Research Of Shape Analysis Algorithm Based On Fourier Descriptor

With the emergence of a large mount of image database and the increase of multimedia information, content-based image retrieval (CBIR) technology has becoming a very important research and application field. Among them, the content of image information is the lower visual features, such as color, texture, shape and their combination characteristics. Due to the shape image feature has some advantages which the other characteristics cannot match, the main research content of this paper is the shape analysis algorithm used in image retrieval.Shape analysis algorithm mainly contains the edge detection algorithm used in the pretreatment of target shape and the shape representation algorithm when run in the step of shape feature extraction. Firstly, the edge detection algorithm is the basis of the analysis of a shape. And whether the boundary shape of the target object can be accurately extracted or not will directly affect the performance of the feature extraction algorithm and the effect of image retrieval. After the comparison and analysis of a variety of edge detection algorithm, we choose Canny operator to detect the edges of image. And also achieve good results in the simulation experiments. Secondly, shape representation algorithm is the key to the shape analysis technology, and its main purpose is to maintain and accurately represent the important shape information used for image retrieval. About the algorithm, we choose the Fourier description, which has the most widely application and potential, to shape representation algorithm research. We put forward Gauss multi-scale elliptic Fourier descriptor(GMEFD) and two-dimensional Fourier descriptor with the phase information(PGFD) based on the mature Fourier theory. In the MATLAB simulation experiments, we prove that the GMEFD has such nature as simple calculation, fast computation and well generality, and the PGFD has the properties of very high retrieval accuracy and strong ability of distinction between different shapes. Thus, they are suitable for different environment of retrieval requirements. GMEFD is applicable to the environment with simpler shapes and higher retrieval speed demand, however, PGFD is suitable for more complicated shapes and higher retrieval accuracy. Finally, we design and implement an image retrieval prototype system based on the target shape used the GUI development environment GUIDE provided by MATLAB, which help us realize an interface between users and the retrieval system.Fourier descriptors are the shape analysis algorithm with the most rapid development and most widely used. We believe that the combination between Fourier descriptors and the other kinds of descriptors even the other contents in a image is a very important development trend in the future, and which also is the main content of our next step work.