Aiming at the color features of human skin and mucosa, a color constancy model is established to collect and analyze medical true color images under natural light condition. In LAB color space, a TRM (topology resolve-map) model is established to restore these colors from the gathered images to standard values by linear mapping to the one-dimensional L* space and triangulation mapping to the two-dimensional a* b* space. Via regulating and restoring the 13 skin colors in SG color codes, the experimental results show that the TRM model has good accuracy and stability and the differences of delta L, delta ab and delta E are dramatically decreased.