Image Segmentation Algorithm Research

Image segmentation is the techniques and the process to separate the image into regions which have different characteristics and to extract the interested objects from the image. Meanwhile, image segmentation is a vital important and very difficult problem in many fields such as image processing, pattern recognition and artificial intelligence etc. It is also the principal and crucial step in computer vision. The application of image segmentation is extraordinary wide, which involves nearly all fields related to image processing. Therefore, image segmentation has been a research focus in image approaches since 1960s.Firstly, in this thesis, several aspects are summarized, which involves the objective and the significance of image segmentation, the status of image segmentation in image engineering, the research actuality to the segmentation techniques and the classification to the algorithms of image segmentation. And the classical algorithms and the modern ones are also retrospected.Furthermore, the contents, including the approaches and theories related to image processing and the typical algorithms of edge detection, are briefly introduced in this paper. On the groundworks of the former researchers, the performance variation of the algorithms based on edge detection are mainly discussed in image and noise contained image processing under different resolution. Several experiential conclusions are drawn from the experiments. This provides the premises to improve the approaches and to exert their strong suits in practical application.In addition, it is discussed that the basic theories of wavelet, including the definition of wavelet transform, the dyadic wavelet, multiresolution analysis, and the research evolvement of wavelet analysis in image segmentation. And it is also researched that two multithreshold image segmentation algorithms based on wavelet analysis, meantime, the performance (i.e. the characteristics and the advantages) of the algorithms is drawn through comparison with the typical multithrehold technique.Finally, the research achievements of the paper are summarized, and the future works are also provided.