Analysis of Colorimetric Stationary Criterion: Application to Colour Image Quantization

An important storage capacity is necessary for the growing information need in images processing applications. But, some applications like cartography and electronic surveillance can tolerate little distortion of this information. Thus, quantization algorithms have been proposed to reduce memory space occupied and processing time. Our quantization algorithm is based on the criterion of colorimetric stationary state defined from the covariance matrix of a first order Markovian process. The colour space is divided iteratively into sub-cubes, each containing homogeneous colours obtained by the colorimetric stationary criterion. After division, the process continues by the creation of a colour palette with the number of colour being initially fixed by the operator. The value and originality of the method is demonstrated by applying it to different types of colour images.