NJIT
Adaptive transform coding of video with motion compensation and vector quantization
Abstract
dc:description.abstract<p>This thesis proposes an efficient transform coding scheme for the low bit rate video sequence. The redundancy within adjacent video frame is exploited by motion compensated interframe prediction using the efficient independent orthogonal search technique, and the motion compensated frame difference (MCFD) signal is transform coding by 1-D DCT, 1-D HDCT and 2-D DCT, the coefficients are zonal masked, and vector quantizatied adaptively. The encoded MCFD as well as the motion information and variance grouping (side informations) are transmitted to the receiver to reconstruct the frame. The transform MCFD signal is classified into the four groups according to the coefficient band variance. A multi-codebook vector quantizer is designed based on all groups. The overall average PSNR for 1-D DCT, 1-D HDCT and 2-D DCT are 36.14dB, 36.26dB, 36.30dB respectively. The results are obtained for the "CINDY" video sequences with the average bit rate 0.3318, 0.3397, 0.2655 bits/pixel, over 39 frames. It is shown that the vector codebook on the set of training sequence generated by the MCFD signal is more independent, general than the original video signal. It is also shown that the performance of 1-D HDCT technique for MCFD signal coding is as good as 2-D DCT technique. Scene change detector algorithm based on cross correlation criterion is proposed in this thesis for the "MIX" video sequence.</p>
Degree
thesis:*- Name thesis:degree_name
- Master of Science in Electrical Engineering - (M.S.)
- Discipline thesis:degree_discipline
- Electrical Engineering
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Chien, Jung-Hui
- Contributors dc:contributor
-
- Ali N. Akansu
- Chung H. Lu
- Nirwan Ansari
Subjects
dc:subject × 4Identifiers
dc:identifier.*- Repository record dc:identifier
- https://digitalcommons.njit.edu/theses/2738