一种新的JPEG-LS压缩算法的低延迟实现方案  

A New Proposal of Low Latency Solution for JPEG-LS

作　　者：阙恒 王渊峰 武凤霞 QUE Heng WANG Yuanfeng WU Fengxia(Shanghai Zhaoxin Semiconductor Co., Ltd, Shanghai 201203, China.)

机构地区：[1]上海兆芯集成电路有限公司,上海201203

出　　处：《集成电路应用》2017年第6期23-27,共5页Application of IC

基　　金：国家核高基(核心电子器件;高端通用芯片及基础软件产品)专项基金(2014ZX01029101)

摘　　要：JPEG-LS的实现复杂度相对较低,同时拥有较高的压缩率,因而被广泛使用在连续色调图像的无损压缩领域。然而JPEG-LS的算法本身决定了它必须逐像素压缩,这种强依赖关系使得JPEG-LS单幅图像的压缩延迟过大。虽然利用分块的方法可以对JPEG-LS进行有效的加速,但分块压缩法无法降低一个子块内的压缩延迟,同时较小的分块也会造成压缩率的显著损失,因此不适用于对延迟敏感的系统。为了解决这个问题,提出了一种新的降低压缩延迟的方法,即推迟上下文更新法,它打破了JPEG-LS的强像素依赖,大大缩短了压缩延迟,即使不做图像分块,也使得像素间的并行压缩成为可能。JPEG-LS has a relatively low implementation cost while keeping high compression ratio. It is widely used in the domain of Iossless compression of continuous-tone images. However, the core algorithm of JPEG-LS itself needs to process those pixels of target images one by one, sequentially. This strong dependency between pixels makes the compression latency for one single image is very big. Though dividing the original image into sub blocks could speed up the compression task, thus reduce the latency, it is hard to reduce the compression latency for one single sub block. Meanwhile, smaller sub blocks will affect the image compression ratio remarkably, with negative effect. In a word, the sub-block method is not effective for latency-critical system. To solve this problem, a new proposal is made to reduce the compression latency, i.e., the idea of deferred context update. This new method breaks through the strong pixel dependency introduced by the original algorithm of JPEG- LS. Depending on the size of the deferred context update window, it could reduce the latency remarkably. Even without dividing the original image into sub blocks, this new method could compress multiple pixels from the original image in parallel.

关 键 词：压缩算法 推迟的上下文更新 并行压缩 低延迟压缩 帧缓存压缩 JPEG-LS 

分 类 号：TN402[电子电信—微电子学与固体电子学] TP391[自动化与计算机技术—计算机应用技术]