Wavelet transform for still color image compression

We consider here the compression of still color image with very low distortion from the human eye point of view. The basic idea in this work is to take into account the variations of human eye/brain spatial resolution with color. The mosi natural way for an image processing researcher to perform such a scheme is to use a multiresolution analysis of the image to be coded before quantization and coding. Previous experiences connected with still grey value image compression/decompression scheme design have shown that the wavelet transform, Mallats algorithm is a very efficient method for this purpose, particularly if real time implementation is under consideration. Hence we present in this paper a wavelet transform algorithm for color image and we show how and with what performances tile transformed image can be altered and reduced. We show that a quasi lossless compression/decompression scheme can be easily obtained with compression ratio up to 1:10 (quantization step was not considered here). The results obtained after a large series of tests based on psychovisual estimations rather than on pure PSNR evaluation are in good accordance with the assumed properties of the human visual perceptive system.