Research On Image Retrieval Methods Based On Description Of The Contour Of Autoregressive Model

In recent years, due to the large number of multimedia database applications, make content-based image retrieval has been an active area of research. Feature-based image retrieval content-based image retrieval is an important component of the feature-based image retrieval in the shape of the object's description. Describes the current shape of a variety of methods have been proposed, most of them are based on object contour. Fourier descriptors, wavelet descriptors outline the boundary moments, etc. But there are two problems, firstly store the shape need volume of data, secondly the problem of translation, scaling and rotation invariant transform still not been solved.To solve the above problem, the paper proposed a multiple linear regression model to describe the shape of the image boundary feature extraction and image retrieval. First, using Canny edge detection operator extracts the edge information, in order to obtain the object closed boundary curve, then analysis of contour with the minimum maximum method, and decomposed into smooth closed curve to the concave or hole boundary. Second, take two-dimensional plane into a one-dimensional graphics data using polar coordinates, and get a very one-dimensional path length sequence from the object boundary contour, to the sequence as a time series model from the regression model, using the model parameters as the shape feature, and the parameters have translation, rotation, scale invariance properties. Third, for the shape data just stored data before the m sampling points, to estimate other data based model, it can reconstruct the original boundary curves without distortion and distinguish the outline of complex targets, achieved with less data to describe the shape of a certain precision contour. Using chain structure for extracted features, and on this basis, using a coarse to fine hierarchical search methods:first, get number of images using feature-based methods, then conduct a more precise match in a smaller subset with the scaling function.Make up the color and texture as the search feature some of the deficiencies, with image retrieval algorithm based the advantage of spatial correlation. Shape-based image retrieval techniques will make life more convenient for people.