The Research Of Image Retrieval Algorithm Based On Overall Local Color Features And Texture Features

With the great development of multimedia and Internet technology, image data resources increasing dramatically and being used in various fields, the traditional text-based image retrieval method has become increasingly unable to meet the image information needs. Nineteen ninties, the content-based image retrieval technology was born,and quickly became the research focus in the field of the image information retrieval.This paper studies mainly in the following aspects:(1)the current development and research focus of the content-based image retrieval technology are introduced systematically,the related key technologies of the content-based image retrieval are learned and explored, which provides a theoretical basis for the algorithm of this paper.(2) In view of the defects that the traditional color histogram can’t express color spatial distribution information and the fixed block is sensitive to image rotation transformation, this paper presents a overall local color feature algorithm.First, use color moments to extract the image global color feature, and in order to describe the image color feature accurately, we use non-uniform color block to extract the image color spatial feature, then combine the two as the integrated color feature of the image.The experimental results show that:compared with the traditional fixed block, this algorithm can not only retain the image color spatial feature but also improve the rotation invariant properties to a certain extent; compared with the traditional global color histogram、color moments and color coherence vector, the algorithm significantly improves the image retrieval performance.(3)Aiming at the defects that a single color feature can’t describe the image information completely, which cause low performance of image retrieval,this paper has researched on the fusion technology based on color and texture.This paper puts forward two kinds of comprehensive features algorithm according to the fusion research of the color feature extraction algorithm in (2),gray level co-occurrence matrix and intersecting cortical model. The experimental results show that:the combined overall local color feature and gray level co-occurrence matrix algorithm,the combined overall local color feature and intersecting cortical model and gray level co-occurrence matrix algorithm greatly improve the performance of the image retrieval,and the latter altorithm performs better in these two algorithms;compared with the single feature algorithm,the comprehensive features based on color and texture can better represent the image information, enhance the flexibility of image retrieval and greatly improve the image retrieval performance.