基于结构相似性的图像质量评价方法  

Metric of image quality based on structural similarity

作　　者：朱里[1] 李乔亮[1] 张婷[1] 汪国有[1] 

机构地区：[1]华中科技大学图像识别与人工智能研究所生物与医学图像研究室,湖北武汉430074

出　　处：《光电工程》2007年第11期108-113,共6页Opto-Electronic Engineering

基　　金：国家自然科技基金项目(60372066)

摘　　要：分析现有图像质量评价方法特点,针对基于失真敏感度的质量评价方法的局限性,提出了一种新的基于结构相似度的图像质量评价方法,从亮度、对比度、结构三个子方面得到一个总的相似性度量作为质量客观评价标准。该方法充分考虑了图像的结构信息和人类视觉的特性,从图像内容的理解功能出发,通过数学建模估算出人眼对图像质量的主观视觉感受,使结构相似性计算模型符合图像处理应用的本质。通过理论推导和算法验证,该方法可以为选择图像压缩算法和评价图像质量提供依据。将采用压缩算法SPIHT编解码后的图像与传统的峰值信噪比方法评价图像相比,实验表明,本文提出的算法是一种更为有效的图像质量评价方法。A new evaluation method of image quality based on its construction simulation is proposed to solve the limit of evaluation of perceptual distortion by analyzing the current characteristic of image quality evaluation method. Whole similarity obtained from luminance, contrast and image construction is the objective evaluation standard of image quality. The method fully considers the characteristic of structure information of image and vision of people, starts from the comprehension function of image context, and sets up the structure simulation computing model to evaluate the subjective perception to image quality. By theory deducing and algorithm validation, the evidences for selecting the image compressed algorithm and evaluating image quality are obtained. Reconstructed image after encoding by compression algorithm SPIHT （Set Partitioning in Hierarchical Trees） is compared with the traditional evaluation image based on Peak Signal-to-noise Ratio （PSNR）, and experiment shows that the method proposed in the paper is a more effective evaluating method for image quality.

关 键 词：图像质量评价 结构相似性 失真敏感度 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]