Spatial Frequency Filtering Using Sofm For Image Compression

Abstract

The aim of the research is to propose a new approach to image coding using SOFM and spatial frequency band-pass filter to investigate the Artificial Neural Network. The approach is based on SOFM which is similar to vector quantization (VQ) and it is adopted the technique to improve the image compression effectively. In the approach has been using the band-pass filter for image compression by SOFM based on vector quantization by components as the original image and the spatial frequency image component, which is derived from the adaptive to the contours of the 2D analysis and synthesis. The calculation of the computational cost is compression based on SOFM. The new approach of image coding using a band-pass filter, where is used as a first stage of proposed method of image encoding and as well as the image decoding has been presented with De-quantization with entropy coding based on arithmetic coder and high pass filter, the evaluation with jpeg format compression shows, that using 16x16 image block of pre-processing in SOFM has given the best compression ratio with small SNR. On the given experiment shows the different pixels presented by Lena.bmp, girl256.bmp and compared with a compression ratio of the Iena.jpeg file.