Abstract: Based on the Human Visual System(HVS) with different sensitivities in the frequency domain, this paper presents an image quality assessment model. Discrete Cosine Transform(DCT) coefficients are divided into low and high frequency components. Distortion of low frequency is showed by absolute square sum of low frequency error, but distortion of high frequency is showed by relative square sum of high frequency error. The product of two distortion is adopted as objective evaluation of quality. The model is validated with Differential Mean Opinion Score(DMOS) on LIVE database. Experimental results show that the performance of this method is superior to the algorithms of Peak Signal to Noise Ratio(PSNR) and Structural SIMilarity(SSIM).

Key words: image quality assessment, Discrete Cosine Transform(DCT), Sense Capacity(SC), linearity correlation coefficient

摘要： 根据人眼对图像高低频失真的敏感度不同，提出一种图像质量评价模型。将图像离散余弦变换后的频率分量分为低频和高频分量，利用低频系数变化量的绝对平方和表示低频失真量、高频系数变化量相对平方和表示高频失真量，通过2个失真量相乘得到客观评价值。实验结果表明，该模型与LIVE图库上差异主观评价分的线性相关性优于峰值信噪比和结构相似度模型。

关键词: 图像质量评价, 离散余弦变换, 感觉容量, 线性相关系数

CLC Number: 

• TP391