基于感知对比度的图像清晰度客观评价模型

Objective image sharpness metric based on perceptual contrast

图像清晰度是评价图像质量时常用的指标之一。现有的清晰度评价模型未能充分考虑人眼视觉的亮度掩盖特性。为此,在均方根对比度基础上,考虑人眼亮度掩盖特性,通过计算图像中人眼感兴趣区域(包含细节、边缘和纹理)的感知对比度构造一种无参考的图像清晰度客观评价模型。并利用IVC数据库来验证模型,结果表明,与已有的4种清晰(模糊)度评价模型相比,该模型的评价结果更接近人眼主观感受,且计算量小,运算耗时短,是一种简单有效的图像清晰度评价模型。

Image sharpness is one of the common metrics for image quality assessment. Existing sharpness metrics have not given enough consideration to the human luminance masking effect. The root mean squared contrast model is im- proved by considering the human luminance masking effect. A perceptual contrast model is presented. A no-reference sharpness metric is established by averaging the perceptual contrast over the region of interest （contains image details, edges, and texture）. IVC database is used to test the proposed sharpness metric. Experimental results show that, com- pared with four existing sharpness （blur） metrics, the proposed perceptual sharpness metric provides better predictions which more closely matches to the human visual perception, and has lower computational complexity. The sharpness metric is simple but effective.