Codebook Generation using Hierarchical Radial Basis Function Neural Network with Wavelet Transform and Vector Quantization for Better Image Compression

Abstract: Image compression is the technique of reducing the size of the image file without degrading the quality of the image. The compression in file size permits more images to be stored in theavailable amount of memory space. It also decreases the time needed for images to be uploaded over the Internet or downloaded from it. There are many techniques available in the lossyimage compression, in which wavelet transform based image compression is the best technique. Vector Quantization (VQ) is the most powerful toolfor image compression. One of the major steps in the Vector Quantization is the generation of the code book.In this proposed approach, a popular neural network technique called Hierarchical Radial Basis FunctionNeural Network(HRBFNN) approach is used to generate the code book. A combined approach of image compression based on vector quantization and wavelet transform is proposed using HRBFNN. This approach will be very helpful for medical imaging, criminal investigation where high precision reconstructed image is required. The experimental result shows that the proposed technique provides better PSNR value and also reduces the Mean Square Error value than the modified SOM and RBF.

Keywords: Image Compression, Neural Networks, Vector Quantization (VQ), Wavelet Transform, Radial Basis Function (RBF), Hierarchical Radial Basis Function Neural Network (HRBFNN)