Quality Analysis of Color Images Compressed with Enhanced Vector Quantizer Designed Using HSI Color Space

This paper presents a novel technique of improving the quality of decompressed color images. For compression of images vector quantization (VQ) technique is used in which an enhanced vector quantizer (codebook) is designed using Kohonen ´s self organizing feature maps(SOFM) neural networks. While designing the vector quantizer, the training of the self organized network is done by developing an enhanced training image and also by selective training of input vectors. The quality of reconstructed image is evaluated by using image quality measures such as, Structural content, Image fidelity, mean structural similarity index along with conventionally used PSNR and entropy of the images on RGB color space. The vector quantizer is design for HSI color space also and performance is observe as well as compared with RGB color space. The comparison with existing JPEG with respect to storage and quality is noted. This paper aims towards development of dedicated hardware for compression and decompression using VLSI techniques for which Vector quantization is most simple and flexible approach.