Structure-based motion estimation for video compression

摘要

Motion estimation is a crucial process in video compression algorithms, aiming at reducing the temporal redundancy in video data. Traditional block-based motion estimation methods lead to satisfactory results in different situations but fail in others when dealing with high local structural variations. In this paper, a structure-based approach for motion estimation is proposed, taking into account structural information of the frames throughout the video sequence. Results show that using the proposed method, relevant local information can be better reconstructed comparing to the standard motion estimation approach, thus reducing the residual error, which may lead to improved rate-distortion performance of video compression algorithms.