Shape Extraction And Recognition,

With the development of multimedia techniques and the availability of high-speed computers and large capacity storage devices, images play an important role in data processing of computing system. Digital image acquiring devices produce a large amount of images. Recognizing an object by its shape is a fundamental problem in computer vision, and typically involves finding a discrete correspondence between object model and image features. It can be utilized in many kinds of image context analysis and retrieval system. This thesis focuses on the problem of the shape-based representation and extraction of image context, and object location.The thesis presents an efficient shape-based algorithm for object recognition and location, including shape features extraction, the correspondence between object model and image features, and object location.Of many methods, straight-edge-line-based shape description is the best one not only because it is abstract sufficiently, but also because it provides abundant information. Shape is described as the distribution of straight edge lines with specific properties such as position, length, orientation, width, contrast, steepness, and so on.Object model consisting of the set of straight edge lines with some specific values of properties or within some ranges is the object shape feature derived from priori and process target.Besides values of properties such as position, length, and orientation of the straight edge line, the algorithm of straight lines extraction, which is based on the concept of line support regions, can also compute contrast and steepness across line support regions. With global analysis and parametric control, all above provide abundant information so that speed and correctness of corresponding object model with image features can be increased dramatically.The key problem of correspondence between line segments forming an object model, called model lines, and line segments extracted from an image, called data lines, is to quantify the degree of match. A match error ranks matches by summing a fit error, which measures the quality of the spatial fit between corresponding model lines and data lines, and an omission error, which penalizes matches which leave portions of the model omitted or unmatched. Inclusion of omission is crucial to success when matching to corrupted and partial image data. Because of the fragment, any number of model segments may map to a data segment, and any number of data segments may map to a model segment.New optimal matching algorithms use a form of combinatorial optimization called local search, which relies on iterative improvement and random sampling to probabilistically find globally optimal matches. A novel variant has been developed; subset-convergent local search finds optimal matches with high probability on problems known to be difficult for other techniques. Specifically, it does well on a test suite of highly fragmented and cluttered data, symmetric object models, and multiple model instances.