重视边缘区域的结构相似度图像质量评价

Structural similarity highlighting edge regions for image quality assessment

基于结构相似度的图像质量评价方法简单高效、准确性较高,但是对严重失真和交叉失真类型评价不够准确。考虑到边缘是图像的主要信息和能量成分,人眼对边缘信息的丢失更加关注,提出一种重视边缘区域的结构相似度图像质量评价方法(HESSIM),采用动态阈值(Otsu)法提取边缘区域,结合JND确定边缘区域的明显失真,并对其予以重视。实验结果表明,HESSIM比SSIM有更准确的评价,特别是对模糊类失真和噪声类失真的评价,HESSIM的优越性更加明显。

Structural similarity （SSIM） is an image quality assessment algorithm with the advantage of simplicity, high efficiency and better consistency with human subjectivity. However, it often fails when measuring badly distorted or cross distortion images. In this paper, an improved algorithm called structural similarity highlighting edge regions （HESSIM） is proposed based on the idea that edges are the most important information in an image. The humans eye is very sensible for distorted edge information. In the proposed HESSIM, the edge regions are first divided from an image by Otsu＇ s method, then those with obviously perceptual distortion are chosen by the JND model, and their distortion measures are highlighted. Experimental results show that HESSIM is more consistent with HVS than SSIM, especially for distorted images which are blurred or comtaminated with white noise.