largest representative values for the

unequal-spacing, minimum mean square error
quantizers.

Rate, log2M

Rate, log2 M

I-'

Figure 9.-

Optimum equal-spacing for the magnitude error quantizers.

0 Normal

0
1.0

I-

Exponential
Gomma

,001

Exponential
Gamma, M even only
(See figure 6 for M odd)

.01-

0 Normal

8
Rate, log2 M

Figure 10.- Distortion for the equal spacing,

minimum magnitude error quantizers.

.OOl0

I
8

"
1

'

Rate, log2 M

Figure 11.- Optimum equal spacing for the minimum relative error quantizers for M odd
only.

I .o,

.I

Q)

>
.t

-0

2
.01

c-

0
.+
L
0

4v)

_0

0 Normal
I7 Exponential
0 Gamma

,001

.0001

I
4

l
l
1
5
6
7
Rate, log2 M

IO

II

Figure 12.- Distortion for the equal spacing,

minimum relative error quantizers for M
odd only.
115

1.0'

.I

.01

E
0

0v)

c
0

0)

,001

0
.c
L

0
+

v)
.n

.0001

.00001

.000001
0

2 3 4 5 6 7 8 9 1 0 1 1
Entropy, bits

Figure 13.- Mean-square distortion vs. entropy

for equal-spacing quantizers.
116

.I

.01

.OOl

.oooI
0 Gamma

.oooo

I
1

1 1 -

3 4 5 6
Entropy, bits

Figure 14.- Mean-square distortion vs. entropy

for unequal-spacing quantizers.

117

-Rate

distortion bound

0 Equal spaced quantizers

Unequol spaced quantizers
spaced, entropy coding
A Unequal spaced, entropy coding
[7

0 Equal

...

Distortion, mean square

Figure 15.- Rate-distortion curves for normal input and mean-square distortion.

118

-Rate
distortion bound (lower bound)
0 Equal spaced quantizers
Unequal spaced quantizers
0 Equal spaced, entropy coding
A Unequal spaced, entropy coding

I
.01
Distortion, mean square

,001

.OOC)I

IO

.I

Figure 16.- Rate-distortion curves for exponential input and mean-square

distortion.
7

I 4
N

-ol
0)

z 3

-Rate

C
.(

distortion bound (lower bound)

(M 64,128 used for unequal)

1-

01

,001

.01

.I

1.0

Distortion, mean square

Figure 17.- Rate-distortion curves for gamma input and mean-square distortion,
M odd.
119

-Rate
distortion bound (lower bound)
0 Equal spaced quantizers
0 Unequal spaced quantizers
0 Equal quantizer, entropy coded
A Unequal quantizer, entropy coded

0
.01

.03

_I
Distortion, magnitude

.3

Figure 18.- Rate-distortion curves for normal input and magnitude distortion.

..

Distortion, magnitude

Figure 19..- Rate-distortion curves for exponential input and magnitude

distortion.

120

5 4
N
0

z 3

-Rote distortion bound (lower bound)

0 Equal spaced quantizer
0 Unequal spaced quantizer
0 Equal quantizer, entropy coded
A Unequal quantizer, entropy coded

.03

.I
Distortion, magnitude

.3

Figure 20.- Rate-distortion curves for gamma input and magnitude distortion,
M odd.

0 Equal spaced quantizer

0 Unequal spaced quantizer
0 Equal quantizer, entropy coding
A Unequal quantizer, entropy coding

Distortion, relative

Figure 21.- Rate-distortion curves for normal input and relative distortion,
M odd.

121

0 Equal spaced quantizer

0 Unequal spaced quantizer

Equal quantizer, entropy coded

A Unequal quantizer, entropy coded

\

F i g u r e 22.-

Rate-distortion

c u r v e s o r e x p o n e n t i a l i n p u t and r e l a t i v e d i s t o r t i o n , M odd.

0 Equal spaced quantizer

0 Unequal spaced quantizer

Equal quantizer, entropy coded

A Unequal quantizer, entropy coded

F i g u r e 23.-

122

_I
Distortion, re lot ive

.3

D i s t o r t i o n c u r v e s o r gamma i n p u t and r e l a t i v e d i s t o r t i o n , M
NASA-Langley,

1977

odd.

A-6714

N A T I O N A L A E R O N A U T I C S A N D S P A C E ADMINISTRATION
W A S H I N G T O N . D.C. 20546
-OFFICIAL B U S I N E S S
PENALTY

FOR PRIVATE USE 5300

S P E C I A L FOURTH-CLASS R A T E
BOOK

POSTAGE AND FEES P A I D

N A T I O N A L AERONAUTICS A N D
SPACE A D M I N I S T R A T I O N
451

D 7719218 S33903DS
148 0 0 1 Cl U
DEPT Or THd A I R FORCd
A F IUi3APONS LABORATOHY
BTTM: T E C H N f C A E L1BiME.Y (SUL)
KIBTLAND AFB N M 67117

If Undeliverable (Section 158

Postal Mnnnnl) Do Not Return

-~

The aeronautical and space activities of the United States shall be

conducted so as t o contribute . . to the expansion of human Knowledge of phenomena in the atmosphere and space. T h e Administration
shall provide for the widest practicable and appropriate dissemination
of information concerning its activities and the results thereof.
-NATIONAL AERONAUTICSAND SPACE ACT OF 1958

NASA SCIENTIFIC AND TECHNICAL PUBLICATIONS

TECHNICAL REPORTS: Scientific and ,
technical information considered important,
complete, and a lasting contribution to existing
knowledge.

TECHNICAL TRANSLATIONS: Information

published in a foreign language considered
to merit NASA distribution in English.

TECHNICAL NOTES: Information less broad

in scope but nevertheless of importance as a
contribution to existing knowledge.

SPECIAL PUBLICATIONS : Information

derived from or of value to NASA activities.
Publications include final reports of major
projects, monographs, data compilations,
handbooks, sourcebooks, and special
bibliographies.

TECHNICAL MEMORANDUMS:
Information receiving limited distribution
because of preliminary data, security classification, or other reasons. Also includes conference
proceedings with either limited or unlimited
distribution.

CONTRACTOR REPORTS: Scientific and

technical information generated under a NASA
contract or grant and considered an important
contribution to existing knowledge.

TECHNOLOGY UTILIZATION
PUBLICATIONS: Information on technology
used by NASA that may be of particular
interest in commercial and other-non-aerospace
applications. Publications include Tech Briefs,
Technology Utilization Reports and
Technology Surveys.

Details on the availability of these publications may be obtained from:

NATIONAL

SCIENTIFIC A N D TECHNICAL INFORMATION OFFICE

AERONAUTICS

AND

SPACE

Washington, D.C. 20546

ADMINISTRATION