The LAR method as a new scalable region-based technique for color images compression at low bit rates

This paper presents a very efficient solution for chromatic image coding using a region technique. The originality relies in the fact that the region's description is not transmitted, but deduced from the low resolution luminance representation, and thus does not induce any overhead. The Y component is compressed by the basic LAR (locally adaptive resolution) codec, a content-based approach composed of a spatial coder, to achieve low bit rates, and a spectral coder, to compress the error image. The two coders, as well as the region description, share the same data representation structure, avoiding the contour or block artifacts of traditional techniques.