DocumentCode :
2281053
Title :
LCU-based adaptive interpolation filter
Author :
Matsuo, Shohei ; Takamura, Seishi ; Jozawa, Hirohisa
Author_Institution :
NTT Cyber Space Labs., NTT Corp., Yokosuka, Japan
fYear :
2012
fDate :
7-9 May 2012
Firstpage :
393
Lastpage :
396
Abstract :
Interpolation filter that supports motion estimation with fractional-pel accuracy is one of important coding tools for improving efficiency of inter frame coding. A lot of interpolation filters were proposed to standardization activity of Joint Collaborative Team on Video Coding (JCT-VC) to establish a next video coding standard called High Efficiency Video Coding (HEVC). DCT-based interpolation filter (DCT-IF) is currently employed in the HEVC test model (HM) software. However, the values of the filter coefficients are constant regardless of the characteristic of the input video. The authors proposed a region-based adaptive interpolation filter (RBAIF) that optimizes the filter coefficients on a region-by-region basis. In this paper, some modifications are added to the conventional RBAIF. Simulation results showed that, compared to the HM software, the proposed filter offers about 0.5% and 0.4% average coding gains for luminance and chrominance, respectively. Runtime of the encoder and the decoder were 101.4% and 102.5% on average, respectively.
Keywords :
adaptive filters; data compression; interpolation; motion estimation; video coding; DCT-IF; DCT-based interpolation filter; HEVC test model software; HM software; JCT-VC; LCU-based adaptive interpolation filter; conventional RBAIF; fractional-pel accuracy; high efficiency video coding; inter frame coding; joint collaborative team; motion estimation; region-based adaptive interpolation filter; Decoding; Encoding; Indexes; Interpolation; Proposals; Software; Video coding;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Picture Coding Symposium (PCS), 2012
Conference_Location :
Krakow
Print_ISBN :
978-1-4577-2047-5
Electronic_ISBN :
978-1-4577-2048-2
Type :
conf
DOI :
10.1109/PCS.2012.6213372
Filename :
6213372
Link To Document :
بازگشت