Title :
Approximate Computing in Chrominance Cache for Image/Video Processing
Author :
Fei Qiao ; Ni Zhou ; Yuanchang Chen ; Huazhong Yang
Author_Institution :
Dept. of Electron. Eng., Tsinghua Univ., Beijing, China
Abstract :
An approximate computing method for chrominance data of the video/image sequences is proposed in this paper, which is implemented in one hybrid cache architecture. This approach fully exploits the fact that the sensitivity of chrominance is much lower than that of luminance to reduce off-chip memory access of chrominance data. For the 4:2:0 image format, the proposed approximate computing method can achieve 25% reduction in off-chip memory access with the 2×2 basic approximate unit. Moreover, for the image formats of 4:2:2 and 4:4:4, it has about 37.5% and 50% access reduction of off-chip memory, respectively, with acceptable degradation of image quality to human visual system.
Keywords :
cache storage; image sequences; video signal processing; approximate computing method; chrominance cache; human visual system; hybrid cache architecture; image processing; image quality; image sequences; off-chip memory access; video processing; video sequences; Codecs; Memory management; Multimedia communication; PSNR; Standards; Streaming media; approximate computing; chrominace; hybrid cache; image/video processing;
Conference_Titel :
Multimedia Big Data (BigMM), 2015 IEEE International Conference on
Conference_Location :
Beijing
Print_ISBN :
978-1-4799-8687-3
DOI :
10.1109/BigMM.2015.68