近無失真2倍顯示壓縮設計

Abstract

隨著消費性產品顯示尺寸的增大，顯示功能所佔的功率成為一個低功耗要求的關鍵問題。本論文提出一個簡單而有效率的近無失真2倍顯示壓縮引擎。此引擎針對自然影像使用適應性差動混合Golomb-Rice編碼、內容基礎影像使用索引式編碼、以及複雜影像使用截短式編碼,以確保2倍壓縮率。依照此機制，對任意尺寸的影像，在近無失真、40dB PSNR的影像品質之下，記憶體所需頻寬至少可節省50%以上。硬體設計使用0.18um製程，使用的邏輯閘數目為23K，在100Mhz的操作頻率下可支援每秒30張尺寸為1920x1080的影片，消耗功率為7.4mW。

With increasing display size in modern consumer devices, the corresponding power consumption is one of the critical issues for low power concerns. To save bus transaction power, this thesis presents a simple yet efficient near lossless two times embedded compression engine for display. The engine uses a hybrid approach that combines adaptive differential coding with Golomb-Rice coding for nature images, index based coding for screen content and truncated coding to ensure two times compression ratio. With above mechanisms, the memory bandwidth can be saved by at least 50% for any video size and content with near lossless quality, greater than 40dB PSNR. The hardware design with 0.18um CMOS process can easily support 1080P@30fps video encoding and decoding with only 23K gate count at 100MHz with 7.4mW power consumption.