Studies And Improvements On The Algorithms Of Classic Edge Detection

Digital image edge is the most basic feature of the image, which contains a lot of useful information for identifying images. Edge identification and extraction for the entire image scene recognition and understanding is very important, and edge detection of objects in pattern recognition is an important part of the feature extraction. But image analysis in the field of classical studies has not been completely resolved. How to eliminate the false edges brought by image noise, and to make sure acquiring the accuracy information of the image edge, is one of the problems.This thesis focuses on digital image edge detection in the analysis of difficult problems which have not been solved. Studying on edge detection produced a number of classic algorithms. This thesis introduces the classical algorithms in detail and improves these classical algorithms. This thesis first introduces the widely used edge detection methods such as Sobel, Roberts, and Canny algorithm, simulation experiments of these algorithms in noise and noise-free cases, and a comparative analysis of test results of these algorithms. However, these algorithms are very sensitive to noise. When the image contains noise, the noise will be detected as edge points, while the real edge of the interference due to the noise may also be undetected. According to these disadvantages of the above algorithms, this thesis improved the traditional Sobel edge detection algorithm. This improved algorithm extends the traditional Sobel operators from2templates on the horizontal and vertical directions to8directions. And this improved algorithm uses the image gradient mode, the edge point of the relevant and binding on the edge of the image edge tracking, and gradually the exclusion of false edges, accurate positioning edges to exclude noise, and thus complete the whole image edge detection. Next this thesis describes an algorithms of edge detection based on fuzzy set theory, and points out the advantages and disadvantages of this algorithms. Then this thesis improves this algorithm and get well test results. At last this thesis introduces a digital image edge detection algorithm, which is based on mathematical morphology. This algorithm combines the scale elements and structure elements of the image to get the edges from image. This algorithm can maintain the image edge integrity and continuity in high quality.