Research On Various Image Edge Extraction Algorithms

Image edge is one of the most fundamental characteristics of an image, including the most information of the image. Edge extraction is the classic technology in the image processing. Its performance will have direct and significant impact on image characteristics description, recognition and understanding.The paper, based on the comprehensive survey of the research situation and development trend of edge extraction, focuses on the theoretical analysis, simulation experiments and performance comparison of different algorithms such as several classical algorithms, fuzzy theory and mathematical morphology.The paper firstly does simulation experiment on several classical algorithms using various images, and then analyzes their performance. Then two improved methods are proposed based on the advantages of different algorithms. One is the combination of Laplacan operator with the function of sharpening and mathematical morphological which can extract continuous and complete edge. The other is improved Pal.King algorithm which replaces matrix transformation with various threshold segmentation technology, where Otsu threshold method is chosen as the representative of these improved methods. Finally the experiment results of all algorithms are compared and evaluated.There are two kinds of evaluation criterion for image edge extraction, i.e. subjective and objective criterion. Firstly, the paper adopts the subjective method to evaluate the continuity, smoothness, refined degree and edge localization performance. Secondly, the paper evaluates anti-noise, positioning performance and running time using objective criterion. Finally, both criterions are unified.