Image processing quality evaluation analysis combining human visual perception with color appearance scale phase congruency
A highly consistent image quality assessment method that integrates human visual mechanisms and multimodal perceptual features is built to enhance its sensitivity to complex distortions such as structural degradation, color shift, and phase perturbation. The limitations of traditional pixel-difference-dependent metrics in representing human perception and exhibiting limited subjective consistency in applications such as intelligent vision, coding optimization, and image enhancement are solved. A structural consistency model of luminance, contrast, and local structural features is built firstly. Then, color appearance scale and phase congruency features are employed to enhance sensitivity to color and edge phase changes. Finally, a regression framework is utilized to achieve multi-feature fusion and parameter optimization. The proposed model demonstrates excellent performance under various distortion conditions: its PLCC is around 0.94 and 0.95 under motion blur and contrast scaling distortion, respectively; and its SRCC exceeds 0.93 under multiple distortion conditions. Simultaneously, all RMSE values remain low, specifically around 6.7, 5.7, 8.1, and 5.0 under Gaussian noise, motion blur, compression, and contrast scaling distortion, respectively. These results indicate that the proposed method more closely approximates subjective perception, providing reliable support for visual enhancement, intelligent compression, and perceptual coding.