Comparison of Kohonen feature map against K-mean clustering algorithm with application to reversible image compression

A proposed criteria based on the concept of normalized entropy is used to evaluate the performances of the Kohonen feature map and the K-mean clustering algorithm and the experimental results are discussed. Then a newly proposed efficient reversible image compression method is described briefly. It turned out that a problem which severely handicaps the practical realization of the method can be resolved by the Kohonen´s algorithm at a satisfactory level