Shape-based Image Retrieval Algorithms

As development of the database and computer vision, content-based image retrieval technology has become an active field of research. Content-based image retrieval is the information from low-level to top understanding, analysis and abstraction, obtain the contents in the images. The content of information including the image color, texture and shape features. Content-based image retrieval is a product of computer vision processing technology and database technology. The basic idea is to through the analysis of the image itself contains inherent characteristics and related image connected to the search. Image retrieval technology on mass image information management and access can provide powerful support; have the extremely broad prospect of application, related to social and life in many areas such as digital library, web image search, and biomedical domain, military and national security. Content-based image retrieval technologies include feature extraction, features presentation or description, similarity measure and the evaluation of the arithmetic performance.Shape is an important visual feature and it is one of the basic features used to describe image content. However, shape representation and description is a difficult task. This is because when a 3-D real world object is projected onto a 2-D image plane, one dimension of object information is lost. As a result, the shape extracted from the image only partially represents the projected object. To make the problem even more complex, shape is often corrupted with noise, defects, arbitrary distortion and occlusion. Shape-based image retrieval is a difficult in CBIR, and be the key of in our research.Shape representation and description techniques can be generally classified into two classes of methods:contour-based methods and region-based methods. The classification is based on whether shape features are extracted from the contour only or are extracted from the whole shape region. Under each class, the different methods are further divided into structural approaches and global approaches. This sub-class is based on whether the shape is represented as a whole or represented by segments/sections. Introduce the main categories of these descriptions; give the approach of matching, advantages and disadvantages of various methods.This paper is the main contribution of the paper proposes a method:fast tensor scale descriptor. This approach is to calculate the tensor scale descriptor base on image foresting transform, geometrical shape features are also constructed by using the normalized histogram of the local orientation of tensor scale descriptor, and then, combined a similarity measure approach, a content-based image retrieval method invariant to translation, rotation and scaling transforms. The tensor scale features methods of extraction algorithm complexity and previous two kinds of methods are compared, and show that the method of computing complexity lower. Experimental results show that fast tensor scale descriptor in the image retrieval has very good retrieval results.Review and to describe six major descriptors:geometric moments descriptor, Zernike moments descriptor, Fourier descriptor, curvature scale space descriptor, beam angle statistics descriptor, fast tensor scale descriptor, carries on the experimental analysis, comparison and summarized. Experimental results show that the Fourier descriptor has the storage space is small, short response time and retrieval rate high characteristic, suitable for real-time image retrieval based on shape. Beam angle statistics descriptor and fast tensor scale descriptor have high retrieval rate, but larger storage space and response time is longer, suitable for retrieval system, have smaller images and demand high retrieval rate. With the further development of hardware technology, these two descriptors will have described the research and application of comparative higher the value.