Study of blocking effect elimination methods by means of intraframe video sequence interpolation

Abstract

The paper deals with image interpolation methods and their applicability to eliminate some of the artifacts related to both the dynamic properties of objects in video sequences and algorithms used in the order of encoding steps. The main drawback of existing methods is the high computational complexity, unacceptable in video processing. Interpolation of signal samples for blocking - effect elimination at the output of the convertion encoding is proposed as a part of the study. It was necessary to develop methods for improvement of compression ratio and quality of the reconstructed video data by blocking effect elimination on the borders of the segments by intraframe interpolating of video sequence segments. The main point of developed methods is an adaptive recursive algorithm application with adaptive-sized interpolation kernel both with and without the brightness gradient consideration at the boundaries of objects and video sequence blocks. Within theoretical part of the research, methods of information theory (RD-theory and data redundancy elimination), methods of pattern recognition and digital signal processing, as well as methods of probability theory are used. Within experimental part of the research, software implementation of compression algorithms with subsequent comparison of the implemented algorithms with the existing ones was carried out. Proposed methods were compared with the simple averaging algorithm and the adaptive algorithm of central counting interpolation. The advantage of the algorithm based on the adaptive kernel size selection interpolation is in compression ratio increasing by 30%, and the advantage of the modified algorithm based on the adaptive interpolation kernel size selection is in the compression ratio increasing by 35% in comparison with existing algorithms, interpolation and quality of the reconstructed video sequence improving by 3% compared to the one compressed without interpolation. The findings will be widely used in video processing tasks, various codecs of video compression and streaming systems.