Development of a new filter for MediaSDK encoder

Annotation to the thesis Topic: " Development of a new filter for MediaSDK encoder" The author of the work: a student of the Faculty of Informatics, Mathematics and Computer Science of the National Research University "Higher School of Economics" Losev Ivan Sergeevich Subject of research: video coding The object of study: noise reduction filter The purpose of the study is to analyze the possibilities of CPU and GPU optimization of the noise reduction algorithm in order to decide on the inclusion of this filter in the industrial video encoding library MediaSDK. The volume of the thesis is 36 pages and includes: an introductory part, a description of the algorithm, 4 chapters of its optimization, summing up and a list of sources. The relevance of the work is due to possible improvements in the efficiency of the encoders, because after applying the noise-canceling filter, unnecessary information is removed from the frames (noise), the interframe difference is also reduced, which allows encoding video with a lower bit rate (the number of bits used to transmit / process information per unit time). In addition, the visual quality of the video improves and its clarity increases. Also, there is a need to check the optimization possibilities of the chosen algorithm, since it is necessary to take into account not only the degree of improvement in video quality, but also the speed of the chosen approach, since in real-time systems the absence of delays in video playback is more important than its quality. The result of the work is a comparative analysis of the use of CPU and GPU optimized versions of the selected algorithm on various video formats.