Research On Image Understanding Methods Based On Visual Perception

Image Understanding is the subject of automated extraction of information from images. The essential technologies of the science include image segmentation, image component modeling, image analysis and reasoning. The research in this field involves development of automated ways to detect various characteristics of objects in a picture (indoor/outdoor, people, faces, trees, buildings, etc.) that may be useful in future applications such as database management or automatic creation of photo albums. Image understanding is a major step towards the development of Artificial Intelligence, which involves systems that will make certain decisions or perform some activities based on their interpretation of an input image. Image understanding also has widespread applications in medicine and defense.By summing up the research result of the rescent years , this dissertation presents the fundamental frame structure of image understanding , and analyzes the function of every section with to achieve. This thesis consists of six major components. The first addresses the method of image understanding. In this component current theories of the visual perception are introduced. Starting with a brief overview of low-level visual processes, which contribute to the recognition of objects, such as the perception of structure from shading or texture, this component mainly concentrates on the method of image understanding. The described theories are assessed with respect to their biological plausibility. Evidence from psychological studies is given, which either supports or contradicts the different computational models. Besides this, recent results from physiological studies reflect the hierarchical processing of 3D information in the primate brain. The second component deals with the implementation of segmentation technique, which is applied in order to detect the boundaries of the objects that define the scene. This component mainly gives out the algorithm of resegmentation , moreover the evaluation of the performance of the result is carried out by analyzing the information associated with the singular edge points. The third component surveys the various techniques for shape representation and analysis. The representation can be either edge based (descriptions of edges and the relationship between edges) or region based (you may have to fiddle with your edge detection algorithm to produce closed regions). Specifically, a review of boundary shape analysis methods and chain-code representation techniques is presented. The fourth component focuses on integrating advanced knowledge representation technology with image understanding technology and perceptual reasoning techniques. The fifth component is concerned with the use of feature-based match similarity measures and feature match algorithmsin image understanding. The component addresses the object recognition and reasoning by using the information resulting from the application of the segmentation algorithm. The recognition stage consists of matching the features derived from the scene regions, while the reasoning is addressed using uncertain reasoning. The last gives a simulation tracking system is designed based on runway recognition algorithm.In this dissertation ,lots of research work has been done around some key techniques of IU (Image Understanding). The presented study is the current research focus of image understanding and image processing. Thus its research has both the theory and the application value. The contribution of this thesis was that we presented a realize method of image understanding based on visual perception, and this study lay foundation for further research.