New hybrid coding for video sequences

Pao Yue-kong Library Electronic Theses Database

New hybrid coding for video sequences

 

Author: Wong, Kai-yin
Title: New hybrid coding for video sequences
Degree: M.Phil.
Year: 2007
Subject: Hong Kong Polytechnic University -- Dissertations.
Video compression.
Image transmission.
Streaming technology (Telecommunications)
Department: Dept. of Electronic and Information Engineering
Pages: xix, 176 leaves : ill. ; 30 cm.
Language: English
InnoPac Record: http://library.polyu.edu.hk/record=b2116786
URI: http://theses.lib.polyu.edu.hk/handle/200/483
Abstract: During the encoding process of traditional video coding standards, the motion estimation consumes most of the computational effort. For the wavelet video coding, this problem becomes more serious since motion estimation in the wavelet domain involves floating-point computation. Hence, it is necessary to develop some fast motion estimation algorithms in wavelet domain in order to reduce the computational burden of the wavelet video encoder during encoding procedure. One of the traditional motion estimation algorithms in wavelet domain is to make use of the correlations among the corresponding subbands in the wavelet pyramid to enhance the speed of motion estimation. This algorithm is entitled as Multi-resolution Motion Estimation (MRME) algorithm which is based on the fact that an object in a subband of the lowest resolution level actually specifies the same object in the subbands of the higher resolution levels. Thus, the computational complexity of motion estimation in the wavelet domain can be reduced significantly by exploiting the relationships between the subbands of different resolution levels. On the other hand, the pixels with similar matching error tend to group in a cluster in both spatial and wavelet domains. Besides, a cluster which appears in a certain position of a subband at the lowest resolution level and there also exists a cluster in the similar position of the corresponding subbands at the remaining higher resolution levels. Thus, the Clustered Pixel Matching Error for Partial Distortion Search (CPME-PDS) algorithm can be applied in the MRME scheme to further improve the speed of motion estimation in the wavelet domain. The CPME-PDS involves the sorting procedure in order to obtain the coefficients with large matching errors and the sorting is required to perform in every subband at each decomposition level. As we found that the clustering property is in a hierarchical nature in the wavelet pyramid, the sorting order in the subband of the highest-resolution level can be re-used to predict the sorting order of the subbands in the lower-resolution levels. Since the sorting operations are only carried out in the three subbands at the high-resolution level only, so the computational effort for motion estimation can be further reduced. From the experimental results, the proposed algorithm (Backward CPME-PDS) can achieve speed-up factors from 2 to 5 and from 1.1 to 1.2 as compared to the Full Search Algorithm (FSA) and Partial Distortion Search (PDS) algorithm respectively. Recently, the three dimensional discrete wavelet transform (3D-DWT) video coder becomes more popular since it can attain both spatial and temporal scalabilities. It involves motion estimation in the wavelet decomposition, so the computational burden for motion estimation is still a major concern. Due to the fact that there exists high spatial and temporal correlations between the motion vectors in the neighbouring blocks in a frame and between the motion vector fields of the low frequency frames at successive temporal levels, the computational complexity of motion estimation can be reduced considerably. By making use of these correlations, an accurate motion estimation predictor can be obtained and a refinement process is performed within the reduced search area based on the accurate motion predictor. From the experimental results, the proposed algorithm can achieve a speed-up factor of 3 to 5 as compared with the FSA using the Haar and Bi-orthogonal 5/3 kernels during temporal decomposition. Besides, quality of the reconstructed video sequence using the proposed algorithm is comparable to that of the FSA. The Embedded Zerotree Wavelet (EZW) coding algorithm is often used to encode the wavelet coefficients. We proposed a modified EZW algorithm to improve the coding gain of the wavelet video coder by discarding some less important wavelet coefficients using a new criterion. Experimental results show that the proposed EZW algorithm can improve 0.2 to 0.5 bit per pixel (bpp) as compared with the original EZW algorithm for the same PSNR value in lossy coding and its reconstructed quality is comparable to that of the conventional EZW algorithm. Furthermore, the proposed algorithm can apply in the Set Partition Embedded Block Coding (SPECK) algorithm to further enhance the compression efficiency from 0.1 to 0.5 bit per pixel (bpp) as compared with the minimum subband approach applied in the SPECK algorithm for the same PSNR value in lossy coding and its visual quality of the reconstructed image using the proposed algorithm can be preserved.

Files in this item

Files Size Format
b21167862.pdf 8.136Mb PDF
Copyright Undertaking
As a bona fide Library user, I declare that:
  1. I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
  2. I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
  3. I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.

     

Quick Search

Browse

More Information