The Research And Implement Of Color-Based Image Retrieval System

The development of multimedia technology has promoted the integration between people and computers. As an important form of multi-media, digital image has been used more and more widely. And the number of digital images is growing rapidly. How to process the digital images has become an urgent issue. Image process includes a lot of directions. Image retrieving techniques is an important content of the study of image processing. Image retrieving has two different methods. Since traditional method of image retrieval has some drawbacks, the Content-Based Image Retrieval(CBIR) is a new method, which is worthy to study and develop. This paper focuses on the color-based image retrieval, which is an important method of CBIR.The method of CBIR extracts the characteristics of the image and by matching the characteristics of the target image with those in the image library, retrievals the image which is similar to the target image. Characteristics extraction and characteristics matching are the core contents. CBIR has been widely used in public security, medical, military, aerospace and some other fields.Color-based image retrieval is an important method of CBIR. The color of an image is the most direct and easiest of all the characteristics to be extracted. Color histogram is the easiest and most commonly used methods to describe the characteristics of an image.After studying the theory based on color-based image retrieval, this paper describes the design and realization of a color histogram-based image retrieval system .The system's developing platform is Visual C++ 2005, using MFC class library. With the system, the user can input a BMP format image, and then select a picture library for image retrieval. The computer calculates the values of its similarity and displays them to the users.This system has utilized two algorithms. The method of local cumulative histogram is used for characteristic extraction. This method improves the common histogram. When extracting common histogram, the system divides the value of color into several grades. However, all the full values of the colors of an image often can not cover all the color grades. Therefore, there will be zero in the values of characteristics. These zeros make troubles for the next step. On the other hand, the method of cumulative histogram cumulates the values of characteristics, which can avoid zeros. Furthermore, common histogram reflects the overall characteristics of an image but neglects the position relationship among pixels. To some extent, the local Histogram takes into account the relationship. The method divides the image into some blocks, marks every block and gets the histogram of every block. Then, after comparing the same blocks of different images, the system matches them. Cumulative local histogram method combines the advantages of the two methods. Therefore, the method is more accurate for characteristic extraction.The system uses multithreading approach for the image retrieving. Multithreading is the method that the system uses many thread to finish a single program. After extracting the characteristics, the system creates a new thread for retrieving, and halts the old thread. When the new thread has finished retrieving, the system continues to process with the old one. Using multithreading approach can improve the efficiency of the program and reduce the response time.The system is mainly comprised of two parts. The first part finishes the function of displaying histograms. The system loads the target image into the memory, and copies it. Then, the system gets the information of pixel in the target image and stores them into different arrays. With the information, the system draws the histogram of the target image from the memory. At last, the system displays the image and its histogram with its relative control. The function of image retrieval is executed as the second part. With the image library chosen by the user, the system uses the method of multithreading to match all the histograms of the images in the library with the histogram of the target image, and stores the values of similarity. When all the images in the library are processed, the system ranks the similarities and displays the values sequentially in the controls. In one word, image retrieving is a kind of image processing technology, which is closely related to people's daily lives. CBIR is the direction of the development of image retrieving technology. Therefore, the research of CBIR is of significant importance and requires much further efforts.