Research On The Technologies Of An Adaptive Watershed Digital Image Segmentation

To human beings, image is one of important information gained from external world and the main information source to cognize the world. In recent years, the demands from many industries are increasing with each passing day; at the same time, computer and image processing cell provide a strong ensuring platform for the research and application of image processing technology.Image segmentation is a basic technology of image processing and key step of later image processing and analysis, yet always a classical hot potato. Because of the complex self-nature of image information and influence from external factors, the segmentation schemes might bring mistaking segmentation, segmentation excursion, difficulty of constructing models, etc.Facing on the disadvantages of image segmentation algorithms and requirement of current image automatic and real-time segmentation, this thesis conducts corresponding researches, which is mainly described as follow:Firstly, the developing process of digital image segmentation technology is all-sidedly surveyed, and the classical solving schemes in this research area are analyzed and compared. At the same time, this thesis explains systematically advantages and disadvantages of different schemes, and gives corresponding feasible improving schemes.Secondly, in view of weak antinoise ability existing in all digital image segmentation schemes, this thesis conducts a deep and systematic research. Through comparing performances with other image segmentation schemes, this thesis discusses advantages and potential problems of the image segmentation scheme, which is based on watershed algorithm. Meanwhile, this thesis also compares the advantages and disadvantages of several kinds of improved watershed segmentation technologies, and establishes the foundation of my self-adaption watershed segmentation.Thirdly, oriented the application of image segmentation technology, an improving watershed image segmentation scheme is proposed systematically and modularized to three steps: self-adaptive noise filter, quick-sort & adaptive delamination and area growth, to complete a whole process of image segmentation. Meanwhile, the feasibility of this improved scheme is discussed theoretically. Finally, applying the self-adaption image segmentation scheme, an effective and high real-time image auto-segmentation system is designed. And the validity of my improved self-adaption watershed segmentation scheme is verified by experiment with two kinds digital images, simple and complex one. Meanwhile, the availability of improving segmentation scheme proposed is verified. And then, this thesis simulates the improved segmentation scheme with image processing tool-box of Matlab6.5, and compares with several classical improved watershed segmentation schemes, verifies the feasibility, self-adaptability and segmentation accuracy of the improved segmentation scheme in this thesis.