Development and analysis of block-based motion estimation techniques for efficient video compression

Date

1992

Authors

Chana, Jatinder Singh

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

Block-based Motion Estimation and Compensation (ME/MC) is a popular inter­frame coding technique and is used in video compression systems, e.g. the MPEG video coding standard, to exploit similarities between adjacent frames in typical video sequences. It provides additional compression and also improves fidelity of decoded video in low bit-rate applications. In a typical application, the frame being coded is parti­tioned into fixed sizes blocks and only motion information of these blocks, with respect to a reference frame, is coded. Search range is a parameter that directly controls the accu­racy of estimated motion information and the computational effort required for motion estimation. Exhaustive ME using larger search range results in greater accuracy at the expense of greater computation. Although exhaustive ME is very accurate, it is also very demanding in terms of computation. Efficient algorithms, collectively referred to as Block Matching Algorithms (BMAs), have been proposed as alternatives to exhaustive ME to provide computation­ally efficient ME. BMAs use efficient search algorithms to obtain motion information and require substantially less computation. A comprehensive evaluation of a select num­ber of such BMAs is present in this thesis. The BMAs selected for evaluation are repre­sentative of the general approaches used in designing such algorithms and are evaluated based on their estimation accuracy and computational efficiency. The evaluation is per­formed using two real video sequences that contain motion present in typical video sequences. The evaluation results indicate that BMAs are effective only for ME involv­ing small motion displacements. A predictive motion search method, the Adaptive Search Window Method (ASWM), is proposed to overcome this limitation. This method dynamically controls the location and search range for ME for individual blocks based on motion information of surrounding blocks. If correlation is detected in surrounding motion information, a pre­diction is made for the corresponding block and ME is performed around the predicted block using a reduced search range. The reduction in the search range results in accurate ME using BMAs or, alternatively, in less computation for exhaustive ME. The application of BMAs incorporating ASWM for ME is evaluated and a significant improvement in estimation accuracy, compared to application of BMAs without ASWM, is observed for ME involving larger motion displacements. Two BMAs incorporating ASWM are also implemented in a software video coder and compression gains over respective BMAs without ASWM are demonstrated.

Description

Keywords

UN SDG 16: Peace, Justice, and Strong Institutions

Citation