Research On Shape Description And Matchingalgorithm Based On Contour Curve

Shape matching technology is a important task of machine vision, patternrecognition which have a very wide range of applications in target tracking, visualnavigation, image registration and stitching, content-based image retrieval and otherareas. Although many scholars have already done a lot of research work on the shapematching, there are still many unresolved problems, such as the shape of the occlusion,matching efficiency and anti-noise performance and other issues. This paper from theoutline of objects, deeply study on the description and matching algorithms, and thealgorithms are applied research in graphic stitching and image retrieval. Main contentsand works of this paper contained three parts as follows:1. Several key technologies and methods of shape matching are introduced. Thereare a deeply research and analysis on shape matching, first to the shape descriptionmethod, respectively, from the shape descriptor based on contour and region-basedshape descriptor and have a detailed analysis on the advantages and disadvantages ofeach method; and finally we do a deeply research on shape matching algorithm,especially on the similarity measure criteria.2. Improvement of the traditional concentric method for the shape description, themethod is basic on concentric circles firstly, and the description set are consisted by agroup of radius and three groups of each angle, which groups angle arei, i, irespectively, then to determine the shape similarity through the comparison ofdifference of concentric circles descriptions.The improved method is robust to rotation,translation, scaling.3. A representation and matching algorithm for planar curve based on sub-matrixand concentric circles is presented in this chapter. The algorithm includes two steps,namely rough matching and exact matching. The rough matching uses a sub-matrixmatching method. As for exact matching, first, the representation of curve usesconcentric circles and then measures their similarity through two curve representationsets of concentric circles. The algorithm which is robust to the translation, rotation andscaling can be used to match block objects and reassemble the graphic. The experimentresults show us the effectiveness and feasibility of algorithm.4. Multi-scale Distance Coherence Vector Algorithm for shape matching isproposed to improve the robust of Distance Coherence Vector Algorithm in this chapter.Firstly, the improved algorithm convolutes Gauss, and than extract Distance Coherence Vector for Original image and evolved images respectively. Similarity of images ismeasured by reasonable weight distribution of the each vector. The improvedalgorithm not only is robust to translation, rotation and scaling, but also has the goodperformance to the noise. There are a better result when make it to image retrieval andthe contrast experiment results show us effectiveness and robustness.