Robust edge detector using back propagation neural network with multi-thresholding

Edge detection is one of the prominent preprocessing stages in many image processing applications like Image Segmentation, Machine vision, Image Analysis and Feature Extraction etc. In order to get optimally true edge response in these applications, a particular edge detection technique shall be vulnerable to errors even when the input image gets contaminated due to presence of high frequency noise or become hazy due to blurriness. In this paper, a robust edge detection technique based on Back-propagation Neural Network with Multi-Thresholding, applicable on both Gray scale and Colored images, is presented. It is demonstrated that the proposed technique performs qualitatively and quantitatively better than Sobel, Robert's, Prewitt's, Canny and Neural based (without Multi-Thresholding) Edge Detectors under both Noisy & Blurred input conditions.