Research On Cloud Model And Tolerance Granule-Based Image Retrival Methods

Granular computing is an effective tool to research complex problems, huge amounts of data mining and fuzzy information processing, which fuses multidisciplinary research results about rough set, fuzzy set and artificial intelligence, etc. In recent years, it is widely applied to artificial intelligence, data mining, machine learning, image classification and retrieval, and other fields. With the coming of information era, digital images begin to rapidly expand and gradually become the main carrier of human communication and information transmission. In the face of the vast digital image databases, it is necessary to study an image retrieval technology with high-efficiency. Many scholars deal with image retrieval problem by granular computing theory and have obtained some achievements. However, image features extracted hierarchically from images in existing models of granular computing, with some uncertainties, cannot conform to eye’s perception psychology. Cloud model is an uncertainty transformation model between qualitative concept and quantitative expression, which can use three digital features to describe uncertainties of the qualitative concept, and it is widely applied to many fields, such as data mining, image processing, and others. At present, dealing with image problems by means of cloud model theory has received attentions from researchers both at home and abroad and has become a new research hotspot in the field of image processing.In the paper, uncertainties of extracting image features hierarchically from images are analyzed based on existing models of granular computing, the key technologies of image retrieval and cloud model theory are studied in depth, the cloud model and tolerance granule-based image retrieval novel methods are put forward. The main innovation points of this paper are listed as follows:(1) Texture expresses distribution law of gray space surrounding pixels and is an important feature to describe images. In order to extract texture feature conforming to the human eye’s perception psychology, the gray level-based original object set and tolerance relationship are defined and the gray relationship system is built.In the existing granular computing model, in term of texture feature extracted with uncertainties, an algorithm of extracting cloud model grid point based on gray level is presented. On this basis, a cloud model-based gray tolerance space model is built and a new texture similarity measure is defined. Then an image texture recognition and retrieval method based on cloud model is proposed. The simulation experiments are done and their results show that the proposed method can effectively improve the efficiency of image retrieval.(2) Aiming at the problem that extracting color features in the nonuniform color space RGB cannot depict images well, in this paper we first convert color images to the uniform color space CIELab and construct an object set for color images in this color space. For image features with uncertainties extracted on the basis of the spatial location grid points, this paper applies the related theory of cloud model to extracting grid point of this color space and then a tolerance relationship system based on CIELab color space is constructed. Thus. a cloud model and tolerance granule-based space model is constructed in CIELab color space. Finally, color and texture features of color images are extracted by this space model, then a cloud model and tolerance-based color image retrieval method is proposed, which is applied to image retrieval for color image. The simulation results prove the validity of the proposed method.