Pixel analysis is the primary step for region

shape and texture analysis and even for semantic analysis. The correctness of the color and luminance analysis of certain pixel in an image or a series of video streams is a guarantee to the acceptable result of other image processing. Based on a reasonable mapping operation to vectors in RGB color space

this paper applies the ART2 to the layered detecting approach to categorize color pixels. The processing steps and final results not only demonstrate the functions of “neoteny learning” and adjusting of vigilance value

but also illustrate that the method is coherent with the human psychological and physiological process of observing an image and also has strong adaptability for shadow noise suppression.