Abstract: A novel mechanism based on space domain natural scene statistic is proposed for the problems in the state-of-the-art no reference image quality assessment methods. These algorithms often suffer from the poor universality or high computational cost. The logarithmic statistical features are introduced into the proposed algorithm. To make an evaluation to the testing image, the proposed algorithm makes use of measurable deviations from statistical regularities observed in natural images. The simulation results show that the proposed algorithm gets a good consistency with human subjective perception. Compared with the current no-reference methods, the proposed algorithm is simple and effective, and is suitable for various distortion types.

Key words: image quality assessment, no reference, statistical characteristics, logarithm

摘要： 针对现有无参考图像质量评价算法通用性差，且存在高计算复杂度的问题，提出一种基于图像空域统计特征的无参考图像质量评价算法。该算法引入图像对数统计特性，通过度量失真图像与自然图像统计特征之间的偏差，实现无参考图像质量评价。仿真测试结果表明，与现有无参考技术相比，提出算法简单有效，适用于多种失真类型，且与人类主观感知具有较好一致性。

关键词: 图像质量评价, 无参考, 统计特征, 对数