The chromaticity of an object surface in scene is a constant feature irrelevant to the object shape, which is called color in-varianoe. But that of the corresponding object in image is variable because of the existance of highlight and shading, no color constancy is observed. To construct a physical model of color reflection is a way to understand the image color correctly. This paper describes a method for computing illuminant and object chromaticity coordinate, which is based on the Dichromatic Reflection Model proposed by klinker et al[1]. first deals with plane fitting in 3D color histogram using K-L transform, then separates the interface reflection component from the body reflection component. Also the limitation of practical CCD cameras is discussed and the experimental results are presented. (Fig.1~3: see color inset No.1~9).