Previous [1] [2] 3 [4] [5] [6] [7 ] [8 ]

Journal of Inforamtion Science and Engineering, Vol.8 No.2, pp.207-222 (June 1992)
Neural Networks for Step Edge Detection Using
Simulated Annealing Technique

Hong-Yuan Liao and Mitechell Middler*
Institute of Insormation Science
Academia Sinica
Taipei, Taiwan
*Department of Engineering and Computer Science
Northwestern University
Evanston, IL 60208

In this paper, we modify the networks proposed in [1]1 to enhance their capability of edge detection. The network in [1] consists of two separate modules:the propective edge selection module and the final edge selection module. The prospective edge selection module consists fo two or more networks, working in parallel, that locate "prospective" edges in the image. A separate network is required to detect the edges in each desired orientation. The "prospective" edge map resulting from the preliminary networks is then fed into the findal network to remove spurious edges and to select the desired edges. The proposed modification allows the networks to ignore small changes of pixel values in a region and, thus, to avoid over-detection. In addition, only the edge pixels detected by the preliminary network have a chance of retaining their status (as edge pizels) in the final network. The inclusion of this constraint not only avoids the occurrence of inconsistency between the preliminary and final networks, but also speeds up the computation significantly. Experiments have been conducted on real images to corroborate the propsoed method.

Keywords: neural network, simulated annealing, edge detection

Received August 31, 1991; revised January 11, 1992.
Communicated by Jun-Shon Huang.
1 While Cortes and hertz [1] refer to their result as a segmentation, when applied to real images, many spurious regions occur and objects are not clearly defined, as is required of a segmentation algorithm. It is more accurate to refer to the refer to the results as an edge detection.


  1. Cortes, C. and Hertz, J.A., "A Network Sysstem for Image Segmentation," in International Joint Conf. On Neural Networks, Washington D.C., June 18-22, 1989, pp. 121-125.