Image compression based on Wavelet Support Vector Machine Kernels

摘要

In this recent multimedia world, the major challenges are the optimized use of storage space and Bandwidth. Compression plays the crucial role to reduce the storage space of images and transmission of information with limited Bandwidth availability without degrading the quality of image. Inorder to fulfil the above demand, in literature various compression algorithms were proposed for different applications. In this paper, we evaluated the performance of Wavelet Support vector machines (WSVM) with different combinations of kernel function and wavelets.SVM regression is applied to wavelet coefficients inorder to approximate the obtained coefficients from wavelets so that better compression can be achieved by removing the additional redundancy. From training data and realized compression, Support Vector Machine Regression (SVMR) has the possibility to learn about the dependency by the use of support vectors inorder to represent the real data and to eradicate redundancy. .Run-length coding is used to encode the support vectors and its corresponding weights, obtained from the SVM regression. Performance evaluation of WSVM is done in terms of compression Ratio, MSE and PSNR Experimental results shows that the compression performance can be improved with rbio4.4 combined with RBF kernel gives high compression ratio without loss in the image quality