Image Retrieval Research Based On Color Density Feature And Weighted Region

With the development of the Internet and intelligent terminal equipment such as mobile phone,the network server has a large number of digital images,how to quickly and conveniently retrieve the images of their concern from the database with a large number of images is a natural demand,from a technical point of view is also a very meaningful and challenging topic.Image retrieval technology is a kind of technology to solve these problems.The technology of image retrieval mainly divides into two kinds: text-based image retrieval(TBIR)and contentbased image retrieval(CBIR).Content-based Image retrieval technology is used to identify images by extracting their digital features.Digital feature is a bridge of computer-perceived image.The study of image features has been a key problem in the accompanying image retrieval technology.In addition,the image content is complex and can not be fully expressed by single feature,and comprehensive utilization of multiple features is an effective way to solve such problems.In view of the above problems,this paper focuses on the characteristics of the underlying image.The main research work is as follows:1.The underlying features of existing content-based image retrieval(CBIR)are analyzed and discussed.For the global characteristics,the color histogram feature and the LBP texture feature are discussed in detail,and the generation process of the image feature descriptor and the image retrieval process by using the feature descriptor are discussed.2.In view of the lack of spatial information for the global feature of color histogram,this paper presents an improved method based on color coherence histogram of region pixel density.First,the image is quantified in HSV space to 72 colors,or 72 bins.Then check the space pixel density of each bin,according to the degree of density of a bin of the pixels into sparse and dense two categories,and finally formed the spatial structure information color characteristics.The modified color features are used for image retrieval.3.The LBP texture detection operator is discussed in detail,and the image retrieval performance of 3 kinds of important LBP texture feature detection operators is analyzed and validated experimentally.In order to improve the performance of image retrieval,the central region of the image is divided into key areas of obtaining information according to human visual characteristics.Therefore,this paper proposes a feature region weighting algorithm,in which the key area of the information is obtained,and the feature is given a larger weight value.The experimental results show that the feature weighting strategy has a good effect in image retrieval.4.In order to get the key region of the information to be adaptive to the image content,the saliency region of the image is detected by the method of regional detection,then the image feature is weighted in the saliency region.This method is applied to image retrieval,and the retrieval results are analyzed and summarized.