基于人眼视觉的无参考彩色图像质量评价方法

No-reference Color Image Quality Assessment Method Based on HVS

目的研究如何自动评价相机拍摄图像的质量问题,针对其可能存在的失真类型特点,提出基于人眼视觉的无参考彩色自然图像质量评价方法。方法利用颜色恒常性和人眼视觉掩盖效应,分别计算色偏和模糊程度,并对相机拍摄图像进行测试;利用佳能EOS700D相机拍摄图像,对可能出现的失真进行组合测试,并与主流评价模糊图像的算法如G-SSIM,GSIM及色偏评价方法进行比较。结果实验发现,综合评价指标计算值与图像质量人眼评价值成反比;图像质量越差,综合评价指标计算值越高。结论无参考彩色自然图像质量评价方法,适合相机拍摄图像的无参考图像质量评价,可以开发集成到相机的辅助功能中。

To study the method for automatic evaluation of image quality taken by cameras, and to propose no-reference color image quality assessment method based on HVS targeting at the features of the possible distortion types.Color constancy and human visual masking effect were used to calculate the degree of color shift and blur, respectively, and to test the images taken by cameras. With the images taken by the cannon EOS700 D camera, the possible distortions were tested in combinations, and comparison was made with the mainstream algorithms for assessing blurred images such as G-SSIM, GSIM and color cast evaluation method. The results showed that the calculated value of the comprehensive assessment index was inversely proportional to the assessment value of the image quality given by the observers. As the index of the image quality assessment grew, the quality of the image was declining. The no-reference color natural image quality assessment method was suitable for no-reference color image quality assessment of photos taken by cameras, and can be developed and integrated to the auxiliary function of camera.