Lossless Compression Of Medical Images Based On Neural Networks

With the quick development of the medical technology, more and more medical image devices are applied in clinical medicine for diagnosing the illness. But more and more medical pictures are waiting to store or communicate through the net. To spare the storing space and not losing the important information in pictures, we need to compress them with lossless method; To transmit the pictures in time, the lossless compression cannot be too complex.Based on this, we study the medical image lossless compression.Firstly, we construct the medical image compression model after analyzing the JPEG-LS compression model, then we improve it. We establish three-point predicting 442 model based on theory and experiments, and realize the adaptive adjustment for the prediction. The performance is excellent in compressing ratio and compressing time through tests.Secondly, we select effective Rice encoding in all the entropy coding methods and realize adaptive adjustment. This optimizes the medical image compression algorithm.Lastly, for the true color medical image, the paper analyses the linear transform of color space, and gives the method of extracting color components line-by-line and compressing them sequently. In process of medical image predicting coding, this method enhance the context correlations of same color components, improves the compressing ratios, not increasing the complexity of the predicting coding algorithm. Through experiment tests this algorithm performs perfectly in compressing ratio and compressing time.All the algorithms are tested through VC programs.