Title :
Efficient coding strategy for HEVC performance improvement by exploiting motion features
Author :
Podder, Pallab Kanti ; Paul, Manoranjan ; Murshed, Manzur
Author_Institution :
Sch. of Comput. & Math., Charles Sturt Univ., Bathurst, NSW, Australia
Abstract :
The striking feature of High Efficiency Video Coding (HEVC) Standard is emphasized by 50% bit-rate reduction compared to its predecessor H.264/AVC while keeping the same perceptual image quality. The time complexity - a congenital issue of HEVC has also increased to intensify the compression ratio. However, it is really a demanding task for the researchers to reduce the encoding time while preserving expected quality of the video sequences. Our contribution is to trim down the computational time by efficient selection of appropriate block-partitioning modes in HEVC using motion features based on phase-correlation. In this paper, we use phase-correlation between current and reference blocks to extract three motion features and combine them to determine binary motion pattern of the current block. The motion pattern is then matched against a codebook of predefined pattern templates to determine a subset of the inter-modes. Only the selected modes are exhaustively motion estimated and compensated for a coding unit. The experimental outcomes demonstrate that the average computational time can be down scaled by 30% of the HEVC while providing improved rate-distortion performance.
Keywords :
data compression; feature extraction; image matching; image sequences; motion estimation; rate distortion theory; video coding; HEVC performance improvement; binary motion pattern determination; bit-rate reduction; block-partitioning mode selection; high efficiency video coding; motion estimation; motion feature extraction; motion pattern matching; perceptual image quality; phase-correlation; rate-distortion performance; time complexity; video compression ratio; video sequence; Bit rate; Correlation; Cost function; Encoding; Feature extraction; Standards; Video coding; HEVC; Intermode Selection; Motion Estimation; Motion Features; Phase-Correlation;
Conference_Titel :
Acoustics, Speech and Signal Processing (ICASSP), 2015 IEEE International Conference on
Conference_Location :
South Brisbane, QLD
DOI :
10.1109/ICASSP.2015.7178203