Research On Proximity Space Theory Based Image Retrieval Method

In recent years,with the rapid development of communication technology and computer network technology,the capacity of the digital images is explosive growth at an alarming rate.How to efficiently manage,access and accurate retrieval more and more great image databases have become the research hot topic in the field of image technology.So image retrieval technology is becoming an important research direction.The main works of this thesis is the color-based image retrieval,it includes three aspects:feature extraction,similarity measure and performance evaluation,the main researches are as follows:(1)Color feature extraction method is proposed by combining proximity space theory.First of all,color space conversion is used in the process of image feature extraction,and then 72 kinds of color features will be obtained by the color quantization;and then the idea of the proximity space theory is utilized to extract color histogram features of the query image and images of database respectively;calculate the similarity between the feature vectors of two images by the Euclidean metric of the probe function;finally,the effectiveness of the proposed feature extraction method is verified by simulation experiments.(2)Using the dominant grain structure,the matching method based on improved dominant grain structure similarity is proposed.At first,take Top-ranked color features in 72 kinds of color features to match images,in order to reduce the complexity in the process of image matching;then,according to the properties of the proximity space theory,by comparing the distinguished color feature values in the color feature vectors between images.Judging roughly whether two images are relevant images;design an improved dominance granule structure similarity method,calculate the similarity between query image and images of database;finally,the effectiveness of the similarity matching method is verified by experiments.(3)Combining the feature extraction and similarity matching method,an image retrieval method based on proximity space theory is proposed,in order to achieve more effective precision rate and recall rate,the experimental parameters are varying for different databases,the value of the parameter is determined by the size of the color features frequency of database.Experimental comparison results shows:The main performance indicators are better than the existing typical algorithm,the algorithm's average precision rate of the top 10 reach to 68.30%on the data set which contains 10 categories 1000 images,and the algorithm's average recall rate reach to 83.87%on the data set which contains 20 categories 1440 images by the simulation experiment.