Performance Analysis of Daubechies Wavelet and Differential Pulse Code Modulation Based Multiple Neural Networks Approach for Accurate Compression of Images

Large Images in general contain huge quantity of data demanding the invention of highly efficient hybrid methods of image compression systems involving various hybrid techniques. We proposed and implemented a Daubechies wavelet transform and Differential Pulse Code Modulation (DPCM) based multiple neural network hybrid model for image encoding and decoding operations combining the advantages of wavelets, neural networks and DPCM because, wavelet transforms are set of mathematical functions that established their viability in the areas of image compression owing to the computational simplicity involved in their implementation, Artificial neural networks can generalize inputs even on untrained data owing to their massive parallel architectures and Differential Pulse Code Modulation reduces redundancy based on the predicted sample values. Initially the input image is subjected to two level decomposition using Daubechies family wavelet filters generating high-scale low frequency approximation coefficients A2 and high frequency detail coefficients H2, V2, D2, H1, V1 and, D1 of multiple resolutions resembling different frequency bands. Scalar quantization and Huffman encoding schemes are used for compressing different sub bands based on their statistical properties i.e the low frequency ban