Based On Wavelet Transform And Error Competitive Learning Vector Quantization Method

Image compression which is a crucial technology in the fields of communication and multimedia, is very important to information technology. Image is the most important carrier and media which contains information in the information exchanging. The data quantity is very large in a digital image, and should be compressed, which can be satisfied with the high transmission speed and large memory of the digital image.In recent years, with the high-speed development of multimedia and networks technology, many methods of data compression have been into application. There are some international standards of data compression which is the base of widely using of data compression. Research and application on wavelet transform image compression is very active in the last few years. As has been proved, methods of wavelet transform are better than other methods, and have applied to the fields of still and moved image compression, as well as a key part in some international standards such as JPEG2000, MPEG-4. However, it is still a research hotspot that how to deal with the coefficients transformed from a digital image with wavelet in a more efficient way.Vector quantization(VQ) is always more powerful than scalar quantization in theory. But vector quantization's computing quantity is very large, which is one reason that it can not be widely using. Artificial neural networks(ANN) is a distributed system that computes very fast and can deal with the problem of large compute quantity whichVQ brings. So we combine the ANN and VQ in order to advance the VQ and realize the fast VQ which is used to compress the image data. A new vector quantization algorithm for image compression based on wavelet transform and distortion competitive learning (VQWDCL) was proposed.VQWDCL is a method of wavelet image data compression by VQ trained by ANN. Image data obtained by 3 levels wavelet transform are divided into some vectors, then we use ANN to train the vectors so as to realize fast VQ, and here we take use of the modified Distortion Competitive Learning algorithm.The simulation result shows that, VQWDCL is better than JPEG2000 which adopts modified DCL and fast search algorithm under the compression of 70,and the reconstructed image is much clearer than that by JPEG2000 which adopts three levels wavelet transform.To sum up, VQWDCL scheme, a new vector quantization algorithm for image compression based on wavelet transform and distortion competitive learning is a remarkably advanced image compression scheme with good performance.