Document Type

Thesis

Date of Award

5-31-1989

Degree Name

Master of Science in Electrical Engineering - (M.S.)

Department

Electrical Engineering

First Advisor

Ali N. Akansu

Second Advisor

Chung H. Lu

Third Advisor

Nirwan Ansari

Abstract

An efficient scheme for video sequence coding is presented here. The interframe redundancy present in the video sequence is exploited by motion compensation using the efficient independent orthogonal search technique, which requires least number of computations in the worst case. The motion compensated frame difference signal is encoded with a new technique — subband adaptive vector quantization.

Subband coding like transform coding results in signals of unequal energy patterns, which is efficiently utilized for bit allocation. In addition, the subband approach offers means for efficient (but suboptimal) Vector Quantization (VQ) by splitting the signal into several signals. By rate distortion theory, vector quantization performs better than the scalar quantization. As subband signals are uncorrelated, the loss in performance due to independently quantizing each band is small.

The subbands are vector quantized adaptively. Subbands are divided into 4x4 subblocks, grouped according to their variances and are quantized using tailor made codebooks. By this different featured signal sources will employ different codebooks. An average SNRpp = 37.82 dB, is obtained for highly active monochrome video sequence 'CINDY' with an average bit rate of 0.36 bits/pixel. The subjective tests also show that the picture quality is good at that bit rate. Additional bit rate is obtained if the motion information is entropy encoded.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.