Abstract: We present an algorithm of color image understanding whichsegments the image after analyzing surfaces with color variations due tolighting condition and object colors. The work is based on a theory——theDichromatic Reflection Model——which describes the color of the reflectedlight as a linear combination of light from surface reflection (highlight)and body reflection (object color).A peculiar cluster shape (skewed T) isformed by color pixels of an object area in 3D histogram in color space.The analysis of cluster's property helps to determine the color of lightand objects,but at same time the meaningful cluster is generated by theprojection of certain object area of image. According to the strategy ofhypothesis and test,following the continuity of the image and the featureof color clusters,the algorithm completes the image segmentation and givesforth the physical description of imaging process,including intrinsicimages,segmented image,the color of light and objects. The intrinsicimages are composed of both matte image and highlight image reflectingrespectively the properties of light condition and every objects.

Key words: Dichromatic reflection model, Image dissection, Color analysis, Algorithnl