WT-SOM network implementation on FPGA for the medical images compression

This paper is devoted to the implementation of a new combined method based on wavelet transform and neurons network (WT-SOM) and designed for the compression of medical images on FPGA VirtexII circuit. Medical images present specific characteristics which require to be exploited by an explicit and efficient compression algorithm. Compression is a vital operation for images transmission, since huge volume data is generally presented. The Vector Quantization (VQ) constitutes a crucial stage in Digital images compression. In order to improve the performances of its implementation, the (VQ) allows to create a dictionary on the level "block" by a neuronal approach that of Kohonen (Self Organizing Map: SOM), tools widely used for lossless compression and high dimensional data for their implementation performances on Virtex II FPGA circuit. It is currently a very active field, and the implementation of neurons networks on FPGA circuit with a large number of neurons remains a difficult and costly task.