一种基于DCT域加权处理的图像质量评价方法

An image quality measure based on DCT domain weighted processing

依据人眼视觉系统的多通道模型,评价图像亮度细节的质量,目前已经成为一个活跃的研究方向.由于这些模型采用滤波器组模拟对比度敏感函数(CSF),增加了视觉阈值获取和实时评价算法的复杂度.本文依据离散余弦变换对应子带的视觉权重,利用“之”形扫描方式实现对图像频带的任意划分,并由此定义加权峰值信噪比作为图像评价的测度,从而达到对图像质量快速准确评价的目的.

According to the multi-channel modals of the human visual system, researchers are paying more attention to the assessment of image luminance details. As filter banks are used to simulate the contrast sensitivity function （CSF） in these models, the complexity of visual threshold obtainment and real-time assessment algorithms are increased. In this paper, image subbands are divided at will by zig-zag array based on visual weighted subbands obtained by discrete cosine transform. Consequently, the weighted peak signal-to- noise ratio （WPSNR） is defined as the image quality measure. The results show that the proposed measure scheme correlates better with perceptual image quality.