WAVELET BASED MEDICAL IMAGE COMPRESSION THROUGH PREDICTION

Abstract

As image compression is desirable to minimize storage space and for reduction of transmission cost over the network. There are two types of image compression, lossy and lossless Medical Images are of special type and purpose which require lossless compression as a minor loss can cause very serious consequences. This thesis work is on medical images compression using Wavelet Transform and Prediction technique. Computed Tomography (CT) and Magnetic Resonance (MRI) Images are used for analysis and experiment. Advanced form of wavelet transform, Lifting Wavelet Transform is used for image decomposition. Correlation that is present between the neighboring, parents and parent neighboring pixels is used for analysis of decomposed image. One prediction equation for each sub-band is developed using linear prediction technique. Using the prediction equation of each sub-band, coefficients of the sub-band are predicted, compared and matched with original coefficients through plotting a graph of coefficients. It has been observed that a equation with all variables can causes multicollinearity problem. Different combinations of variables have been analyzed to over come the multicollinearity and to achieve the accurate prediction. At this stage graph of original coefficients is matched exactly with predicted coefficients graph. After modeling the prediction equation and selecting the variables for each equation, different fine sub-bands are predicted using the coarsest sub-bands while the coarsest sub-bands are processed by Discrete Pulse Code Modulation Transform (DPCM). Arithmetic coding of the combined data vector has been performed to achieve the highest compression. Reverse Wavelet transform is applied to obtain the original image at receiving end. The results has been compared with recent methods for medical image compression. The proposed method gives the best results in terms of compression and coding/decoding time. It is a simple method which employs useful methodology for variable selection. MATLAB 7.0 has been used for the implementation of proposed approach. Experiments have been conducted on a variety of standard grayscale images.