The Research Of Image Retrieval Based On The Correlation And Polymerization Histogram

In the Content-based Image Retrieval, color as one of the most important visual information has been widely used. It reflects the gray scale information of every component of the color space in pictures taking the form of histogram, and it has such features as simple, perspicuity, and interference killing etc. But at the same time it counts the color's gray scale information, it loses the color's information of its space distribution. How to combine every component's gray scale statistical information with its space information well is an important subject people have studied for a long time.This paper puts forward a kind of search algorithm based on the correlation and polymerization histogram. By using the concept of set's polymerization degree, the space polymerization information was incorporated into the color statistical information in the algorithm. It has many advantages. For example, it is not sensitive to the picture's transition, rotation and size change; its retrieval performance has little to do with the size of the database.Meanwhile, in order to make the search algorithm of the correlation and polymerization histogram have a better search precision, according to relative documents, this paper makes the algorithm of the correlation and polymerization histogram in the HSL color space. And moreover, every component of the HSL color space was quantified unequally. Then, the distance of the histogram was counted combining the L1 distance, weighted L1 distance, neighborhood weighted L1 distance and weighted L2 distance. In the detail operating process, the picture was also divided into sixteen pieces which are the same in size. This makes a further progress of the polymerization histogram's search precision.As to testifying the validity of the algorithm, this paper builds a CBIR system based on the correlation and polymerization histogram and uses it to carry out some experiments about the search algorithm of the correlation and polymerization histogram. In these experiments, the algorithm's recall ratio, precision ratio, the time of pretreatment, query time and effect of the size of the database to the retrieval performance were computed respectively. Base on this, the results are then compared with the color histogram and the color correlogram. The data from experiments suggest that the search algorithm of the correlation and polymerization histogram has a better retrieval performance compared with the search algorithm of the color histogram or the color correlogram. When this algorithm is used in the retrieval of the satellite nephogram, it works efficiently.