Research On Lossless Image Compression

Lossless image compression technology plays an important role in many applications where the images are subjected to further processing, (e. g. , for the purpose of character subtraction, image enhancement), repeated compression/decompression, high cost of the image acquiring, the desired quality of the rendered image is yet unknow. Recent years there have been a tremendous increase in medical imaging and remote sensing image. Lossless image compression is required or highly desired. In this thesis, two efficient lossless compression algorithms are presented.Neural network model characterized selfdaptive, selfearning. This method ,base on perceptron , can modify the prediction weight adaptively , and capture the spatial and spectral correlation effectively. The model studies and works at the same time. The prediction weight of the current pixel is determined by the study result of the encoded pixels, which is consistent with the property that the correlation of the neighbore pixel is similar. We use Rice algorithm to encode the outcome residuals. Consequently, the prediction is more accurate and the encoding speed is high. On the other hand, the prediction pixels can be selected according to the situation. So it can simultaneously exploit the spatial and spectral correlation. The experiment results show that the algorithm outperforms best lossless JPEG based on arithmetic coding evidently. Furthermore, it even performs better than LOCO based on context model and Rice coding for TM image.Integer Wavelet Transform(IWT) is an efficient multiesolution decomposition model. By factoring wavelet transforms in to lifting scheme, we can map integers to integers . We decompose each image into multicales, the entropy of transformed image is reduced. The experiments results have proven the algorithm outperforms best lossless JPEG. In addition, it can implement progressive transform technology and improves coding speed.