where s(n) is a sequence of arguments of complex numbers,
l is a label, and kι(n) is a sequence of integers which satis­
Analysis of the phase unwrapping algorithm
Kazuyoshi Itoh
Hokkaido University, Department of Engineering
Science, Faculty of Engineering, Sapporo 060, Japan.
Received 2 February 1982.
0003-6935/82/142470-01$01.00/0.
© 1982 Optical Society of America.
Several workers1-3 have recently proposed digital tech­
niques for high-resolution imaging through the turbulent
atmosphere. The basic concept of these algorithms is to
calculate and average phase angles of a series of image Fourier
transforms to suppress the unwanted atmospheric effects on
image resolution. Since computed phase angles contain the
ambiguities of integral multiples of 27rrad, it is necessary to
obtain continuous phase curves without the ambiguities before
averaging. This process of eliminating the ambiguities is
called phase tracking1 or unwrapping.4 A similar problem has
been discussed by Oppenheim and Schafer4 and Tribolet5 in
the context of realization of a certain homomorphic signal
processing system.
Although phase tracking of sampled signals is an important
step in image processing and other related fields, there are as
yet no fundamental studies on the necessary conditions of
tracking the phase sequences correctly. The only exception
is the very suggestive but intuitive study by Mertz.1
This Letter clarifies the exact tracking conditions of a cer­
tain class of phase tracking algorithms. The result of the
analysis suggests the possibility of high-speed execution of
phase tracking.
For the purpose of brief discussions let us first introduce
two operators Φ and Δ. The operator Φ is defined as
2470 APPLIED OPTICS / Vol. 21, No. 14 / 15 July 1982
fies
The operator Δ is simply defined by
Their functions are wrapping and differentiating, respec­
tively.
Now from the definition of Φ it can be readily shown
that
and that if
k2(n) satisfies the equation
Hence, if Eq. (5) is satisfied,
Equation (7) states that the actual sequence can be retrieved
from the sequence of its principal values by a cascade of three
simple operations; differentiating, wrapping, and integrating.
This is identical to the function of Mertz's phase tracking al­
gorithm.1 The necessary condition for exact tracking is
clarified as Eq. (5). It is interesting to note that phase un­
wrapping can be achieved by wrapping its difference. A priori
information on the difference of the sampled phase is essential
for this class of phase tracking. Unfortunately the condition
[Eq. (5)] is not always satisfied in practice due to the presence
of noise in the image transform phase caused by the turbulent
atmosphere and photon fluctuations.
Let us consider the method of fast implementation of the
cascaded operations for phase tracking. The operator Φ,
which is both the first and the last operator in Eq. (7), can be
executed in one machine cycle; that is, the operator Φ can be
automatically realized by truncating the resultant values of
the preceding operation with a register of finite length, ig­
noring the borrow and carry digits. When twos-complement
notation6 with 8-bit word is used, for example, π(1 — 1/128)
and —πrad should correspond to the maximum value
(01111111) and the minimum value (10000000), respectively.
The final integrating stage, however, requires more than 8 bits
to track the unwrapped phase.
In summary, (1) the phase tracking algorithm can be rep­
resented by a cascade of three simple operations; (2) the re­
sultant sequence is exact only when Eq. (5) is satisfied; and
(3) all the requisite operations, including the wrapping oper­
ation, are fundamental for digital computers and may be ex­
ecuted in several machine cycles.
The author wishes to thank Y. Ohtsuka for his encourage­
ment and helpful comments.
References
1. L. N. Mertz, Appl. Opt. 18, 611 (1979).
2. A. H. Greenaway and J. C. Dainty, Opt. Acta 25, 181 (1978).
3. K. Itoh and Y. Ohtsuka, Appl. Opt. 20, 4239 (1981).
4. A. V. Oppenheim and R. W. Schafer, Digital Signal Processing
(Prentice-Hall, Englewood Cliffs, N.J., 1975), Sec. 10.6.
5. J. M. Tribolet, IEEE Trans. Acoust. Speech, Signal Process.
ASSP-25, 170 (1977).
6. Ref. 4, p. 408.