作　　者： ; ; ;

机构地区： 浙江科技学院信息与电子工程学院

出　　处： 《高技术通讯》 2012年第11期1117-1123,共7页

摘　　要： 通过分析人眼对不同图像类型块的敏感度和支持向量回归（SVR）的特性，提出了一种基于块内容和SVR的图像质量评价模型。该模型首先根据图像子块的交流能量的关系，自适应地把图像分成边缘块、纹理块和平坦块3种类型；接着对不同类型块的失真程度采用图像块质量评价方法进行度量，得到图像各类型块的质量分数；其后利用SVR来学习3种类型块的质量分数与主观评价值的关系；最后用训练好的SVR完成图像质量预测。在实验过程中，图像各类型块失真程度评价采用了奇异值分解（SVD）方法、方向投影（DP）方法和投影能量（PE）方法。从实验结果可以得到，该模型克服了图像3种类型子块质量分数的权值讨论，从而使SVD方法、DP方法和PE方法的评价性能得到了有效提高，这表明该模型对图像块质量评价方法性能提高具有一定的普适性。 An objective image quality assessment （IQA） model based on image block content and support vector regression （SVR） is proposed for image signal processing based on analyzing the sensitivity of human visual perception to different types of image block and the characteristic of SVR. The model divides an image into the blocks of edge, texture and smooth adaptively according to the alternating current （AC） energy relation among the image＇ s sub- blocks, then the block IQA method is adopted to measure the quality of three types of image block, and then the relativity of image block quality scores and subjective quality scores can be learned by the training procedure of SVR, and finally, the image quality is predicted by the trained SVR. In the course of the experiment, the singular value decomposition （SVD） method, the directional projection-based （DP） method and the projection energy- based （PE） method were applied to measuring the quality of three types of image block. The experimental results show that this model can overcome the problem of discussing different block weights, and improve the performance of the methods of SVD, DP and PE effectively. Therefore, this model is universal in some extent when used for im- proving the performance of the block IQA methods.

关 键 词： 图像质量评价 人眼视觉特性 图像块内容 支持向量回归

领　　域： [自动化与计算机技术] [自动化与计算机技术]