Predictive Encoding of Digitized TV Pictures

Author: Sharma, Dhiraj Kumar

Year: 1975

Degree: Dissertation (Ph.D.)

Advisor: Pierce, John Robinson

Committee Member: Unknown, Unknown

Option: Electrical Engineering; Engineering

DOI: 10.7907/t5n9-3a07

Abstract

In this report, we consider the problem of efficient intra-frame encoding of digitized TV pictures using Differential Pulse Code Modulation (DPCM) type encoders. Our goal is to reduce the average number of bits used to encode a pixel while subjective quality of the picture still remains acceptable.

To this end, we propose some new predictors and compare them with the existing ones using the error criteria "essential maximum" and "variance" of the prediction errors. The two criteria have also been compared with respect to the subjective quality of the final encoded pictures. We propose to design the quantizers using "mean n-th power" distortion criterion. Effects of "n" on bit rate and subjective quality of the encoded picture have been studied as it varies from 1 to 6. In this way, we achieve reductions up to 6 to 1 in the bit rate with little degradation in the picture quality. We also consider the use of second and third order entropies of the encoded pictures to reduce the bit rate.

Next, we adapt the encoding procedure to local structural variations in the picture by first segmenting it in three areas of varying detail and then using the concepts developed for non-adaptive encoding to encode the three areas differently. The resulting encoder yields better picture quality than before and can be used to encode a broad class of pictures.

Files