Cascade-Forward vs. Function Fitting Neural Network for Improving Image Quality and Learning Time in Image Compression System

Abstract— The backpropagation neural network algorithm (BP) was used largely in image and signal processing. The BP requires long time to train the BPNN with small error. Therefore, in this research, three Artificial Neural Networks models (ANNs) were constructed. Three algorithms: FeedForwardNet, CascadeForwardNet and FitNet were adopted to train the three constructed ANNs models separately. Each one of constructed models consists of input layer to input the original image, hidden layer to produce the compressed image and finally output layer for decompressed image. The training and testing performance of the constructed models with different architecture were compared to identify the model with best compression ratio (CR) and Peak to Signal to Noise Ratio (PSNR). From experiments, we noted that the better results are obtained when we used the FitNet ANN model. According to results, the performance of constructed FitNet ANN for image compression can be increased by changing the number of hidden layer neurons.

[1]  Fatima B. Ibrahim.  Image Compression using Multilayer Feed Forward Artificial Neural Network and DCT , 2010 .

[2]  J. Todd.  Book Review: Digital image processing (second edition). By R. C. Gonzalez and P. Wintz, Addison-Wesley, 1987. 503 pp. Price: £29.95. (ISBN 0-201-11026-1) , 1988 .

[3]  Hadi Veisi,et al.  A Complexity-Based Approach in Image Compression using Neural Networks , 2009 .

[4]  Omaima N. A. AL-Allaf,et al.  Improving the Performance of Backpropagation Neural Network Algorithm for Image Compression/Decompression System , 2010 .

[5]  Anju Vyas.  Print , 2003 .

[6]  Tai-Hoon Cho,et al.  Fast backpropagation learning using steep activation functions and automatic weight reinitialization , 1991, Conference Proceedings 1991 IEEE International Conference on Systems, Man, and Cybernetics.

[7]  Rafid Ahmed Khalil.  Digital Image Compression Enhancement Using Bipolar Backpropagation Neural Networks , 2007 .

[8]  Martin T. Hagan,et al.  Backpropagation Algorithms for a Broad Class of Dynamic Networks , 2007, IEEE Transactions on Neural Networks.

[9]  Senjuti Basu Roy,et al.  Edge Preserving Image Compression Technique using Adaptive Feed Forward Neural Network , 2005, EuroIMSA.

[10]  R. Lippmann,et al.  An introduction to computing with neural nets , 1987, IEEE ASSP Magazine.

[11]  Richard W. Conners,et al.  Fast Back-Propagation Learning Using Steep Activation Functions and Automatic Weight , 1992 .

[12]  S. Anna Durai,et al.  Image Compression with Back-Propagation Neural Network using Cumulative Distribution Function , 2008 .

[13]  Philip D. Wasserman,et al.  Neural computing - theory and practice , 1989 .

[14]  Omaima N. Ahmad AL-Allaf,et al.  Fast Backpropagation Neural Network algorithm for reducing convergence time of BPNN image compression , 2011, ICIMU 2011 : Proceedings of the 5th international Conference on Information Technology & Multimedia.

[15]  Liu Yang,et al.  An Image Compressing Algorithm Based on Classified Blocks with BP Neural Networks , 2008, 2008 International Conference on Computer Science and Software Engineering.

[16]  N. K. Ibrahim,et al.  Artificial neural network approach in radar target classification , 2009 .