基于视觉注意计算模型的图像压缩新方法  

A New Image Compression Approach Based on Visual Attention Computational Model

作　　者：刘伟[1] 汪云九[2] 童勤业[1] 

机构地区：[1]浙江大学生物医学工程与仪器科学学院,杭州310027 [2]中国科学院生物物理研究所,北京100008

出　　处：《科技通报》2006年第6期775-780,共6页Bulletin of Science and Technology

基　　金：国家自然科学基金项目(30170267)

摘　　要：图像可以被分为重要的感兴趣区域和不重要的背景区域。提出了一种新的基于感兴趣区域的图像压缩方法。该方法首先使用一个基于视觉生理和心理物理实验基础的视觉注意计算模型计算图像中的感兴趣区域,然后用JPEG算法对感兴趣区域和背景区域采用不同的压缩比进行压缩。实验结果表明,该算法具有比JPEG算法更高的压缩效率,同时保证了图像良好的视觉效果。An image can be segmented into the region of interesting（ROI）, which is considered important, and the background, which is less important. In this paper, a new ROI based image compression algorithm is proposed. The first step of the algorithm is to find out the ROI in the image. A saliency-based bottom-up visual attention computational model which is motivated by visual physiological and psychophysical experimental results is used. The second step is encoding and which is based on the JPEG algorithm. The ROI of the image is compressed with a low compression ratio and the back ground with a high one. The reconstruction algorithm of the compressed image is like that of the JPEG algorithm. The algorithm proposed has higher compression ratio than JPEG algorithm and the image compressed has a perceptually high quality. We test our algorithm with lots of natural images and satisfy with the result.

关 键 词：视觉注意 感兴趣区域 图像压缩 特征抽取 

分 类 号：Q811.211[生物学—生物工程] TP751.1[自动化与计算机技术—检测技术与自动化装置]