Reducing the real time H.264 Decoder’s Decoded Picture Buffer Size by removing the pixels that are no longer used in time

碩士 === 國立交通大學 === 資訊科學與工程研究所 === 101 === In the H.264 video compression standard decoder, the Decode Picture Buffer (DPB) consumes over 95% memory space and the chip area. If we can reduce the DPB size, it can achieve the reduction in chip area and power consumption especial the embedded video decod...

Full description

Bibliographic Details
Main Authors: Liu, Ying-Heng, 劉盈亨
Other Authors: Fu, Hsin-Chia
Format: Others
Language:en_US
Published: 2012
Online Access:http://ndltd.ncl.edu.tw/handle/42497853033054942391
Description
Summary:碩士 === 國立交通大學 === 資訊科學與工程研究所 === 101 === In the H.264 video compression standard decoder, the Decode Picture Buffer (DPB) consumes over 95% memory space and the chip area. If we can reduce the DPB size, it can achieve the reduction in chip area and power consumption especial the embedded video decoder. In this paper, we focus on the analysis of the reference behavior among the frames. We find that there are 60% to 80 % of pixels of each frame which are no longer used in specific time. Therefore, we can remove the unused pixels after the specific time to reduce the DPB size. Besides, according to the properties of streaming data, we proposed the suitable decoding pipeline to achieve the steady frame output rate. In order to reduce the DPB size, we propose the suitable time to remove the unused pixels for each frame, a pre-decoding mechanism to obtain the lifetime information of pixels, and new memory architecture to store the frames. Without affecting the decoding performance, our design is able to achieve a 62.4% reduction in storage, with a decrease in image quality less than 4.1%. Compared to the traditional H.264 decoder, our design can achieve a 66.7% reduction in chip area and a 56.2% reduction in power consumption.