Image Retrieval System Based on Density Slicing of Colour Histogram of Images Subareas and Colour Pair Segmentation
Jehad Odeh 1, Hjah Fatimah Ahmad 2, Mohamed Othman 3, and Rozita Johari 3
1Faculty of Information Science and Technology, Multimedia University, Malysia
2 Multimedia Department, University Putra Malaysia, Malysia
3 Computer Network Department, University Putra Malaysia, Malysia
Abstract: Techniques to identify objects within an image and searching for similar objects in the database is not claiming a lot of progress, due to the limitations of the capabilities of the existing techniques and algorithms in image processing and computer vision to perform such task. In this paper, a new technique based on slicing the images to equally sub-areas, then applying the density slicing to the colour histogram of these areas combined with the colour pair technique is presented. We tried to overcome problems related to the original colour pair segmentation, as well as overcome problems related to the computational complexity in histogram localization through proposing density slicing or multiple thresholds. In this paper, new techniques proposed, new ranking formula, and a complete framework with the interface consideration.
Key Words: Colour pair segmentation, content-based image retrieval, city block, density slicing.
Received June 24, 2003; accepted August 25, 2003