The Development of A Color Visual Difference Model (CVDM)

A color visual difference model (CVDM) was developed to predict the image quality difference of two images. The model is an extension of the two previously published models: The monochromatic Visible Difference Predictor (VDP) by Scott Daly and the color Spatial-CIELAB (SCIELAB) model by Zhang et al. The CVDM consists of color space conversion, modulation by contrast sensitivity functions, visual masking effect, multiresolution detection mechanisms, and visible color difference calculation. Inputs to the model are a reference image and a processed image, as well as calibration parameters such as viewing distance, resolution of the images, and white point. The output of the model is a CIELAB ∆E map, on which the bright colors represent large ∆E values, and the dark colors represent small visible ∆E values. The model was applied to detect the visibility of blur, noise, grating, and compression artifacts. The results show a better agreement with visual impression than does the S-CIELAB model.