The Applicability of Neural Networks to Non-linear Image Processing

Abstract: In this paper, the applicability of neural networks to non-linear image processing problems is studied. As an example, the Kuwahara filtering for edge-preserving smoothing was chosen. This filter is interesting due to its non-linear nature and natural modularity. A number of modular networks were constructed and trained, incorporating prior knowledge in various degrees and their performance was compared to standard feed-forward neural networks (MLPs). Based on results obtained in these experiments, it is shown that several key factors influence neural network behaviour in this kind of task. First, it is demonstrated that the mean squared error criterion used in neural network training is not representative for the problem. To be able to discern performance differences better, a new error measure for edge-preserving smoothing operations is proposed. Secondly, using this measure, it is shown that modular networks perform better than standard feed-forward networks. The latter type often ends up in linear approximations to the filter. Finally, inspection of the modular networks shows that, although analysis is difficult due to their non-linearity, one can draw some conclusions regarding the effect of design and training choices. The main conclusion is that neural networks can be applied to non-linear image processing problems, provided that careful attention is paid to network architecture, training set sampling and parameter choice. Only if prior knowledge is used in constructing the networks and sampling the datasets can one expect to obtain a well performing neural network filter.

[1]  C. D. Green,et al.  Are Connectionist Models Theories of Cognition , 1998 .

[2]  Fa-Long Luo,et al.  Applied neural networks for signal processing , 1997 .

[3]  Klamer Schutte,et al.  Vehicle recognition in infrared images using shared weight neural networks , 1998 .

[4]  Dick de Ridder,et al.  Shared Weights Neural Networks in Image Analysis , 1996 .

[5]  R. M. Hodgson,et al.  The properties and training of a neural network based universal window filter (UWF) , 1995 .

[6]  Kurt Hornik,et al.  Multilayer feedforward networks are universal approximators , 1989, Neural Networks.

[7]  Herbert Jahn.  A Neural Network for Image Smoothing and Segmentation , 1998, SSPR/SPR.

[8]  Rajesh P. N. Rao,et al.  Efficient Encoding of Natural Time Varying Images Produces Oriented Space-Time Receptive Fields , 1997 .

[9]  A. M. Taratorin,et al.  Digital Processing of Biomedical Dynamical Images , 1988 .

[10]  Jitendra Malik,et al.  Anisotropic Diffusion , 1994, Geometry-Driven Diffusion in Computer Vision.

[11]  Jenlong Moh,et al.  Implementing morphological operations using programmable neural networks , 1992, Pattern Recognit..

[12]  Ruzena Bajcsy,et al.  Adaptive Image Segmentation , 1988 .

[13]  Bart M. ter Haar Romeny,et al.  Geometry-Driven Diffusion in Computer Vision , 1994, Computational Imaging and Vision.

[14]  Yeuvo Jphonen,et al.  Self-Organizing Maps , 1995 .

[15]  Lieuwe Jan Spreeuwers,et al.  Image Filtering with Neural Networks: Applications and performance evaluation , 1992 .

[16]  Anders Krogh,et al.  Introduction to the theory of neural computation , 1994, The advanced book program.

[17]  Hiroshi Katsulai,et al.  Evaluation of Image Fidelity by Means of the Fidelogram and Level Mean-Square Error , 1981, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[18]  Jaakko Astola,et al.  A new class of nonlinear filters-neural filters , 1993, IEEE Trans. Signal Process..

[19]  Kunihiko Fukushima,et al.  Neocognitron: A new algorithm for pattern recognition tolerant of deformations and shifts in position , 1982, Pattern Recognit..

[20]  Makoto Nagao,et al.  A Structural Analysis of Complex Aerial Photographs , 1980, Advanced Applications in Pattern Recognition.

[21]  Heekuck Oh,et al.  Neural Networks for Pattern Recognition , 1993, Adv. Comput..

[22]  Elie Bienenstock,et al.  Neural Networks and the Bias/Variance Dilemma , 1992, Neural Computation.

[23]  R. J. Williams,et al.  Simple Statistical Gradient-Following Algorithms for Connectionist Reinforcement Learning , 2004, Machine Learning.

[24]  Robert P. W. Duin,et al.  Feature extraction in shared weights neural networksDick , 1996 .

[25]  Helge J. Ritter,et al.  A neural 3-D object recognition architecture using optimized Gabor filters , 1996, Proceedings of 13th International Conference on Pattern Recognition.

[26]  Lawrence D. Jackel,et al.  Backpropagation Applied to Handwritten Zip Code Recognition , 1989, Neural Computation.

[27]  F. Tomita,et al.  Extraction of Multiple Regions by Smoothing in Selected Neighborhoods , 1977 .

[28]  Vijaykumar Gullapalli,et al.  A stochastic reinforcement learning algorithm for learning real-valued functions , 1990, Neural Networks.

[29]  William K. Pratt,et al.  Digital image processing (2nd ed.) , 1991 .

[30]  Geoffrey E. Hinton,et al.  Learning internal representations by error propagation , 1986 .

[31]  Kishan G. Mehrotra,et al.  Efficient classification for multiclass problems using modular neural networks , 1995, IEEE Trans. Neural Networks.

[32]  Teuvo Kohonen,et al.  Self-Organizing Maps , 2010 .