DocumentCode
128547
Title
Fast intra-prediction mode decision algorithm for high efficieny video coding
Author
Shiping Zhu ; Dongyu Zhao
Author_Institution
Dept. of Meas. Control & Inf. Technol., Beihang Univ., Beijing, China
fYear
2014
fDate
9-11 June 2014
Firstpage
936
Lastpage
939
Abstract
In this paper, we propose a fast intra-prediction mode decision algorithm based on the probability statistics of the best prediction mode and the relevance of the neighboring blocks for high efficiency video coding (HEVC) to reduce its encoding complexity. The intra prediction candidates are reduced before Rate Distortion Optimization (RDO) procedure by two steps of rough mode decision. Sum of Absolute Difference (SAD) of the residuals is used in the first step and modes with smaller SAD values are retained. We execute the Hadamard transform on prediction residuals from the rough modes to derive the Sum of Absolute Hadamard Transformed Difference (SATD), and then the rate-distortion cost function based on SATD is used in the second rough mode decision. Finally the most probable mode (MPM) is considered for RDO process to select the best intra prediction mode. Experiments results demonstrate that our proposed algorithm is quite efficient for intra prediction speedup. Compared with HM 4.0, the proposed algorithm reduces the encoding time by 30.18% while the PSNR decrease is negligible, 0.06 dB on average. And the bitrate is only increased by 1.97% on average.
Keywords
Hadamard transforms; encoding; probability; video coding; absolute Hadamard transformed difference; absolute difference sum; encoding complexity; high efficiency video coding; intra-prediction mode decision algorithm; most probable mode; probability statistics; rate distortion optimization; rough mode decision; Complexity theory; Correlation; Encoding; PSNR; Prediction algorithms; Rate-distortion; Video coding; HEVC; intra prediction; mode decision; video coding;
fLanguage
English
Publisher
ieee
Conference_Titel
Industrial Electronics and Applications (ICIEA), 2014 IEEE 9th Conference on
Conference_Location
Hangzhou
Print_ISBN
978-1-4799-4316-6
Type
conf
DOI
10.1109/ICIEA.2014.6931297
Filename
6931297
Link To Document