Analysis of Image Compression Approaches Using Wavelet Transform and Kohonen's Network

Since digital images require a large space on the storage devices and the network bandwidth, many compression methods have been used to solve this problem. Actually, these methods have, more or less, good results in terms of compression ratio and the quality of the reconstructed images. There are two main types of compression: the lossless compression which is based on the scalar quantization and the lossy compression which rests on the vector quantization. Among the vector quantization algorithms, we can cite the Kohonen's network. To improve the compression result, we add a pre-processing phase. This phase is performed on the image before applying the Kohonen's network of compression. Such a phase is the wavelet transform. Indeed, this paper is meant to study and model an approach to image compression by using the wavelet transform and Kohonen's network. The compression settings for the approach to the model are based on the quality metrics rwPSNR and MSSIM.