Gray-scale Image Edge Detection

Edge detection is one of the most fundamental aspects in image processing and analyzing,which is still unsolved.Image's edges include image's features such as position and outline,which is the fundamental features of the image. Edge detection is widely used in image analysis and processing such as feature description, image segmentation, image enhancement and pattern recognition etc. so edge detection is the research hot spot in the technology of image processing and analysis. In recent years, many theories and technologies on edge detection are presented. The theories and methods on edge detection still have much defect to be improved. For instance in the detection accuracy and noise removement aspects,these theories and methods are not satisfactory.In this thesis,some traditional and emerging edge detection methods are being summarized.Then some innovative and exploratory work is being done on gray image edge detection as follows:Commonly used image filtering methods are investigated and analysed in detail and a two-stage filtering method which combines improved median filtering and adaptive filtering.Serveral templates are adaptively used in improved median filtering and damage to the image is minimized during the process of eliminating salt and papper noise.Meanwhile, image edges are well protected during the process of eliminating gaussian noise by adaptive filtering.Sobel operator's characteristics and Laplacian operator's characteristics in edge detection are investigated in detail.Then edge detection algorithm that based on gradient multiplication and edge detection algorithm that based on edge classification are designed.They are two edge detection algorithms that combine Sobel operator and Laplacian operator.Characteristic of gray-scale change of the edge is researched.According to the feature that the edges are mainly locate at the position where gray contrast is large and local differences in gray is defined.According to the feature that the gray difference change along the orientation of the edge is small while the gray difference change is significant along the orientation that perpendicular to the edge and edge likelihood is defined. Finally edge detection algorithm that based on local gray difference and edge detection algorithm that based on edge likelihood is defined.These are two template-based edge detection algorithms.Defects of some existing fuzzy enhancement algorithms are investigated in detail. Then Improvements on computation, selection of enhancement transformation, selection of crossover point and adjustment of enhancement transformation are proposed.Finally an edge detection algorithm that based on improved fuzzy enhancement algorithm is designed.In order to guarantee the reliability and integrity of the edge that we detect,dual-threshold is used.