Hierarchical Compression Method for Palettized Images

A.Yu. Bavrina, N.I. Glumov, V.V. Sergeyev, and E.I. Timbay (Russia)


Lossless compression, palettized images, hierarchicalrepresentation.


As usual, the images of the class under consideration contain regions, delineated by contour and filled with one color (or with some texture with small number of colors), with overlaid symbols, lines and text. The number of colors in such images is restricted to a few tens. Smoothly intensity changes, which are typical for grayscale and color (RGB) images, are uncommon for this class of images. A method for lossless compression of palettized synthetic images is described in this paper. The method bases on hierarchical image representation in the form of a set of matrixes or hierarchical levels (HL) of reduced size and coding only the information, needed for reconstruction of the current HL, using the reconstructed values of the previous HL. Different variants of method realization are examined. Experimental investigations show the effectiveness of the proposed method in comparison with well-known lossless compression standards. Widely employed image compression methods are lossy and base on the fact that small error in pixel value results in small visual image distortion. For palettized image compression lossless methods are necessary since indices (usually, in byte per pixel format) instead of intensity values are used. Indices determine color in RGB coordinates with the help of separate stored table (a palette). The error of pixel value even equal to 1 can results in color change, for example, from blue to red.

Important Links:

Go Back