一种基于内容与结构相似性的图像质量评价方法  

AN IMAGE QUALITY ASSESSMENT METHOD BASED ON CONTENT AND STRUCTURE SIMILARITY

作　　者：付燕[1] 马大江 

机构地区：[1]西安科技大学计算机科学与技术学院,陕西西安710054

出　　处：《计算机应用与软件》2014年第10期220-223,共4页Computer Applications and Software

摘　　要：图像质量评价的客观方法弥补了主观方法的费时费力,不具有移植性的问题。提出一种基于图像内容划分,结合结构相似性理论SSIM(Structure Similarity Image Index)的全参考客观图像质量评价方法 PSSIM(Partition SSIM)。该方法结合图像梯度和图像划分,将图像边界区域划分为保持边界区和变化边界区。另一方面摒弃经验参数,结合LIVE图像库中数据计算出保持边界区域权重参数λ=0.35时,主客观评价一致性最高。实验表明,PSSIM方法具有较强的通用性,较好地符合人的主观视觉感知质量。Objective method of image quality assessment has overcome the problems of time and labour-consuming and without portability that the subjective method has. In this paper, we propose an objective image quality assessment method with full-reference, named as PSSIM, which is based on image content partitioning and combines the structure similarity （SSIM） theory. The method integrates the image gradient with image partitioning, divides the image boundary region into boundary preserving area and boundary varying area. On the other hand, it gets rid of the empirical parameters and calculates in conjunction with the data in LIVE image library that when the weight parameter of the boundary preserving area λ= 0.35, the consistency of the subjective and objective assessment reaches the highest. Experiment shows that the PSSIM has quite high universality, and well complies with the quality of human subjective visual perception.

关 键 词：客观质量评价 图像划分 结构相似性(SSIM) 图像内容 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]