M. Emin Yuksel, Erciyes University, Turkey Title: Neuro-Fuzzy Design for Image Processing

Abstract: Digital images are often corrupted by noise during image acquisition and/or transmission due to a number of imperfections caused by image sensors and/or communication channels. In most image processing applications, it is of vital importance to remove the noise from the image data because the performances of subsequent image processing tasks (such as segmentation, feature extraction, object recognition, etc.) are severely degraded by the noise.
  A good noise filter is required to satisfy two conflicting criteria of (1) suppressing the noise while at the same time (2) preserving the useful information (edges, thin lines, texture, small details, etc.) in the image. Unfortunately, a great majority of currently available image filters cannot simultaneously satisfy both of these criteria. They either suppress the noise at the cost of distorting the useful information in the image, or preserve image information at the cost of reduced noise suppression performance.
  In the last few years, there has been a growing research interest in the applications of computational intelligence techniques, such as neural networks and fuzzy systems, to the problems in digital image processing. Indeed, neuro-fuzzy (NF) systems offer the ability of neural networks to learn from examples and the capability of fuzzy systems to model the uncertainty, which is inevitably encountered in noisy digital images. Therefore, neuro-fuzzy systems may be utilized to design line, edge, and detail preserving filtering operators for processing noisy digital images.
  In this tutorial, we will begin by a quick review of the fundamental concepts of fuzzy and neurofuzzy systems as well as their application to digital image data. Then, we will derive a generalized neuro-fuzzy (NF) based operator suitable for a range of different applications in image processing. Specifically, we will consider three different applications of the presented NF operator: (1) noise filter, (2) noise detector and (3) edge extractor.
  In the noise filter application, the NF operator will be employed as a detail-preserving noise filtering operator to restore digital images corrupted by impulse noise without degrading fine details and texture in the image. In the noise detector application, the NF operator will be employed as an intelligent decision maker and utilized to detect impulses in images corrupted by impulse noise. Hence, the NF operator will be used to guide a noise filter so that the filter will restore only the pixels that are detected by the NF operator as impulses, and leave the other pixels (i.e. the uncorrupted pixels) unchanged. Consequently, the NF operator will help reduce the undesirable distortion effects of the noise filter. In the edge extractor application, the NF operator will be used to extract edges from digital images corrupted by impulse noise without needing a pre-filtering of the image by an impulse noise filter.
  In all of these applications, the same NF operator will be used for three different purposes. The fundamental building block of the NF operator to be presented is a simple 3-input 1-output NF subsystem. We will then show that highly efficient noise filtering, noise detection or edge extraction operators may easily be constructed by combining a desired number of simple NF subsystems within a suitable network structure. Following this, we will present a simple approach for training the NF operator for its particular target application. Specifically, we will show that the internal parameters of the NF subsystems in the structure of the presented NF operator may adaptively be optimized by training, and the same NF operator may be trained as a noise filter, noise detector or an edge extractor depending only on the choice of the training images. We will further show that the NF subsystems may be trained by using simple artificial training images that can easily be generated in a computer.
  For each of the three applications of the presented NF operator, we will demonstrate the efficiency of the presented approach by appropriately designed simulation experiments and also compare their performance with a number of selected operators from the literature. We will complete the tutorial with a brief summary of other existing as well as potential applications of the presented general-purpose NF operator in image processing.

SELECTED PAPERS RELATED WITH THE TUTORIAL/TALK
1. M. T. Yildirim, A. Basturk, M. E. Yuksel, “Impulse noise removal from digital images by a detail-preserving filter based on type-2 fuzzy logic”, IEEE Transactions on Fuzzy Systems, vol. 16, no. 4, pp. 920-928, 2008.
2. M. E. Yuksel, "Edge detection in noisy images by neuro-fuzzy processing", International Journal of Electronics and Communications, vol. 61, no. 2, pp. 82-89, 2007.
3. M. E. Yuksel, “A hybrid neuro-fuzzy filter for edge preserving restoration of images corrupted by impulse noise,” IEEE Transactions on Image Processing, vol. 15, no. 4, pp. 928-936, 2006.
4. M. E. Yuksel, “A simple neuro-fuzzy method for improving the performances of impulse noise filters for digital images,” International Journal of Electronics and Communications, vol. 59, no. 8, pp. 463-472, 2005.
5. M. E. Yuksel and E. Besdok, “A simple neuro-fuzzy impulse detector for efficient blur reduction of impulse noise removal operators for digital images,” IEEE Transactions on Fuzzy Systems, vol. 12, no. 6, pp. 854-865, 2004.
6. M. E. Yuksel, A. Basturk and E. Besdok, “Detail-preserving restoration of impulse noise corrupted images by a switching median filter guided by a simple neuro-fuzzy network,” EURASIP Journal on Applied Signal Processing, vol. 2004, no. 16, pp. 2451-2461, 2004.

INVITED BOOK CHAPTER
M. E. Yuksel, “Application of neuro-fuzzy methods for noise filtering, noise detection and edge extraction in digital images corrupted by impulse noise”, Soft Computing in Image Processing: Recent Advances, Edited by M. Nachtegael, D. Van der Weken, E. E. Kerre and W. Philips, Springer, 2006. (ISBN: 978-3-540-38232-4).

Bio Sketch: M. Emin YUKSEL received his B.S. degree in electronics and communications engineering from Istanbul Technical University, Istanbul, Turkey, in July-1990. In February-1991, he joined the Dept. of Electrical and Electronics Eng., Erciyes University, Kayseri, Turkey. He received his M.S. and Ph.D. degrees in electronics engineering from Erciyes University in February-1993 and September- 1996, respectively. Between March-1995 and December-1995, he was with Signal Processing Section, Dept. of Electrical Engineering, Imperial College, London, UK. Currently, he is an associate professor at the Dept. of Electrical and Electronics Engineering, Erciyes University, Kayseri, Turkey. His general research interests include computational intelligence techniques, evolutionary computation and applications of these techniques in signal and image processing.
  Dr. Yuksel was the conference chair of the IEEE SIU-2005 (IEEE 13th Signal Processing and Communication Applications Conference) and conference local chair of the HDM-2008 (International Conference on Multivariate Statistical Modeling and High Dimensional Data Mining). He is a member of the Editorial Board of the International Journal of Reasoning-Based Intelligent Systems. He has served as a member of the technical committees of many national and international conferences. Dr. Yuksel is a Senior Member of the IEEE.