Study on Gamut Mapping Algorithm Based on GRNN

Abstract

The paper proposed a new gamut mapping algorithm based on General Regression Neural Network (GRNN) for the problem that the color of pictures can’t be reproduced perfectly in the cross media transmission. The two groups color samples were created by Matlab. One color group is Adobe RGB as the original gamut, and the other is sRGB as the aim gamut. Then they were converted into a Profile Connection Space (PCS), such as CIE LAB or IPT. The GRNN gamut mapping module, the mapping relationship of the original gamut and reproduction gamut, were calculated with color samples. The color differences of mapping and original pictures are small in general and the saturation of mapping picture is higher than that of original picture. The different PCS has effect on the mapping result of specific type picture. The gamut mapping result of GRNNGM (General Regression Neural Network Gamut Mapping) algorithm is better than the HPMINDE (Hue Perceived Minimum Color Difference Error) algorithm. Lab space is more fit with the face picture mapping and there are no significant differences to the other types of pictures.