Correcolor-Based Image Retrieval

Color is the robust, visualized and basic feather of an image. It is popular to retrieve images in Content-Based Image Retrieval system based on color feather of images. Overall color histogram is a simple and efficient method to compare images. However, it also has some limitations. An overall color histogram provides no spatial information, neither any semantic information in an image; it merely describes which colors are present in the image, and in what quantities.Consider the facts that people are interested in the obvious objects in images if which have; in addition, the semantic relations among objects in the images could be described by the relation among colors in some sense. A new color-based image retrieval method called correcolor-based image retrieval is proposed. It employs the algorithm Apriori to find the relative colors also called correcolors with a row of an image as a transaction, and then extracts the correcolor histogram from the rows that contain all the correcolors. The similarity among images is ranked by the distance of their correcolor histograms.The correcolor histogram has its advantages. First, it locates the interesting rows of images while reducing the interference of the non-interesting parts by pruning operation. Second, the rows of an image contain all the correcolors would reveal the semantic relations among the objects in the image. According to the results of extensive experiments, it is more efficient for correcolor-based image retrieval method than classic ones on the image with obvious objects.