Encoding of color still pictures by wavelet transform and vector quantization

This paper presents research results on color-image compression by the wavelet transform and 3-level quad-tree coding. First a still picture based on an RGB color model is converted to YIQ and YUV color components. At the encoding process each color component is encoded individually. A color component is decomposed to (3/spl times/3+1) subband images using the 2-D discrete wavelet transform three times. Next, the subband image L3LL having the lowest frequency component is encoded by DPCM. All other nine subband images are encoded by a scheme based on zero-point quantization. The zero-tree representation of the coefficients belonging to an area of three different resolution levels is then implemented, and the run-length of the zero-tree is MR encoded. Their coefficients are optimized by 5-dimensional vector quantization based on the LBG algorithm is used to encode the remaining coefficients not included in the zero-tree, and only optimized coefficients are thus encoded. In other words, improvement of the image quality for reconstructed images is established because additional nonzero-tree information has been encoded. These results, including the comparison of each filter, show the effectiveness of color-image compression by the wavelet transform and 3-level quad-tree coding.