Efficient color image compression using integrated fuzzy neural networks for vector quantization

An efficient design of a vector quantizer for image compression is dependent on optimization criteria used in partitioning similar vector groups formed from the image in the spatial or transform domain. Usually a 3D vector quantization is applied to trivariant color spaces. Current trends in image coding exploit the advantage of subband/wavelet decompositions in reducing the complexity in optimal scalar/vector quantizer (SQ/VQ) design. We have used adaptive vector quantization of wavelet coefficients of subimages in each color plane in RGB color space. Our design of vector quantizers using two neuro-fuzzy clustering algorithms namely AFLC-VQ, and IAFC-VQ generates very low bit rate image encoder in color as well as monochrome and outperforms other similar designs in minimizing distortion arising from quantization. Further tuning of these algorithms, and addition of an entropy coder module after the VQ stage could result in extremely low bit rates (compression ratio around 100:1) at minimal distortion