Edge Detection Based On Mathematical Morphology

Edge detection is the foundation of object survey, object segment, image compression, pattern recognition, and so on. Also edge detection is a hot topic in digital image processing. There are many classical edge detection operators , such as Sobel operator, Prewitt operator, Robert operator, Laplance operator, LoG algorithm, Canny algorithm and so on. But all the existing edge detection methods have limitation and applicable scope. We analyze the good points and the bad points of the existing edge detection methods, and propose new edge detection method based on mathematics morphology.First we summarize the elementary knowledge and the nature of binary image morphology and gray-scale image morphology, and illustrate the significance of the morphology operator in digital image processing. Then we review the variant methods of edge detection, and analyze the merits and the existence insufficiency of those methods, and summarize the recent development of edge detection. Finally we propose three methods of extracting edge characteristic and a concept —'highlight edge'. (1) A method of extracting interest point: Based on the morphology top-hat transformation and bottom-hat transformation, we propose a method of extraction interest point. The method is easy and anti-noise. And we can distinguish the convex characteristic and the concave characteristic of a digital image ;(2) A method of extracting irregular straight line segment: We combined the Hough transformation with mathematics morphology to distinguish the partial curving straight line and the dashed line. (3) Multi-scale auto-adapted entropy weight edge extraction method: Information entropy exposes the degree of an image rich in information. We determine the different weight value based on image information entropy. We propose a concept— 'highlight edge', and apply the 'highlight edge' to the extraction of billboard in the video soccer competition.