一种图像增强空域客观评价方法  

An Objective Evaluation Method of Image Enhancement

作　　者：汪祖辉[1] 孙刘杰[1] 邵雪[1] 

机构地区：[1]上海理工大学出版印刷与艺术设计学院,上海200093

出　　处：《包装学报》2016年第4期30-34,共5页Packaging Journal

基　　金：上海市教委科研创新基金资助重点项目(13ZZ111)

摘　　要：针对图像增强效果常用主观评价而没法定量客观评价,提出一种空域结合BRISQUE和JND的图像增强客观评价方法。该方法空域内将测试图像分别进行BRISQUE失真评分和JND视觉评分,然后将所得分数以0.5的权重进行加权处理,所得总分即为增强后图像的客观评价得分。为验证所提方法的有效性,进行了系列实验:首先,认证图像失真相同背景亮度增大BRISQUE值不变,而视觉JND值随之改变;其次,认证相同背景亮度不同失真图像的JND值不变,而BRISQUE值不同;最后对增强后的图像应用所提算法进行评分,得到Score最高分为0.790 5,与主观评价结果一致,而PSNR、SSIM的评分最高为∞和1,但都是和原图像本身比较,不能表明图像增强效果。从而证明所提算法能够定量地对图像增强进行客观评价。Image enhancement results were often assessed by subjective evaluation and the assessment could not be quantitative and objective. A new method of objective evaluation for image enhancement based on BRISQUE and JND was presented. The BRISQUE distortion score and JND visual score were performed in the spatial domain, and the score obtained by the weighted total score of 0.5 would be the objective evaluation score of the post image. In order to verify the effectiveness of the proposed method, a series of experiments were carried out. Firstly, to certify image distortion with same background brightness by increasing BRISQUE value, the visual JND value was changed accordingly; Secondly, to certify with the same background brightness of the same JND value of different distortion images, the BRISQUE values were different. Finally, the evaluation of enhanced image based on the proposed algorithm achieved the highest score of 0.790 5,and the result was consistent with subjective evaluation results. While PSNR, SSIM score was the highest of infinity and1, both were compared with the original image, which could not show the effect of image enhancement. It showed that the proposed algorithm could objectively evaluate the image enhancement.

关 键 词：BRISQUE 人眼感知亮度阈值 人眼视觉 图像增强 客观评价 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]