Neural Networks and Wavelet Transform Based Hybrid Architectures for Image Compression- An Evaluation

Work had always been under process to design efficient algorithms for image compression based on various conventional and soft computing methodologies. This paper aims at exploring the application of multi layered perceptron (MLP) feed forward neural networks (FFNN), wavelet transforms and their combination architectures for image compression. Initially two neural network architectures for image compression are explored – first one using a feed forward neural network of 16-4-16 dimension and the other hybrid architecture using same 16-4-16 FFNN combined with conventional quantizer and Huffman encoder blocks. Next, hybrid architecture for image compression is explored using wavelet transforms together with quantizer and Huffman encoder blocks. Finally, hybrid architecture for image compression combining wavelets and neural networks combined with other traditional compression techniques is tested. In the process, different hand designed wavelets such as the Haar, Daubechies, Coiflets, Symlets and Biorthogonal wavelets etc are taken into consideration. Wide range of bench mark images of varying details are considered for tabulation and graphical analysis of the performance metrics Peak Signal to Noise Ratio (PSNR), Mean Square Error (MSE) and Compression Ratio (CR) obtained with proposed architectures.