The Development of A Color Visual Difference Model (CVDM)

A color visual difference model (CVDM) was developed to predict the imag e quality difference of two images. The model is an extensio n of the two previously published models: The monochromatic Visible Differenc e Predictor (VDP) by Scott Daly an d 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 me chanisms, and visible color difference calculation. Inputs to th e model ar e a reference image and a processed image, as well calibration paramet ers such as viewing distance , resolution of the images, and white point. The output of the model is CIELAB ∆E map, on which the bright co l rs represent la rge ∆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. Th e results show a better agre ement with visual impression than does the S-CIELAB model.