A new segmented image coding method for color images

For the compression of color images, the lossy compression standard JPEG has two major shortcomings. Firstly, it codes the color components separately. This reduces the possible compression ratio. Secondly, the image quality at very low bit rate is seriously degraded by blocking artifacts. The paper describes an extension of an existing Segmented Image Coding (SIG) method for monochrome images to images containing a limited number of colors. The new technique codes the colors more efficiently and performs better at high compression. After conversion to the YUV color space, the corresponding luminance image is compressed using monochrome SIC. At the decoder, this image is reconstructed and the grey-values are translated into colors. Because different colors in the image can have the same grey-values and because SIC is lossy, some grey-valves may be wrongly reconstructed in the decompressed image. To prevent the introduction of foreign colors in a region, i.e. colors which are not present in the region in the original image, bit vectors are constructed indicating which colors are present. The decoder uses this information to replace foreign colors by colors of the region. The bit vectors are compressed wing a lossless bi-level coding scheme. The paper presents experimental results which show that the new method produces a much better subjective image quality than JPEG at high compression due to the absence of block distortion.