Color Difference Evaluation for Digital Pictorial Images Using the Magnitude Estimation Method

Abstract

A new set of color difference data using complex images was collected including small and large color differences. Using 13 images, 182 test images were generated to have various average lightness, chroma or hue values compared to the original images. Twenty observers estimated the overall color difference between the reference and manipulated images shown on an sRGB monitor with 250 cd/m2 as a peak white in a dark room. The magnitude estimation technique was adopted for the visual assessment. The parametric factors of CMC(l:c), and CIEDE2000(l:c) were optimized using the color difference data set. Results showed that CIEDE2000 performs better than CMC. The best ratio for CMC(l:c) is (4:3), while (3.5:1) is the best for CIEDE2000(l:c) for overall color difference prediction.