Image Compression Using Multilayer Feed Forward Artificial Neural Network with Nguyen Widrow Weight Initialization Method

In this paper, Multilayer Feed Forward Artificial Neural Network with weight initialization method is Proposed for Image Compression. Image compression helps to reduce the storage space and transmission cost. Artificial Neural network (ANNs) is a training algorithm has used to compress the image. Artificial neural network is exceptionally Feed Forward Back propagation neural network (FFBPNN) in which Neural network has trained by Back propagation neural network algorithm for image compression and decompression. Many techniques and algorithms are used to train artificial neural network by considering the different number of hidden neurons, epoch and reconstructed image compared with original image. Nguyen and widrow a weight initialization method is used in MLFFANN to compressed the digital image JPEG, PNG and BMP with 16 input neurons,10 in hidden layer that are determining compression rate and 16 output neurons. Training algorithm and developed architecture provide better result. The performance parameter of image compression is evaluated using some standard image. The result of simulation is shown and compared different quality parameter of applying on various parameters. Keywords— Image compression, MLFFANN, PSNR, RMSE, MSSIM and execution time