Vous êtes sur la page 1sur 4

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/313102237

A Study of Various Video Compression Techniques

Conference Paper · January 2017

CITATIONS READS
0 740

3 authors, including:

Mary Shanthi Rani Chitra P.


Gandhigram Rural Institute Gandhigram Rural Institute
42 PUBLICATIONS   90 CITATIONS    10 PUBLICATIONS   7 CITATIONS   

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Mary Shanthi Rani on 31 January 2017.

The user has requested enhancement of the downloaded file.


A Study of Various Video Compression Techniques
M. Mary Shanthi Rani1, P. Chitra2 and K. Anandharaj3
Assistant Professor1, Ph.D Scholar2 and M.Phil Scholar3
Department of Computer Science & Applications
The Gandhigram Rural Institute-Deemed University, Gandhigram, Tamil Nadu, India
drmaryshanthi@gmail.com, bchitra.tr@gmail.com, anand.student@yahoo.com

Abstract
This paper presents the analysis of video 2. Overview of Video Compression
compression using various transform based techniques Techniques
for finding best video compression technique. Video
compression is a process of eliminating redundancy and There are various transform based methods that
irrelevant information for an efficient storage and improve the performance of video compression
transmission .Transform coding s used to represent a techniques. Initially, transform methods are applied to
video file in a compact form. In this paper, we have enhance the image quality in each video frame. The
investigated few transform methods such as Discrete encoder and decoder of transform based video
cosine Transform(DCT), Discrete wavelet compression is shown in Fig.1 and Fig.2, respectively.
transform(DWT) and Discrete fractional
transform(DFT) finally we observe that transform 2.1 Discrete Cosine Transform (DCT)
coding achieves high quality in video than other
DCT method allows a representation of image data
technique.
in terms of frequency components. A discrete cosine
Keywords: DWT, DCT, Quantization, Huffman coding. transform (DCT) conveys a sequence of many data
points in expressions as a sum of cosine functions at
1. Introduction different frequencies. The complexity of DCT methods
is very expensive. DCTs are significant to lossy
Now a days, the usage of videos are essential for the compression of audio (e.g. MP3) and images (e.g.
purpose of entertainment and communication. Normally JPEG).
video file occupies huge amount of memory to store. If
such videos are sent over the transmission media, huge . Hybrid
requirement of transmission bandwidth is needed. In Input Video Frame DWT, DCT
video communication and storage, capabilities are Video Conversion Algorithm
limited and expensive. For example HDTV video signal
requires 1133GB per second but, the compressed video
file requires only 20mbps. Therefore, Video
compression is highly essential for compressing video in
a transmittable size. Compressed Coding
Digital video represents the images moving in the Video Quantization
form of digital data, where, the analog video represents
the moving images in analog video format. Video Fig.1. Encoder Block diagram
compression is a process of reducing the size of video
files. Video files needs to be encoded and decoded from
one form to another form. Eliminating some irrevalant Compressed Hybrid
details is called as lossy process. However the video can Video Inverse coding IDWT, IDCT
Algorithm
be losslessly compressed by individually encoding the
frames in the videos. Basically, Video can be
compressed by exploiting Spatial and Temporal
redundancy. Some of the popular compression
Decompressed Video Dequantization
techniques are Discrete Cosine Transform, Discrete Video Frames
Wavelet Transform, Run Length Encoding and
Huffman Coding. In this paper, an overview of video
compression techniques is presented.
Fig.2. Decoder Block diagram

Computational Methods, Communication Techniques and Informatics


ISBN: 978-81-933316-1-3 293 293
Disadvantages of DCT are the spatial correlation of The length of the applied bits depends on the
the pixels inside the single 2-D block and between the amount of data contained in the source character. Thus
pixels of the neighboring blocks is exploited for the main idea of the Huffman coding is to replace each
compression. This method is impossible to fully of the source symbols with a simpler one and is carried
decorrelate the blocks at their limits using DCT. out step by step. This step is persistent until left with
only two symbols for the optimal code.
2.2 Discrete Wavelet Transform (DWT)
The DWT passes a signal into an image, through the 3. Analysis of Video Compression
filters, such as low pass filter and a high pass filter. The Techniques based on Transforms
low pass filter yields low resolution signal and the high There are many video compression techniques have
pass filter yields variation signal. Then the outputs of been compared.
these filters are down sampled by two. The down Bernatin T et al. (2014)[1] proposed a Hybrid
sampled outputs have the similar number of bits as the Transform and Quantization with Huffman Coding for
input signal. The original signal is reproduced, when the Video Codec. The proposed method uses the motion
up sample output of the low pass filter is added to the up vectors, found from estimation using adaptive rood
sample output of the high pass filter. The output of the pattern search and is compensated globally. The hybrid
high pass filter is fed into another couple of filters and compressed frame is quantized and entropy coded with
the process repeated. Huffman coding for generated bit streams are
The wavelet transform (WT) has gained extensive transmitted to the decoder. The algorithm achieves the
acceptance in signal processing and image compression. size of the compressed frame saving maximum size
Because of their natural multi-resolution, wavelet- storage space.
coding schemes are specially suitable for applications Escalin Tresa L et al. (2014)[2] The proposed
where scalability and tolerable degradation are method uses a Hybrid DWT-DCT algorithm on motion
significant. compensated frame by taking the compensation of both
the method. The performance of the proposed method is
2.3 Quantization evaluated using Compression ratio, PSNR and mean
square error.
Quantization is needed because the significant S. S. Wadd et al. (2014)[3] This paper presents the
visual information is present only in a few transformed performance of Discrete Cosine Transform (DCT) based
coefficients. Therefore, the number of bits representing algorithm for compressing video. The advancement in
the transform should be reduced. This process is called video compression technology with visual quality and
coefficient quantization is implemented to being scales content of video. Block matching algorithms used for
and sub bands. The set of inputs and outputs values are motion estimation in video compression.
scalar quantization or vector quantization. Scalar S.V. Phakade et al. (2016)[4] The author proposed
quantization is a mapping of an input value into a fixed video compression using Hybrid (DCT-DWT)
number of output values. Vector quantization blocks as algorithm. The compression scheme get high
vector for a given rate in a lower distortion than when compression ratio with good reconstructed video quality.
scalar quantization used at the same rate. The hybrid compression technique uses DCT transform
as well as DWT transform. Arithmetic coding is used for
F(u,v) = ( F(u,v) / Q(u,v) ) (1) additional compression. The performance of the system
is evaluated using Compression Ratio, PSNR and Mean
Where, F (u, v) – DCT Coefficient Matrix. Square Error.
Q(u, v) – Quantization Matrix.
4. Experimental Result and Discussion
2.4 Huffman Coding This section performs analysis of existing methods
The entropy coding used in Huffman coding. This is for video compression using DCT and DWT based on
a lossless compression technique which assigns a prefix performance metrics such as PSNR and Compression
code identified as the Huffman code to the two input Ratio.
signals. The basic idea of the Huffman coding is to PSNR (Peak Signal-to-Noise Ratio):
assign to each of the source character alphabet a fixed PSNR is calculated for degradation of video after
number of bits which does not exceed the memory the compression. PSNR is ratio between the maximum
capability.
possible power of a signal and the power of corrupting MSE (Mean Square Error):
noise. The MSE represents the increasing squared error
between the compressed and the original video. The
PSNR= 10 ∗ (2)

294
mean square error is the average of the squared errors local properties within a small area, and this is extremely
between actual and estimated readings in a data sample. beneficial for the non-stationary statistics of video
signal.
MSE (3)
5. Conclusion
CR (Compression Ratio): This paper presents a comparative analysis of
The compression ratio between two videos: the various video compression based on transform based
original video and the compressed video encoder, is techniques. The observation of the results shows that the
measured by equation as below: Hybrid (DCT – DWT) Transform yields high Peak
signal noise ratio (PSNR) and compression ratio for
effective transmission.
CR = (4)
References
Table1 shows comparison result of four video [1] T. Bernatin and G. Sundari ―Video Compression Based on
compression methods based on transforms DCT and Hybrid Transform and Quantization with Huffman Coding
for Video Codec‖, International Conference on Control,
DWT.
Instrumentation, Communication and Computational
The four hybrid methods investigated in Table 1 are Technologies (ICCICCT) , pp.452-456, 2014.
Hybrid Transform using Huffman Coding [1], Hybrid [2] L.Escalin Tresa and Dr. M. Sundararajan ―Video
DWT-DCT algorithm [2], Discrete Cosine Transform Compression Using Hybrid DWT-DCT Algorithm‖,
(DCT) [3], Hybrid (DCT-DWT) algorithm [4]. International Journal of Emerging Technology and
Advanced Engineering, Volume 4, Issue 7, pp.918-922,
Table 1: Comparative analysis of transform based July 2014.
video compression [3] S.S. Wadd and S. B. Patil ― Video Compression Using
DCT‖, International Journal of Advanced Research in
Compression Computer Science and Software Engineering , Volume 4,
Methods Video Frames PSNR Issue 9, pp.394-398, September 2014.
Ratio
Method [1] Xylo phone 70.03 97.48 [4] S.V.Phakade and Harish Patil ―Video Compression Using
Hybrid DCT-DWT Algorithm‖ International Research
Method [2] Xylo phone 36.35 77.66 Journal of Engineering and Technology (IRJET), Volume:
Method [3] Misamerica.avi 34.35 147.51 03, pp.2758-2761, 05-May-2016.
[5] Thazni Aziz and D. Raveena Judie Dolly ― Motion
Method[4] Xylo phone 36.34 11.50
Estimation and Motion Compensated Video Compression
Using DCT And DWT‖, International Journal of Emerging
The experimental result shows that the Hybrid Technology and Advanced Engineering Volume 2, Issue
Transform using Huffman coding [1] method yields 12, December 2012, pp.667-671.
higher PSNR (70.03db) and Compression Ratio (97.48) [6] Navpreet Saroya, and Prabhpreet Kaur. "Analysis of
among other existing methods. IMAGE COMPRESSION Algorithm Using DCT and DWT
From the investigation about the developments of Transforms." International Journal of Advanced Research in
video compression technologies in this study, we can see Computer Science and Software Engineering Volume 4
Issue , pp.897- 900, February 2014.
that many compression improvements were made by [7] Deshlahra, Archana, G. S. Shirnewar, and A. K. Sahoo. "A
Hybrid transform methods. An adaptive algorithm can Comparative Study of DCT, DWT & Hybrid (DCT-DWT)
fit its behavior to the changes of video content, or the Transform." - 2013.

295

View publication stats

Vous aimerez peut-être aussi