一种基于自适应分块八叉树颜色量化的图像压缩技术  

An image compression technique based on adaptive block-based octree color quantization

作　　者：吴振华 沈虎峻 公佐权 冯平 龚彤艳 邓明森[1] WU Zhen-hua;SHEN Hu-jun;GONG Zuo-quan;FENG Ping;GONG Tong-yan;DENG Ming-sen(School of Information,Guizhou University of Finance and Economics,Guiyang 550025;Guizhou Provincial Key Laboratory of Computational Nano-Material Science,Guizhou Education University,Guiyang 550018;State Key Laboratory of Computer Architecture,Institute of Computing Technology,Chinese Academy of Sciences,Beijing 100180,China)

机构地区：[1]贵州财经大学信息学院,贵州贵阳550025 [2]贵州师范学院贵州省纳米材料模拟与计算重点实验室,贵州贵阳550018 [3]中国科学院计算研究所计算机体系结构国家重点实验室,北京100180

出　　处：《计算机工程与科学》2020年第2期291-298,共8页Computer Engineering & Science

基　　金：贵州省科技计划项目(QKHJC[2016]1109号);商务部-贵州财经大学联合基金(2016SWBZD04)。

摘　　要：颜色量化(CQ)是减少图像颜色数量的过程,已广泛用于图像压缩。基于八叉树的颜色量化(OCQ)因其编码效率高、内存使用低和调色板选择效果良好而被认为是最流行的CQ算法之一。然而,OCQ应用的一个严峻挑战是如何有效地管理关键的本地颜色。提出了一种基于分块的自适应八叉树颜色量化(AB-OCQ)算法,实验结果表明,与传统的OCQ算法相比,由于增加了对局部颜色的适当处理,AB-OCQ可以显著提高图像质量。在图像压缩比方面,AB-OCQ的综合性能也优于OCQ的。同时,和主流图像文件格式相比,AB-OCQ算法可以在保持压缩的前提下拥有随机访问图像像素数据的特性,该特性能让应用程序在同等内存下存储更多的图像数据,为提高应用程序的效率提供了一种方法。Color Quantization(CQ)is a process of reducing the number of colors in an image and has been widely used in image compression.Octree-based Color Quantization(OCQ)is considered to be one of the most popular CQ algorithms due to its high encoding efficiency,low memory usage,and good color palette selection.However,a serious challenge for OCQ applications is how to efficiently manage key local colors.In this paper,we propose an adaptive block-based octree color quantization(AB-OCQ)approach to overcome the challenge.Our results show that AB-OCQ can significantly improve the image quality in comparison to the traditional OCQ approach,owing to the proper treatment of some local co-lors.In image compression ratio,AB-OCQ has better comprehensive performance than OCQ.At the same time,compared with the popular image file format,AB-OCQ has the feature that it can randomly access the image pixel data while maintaining the compression ratio.This feature allows the application to store more image data under the same memory,which provides a way to increase the efficiency of application.

关 键 词：图像压缩技术 颜色量化 八叉树 自适应分块量化 AB-OCQ 

分 类 号：TP37[自动化与计算机技术—计算机系统结构]