Comprehensive Contourlet Transform And The Color Histogram Image Retrieval

With the ongoing development of multimedia technology, the number of digital images is increasing rapidly. There is an urgent need for efficient techniques to search and query the needed image in huge database rapidly and effectively. Traditional image retrieval technique refers to the retrieval of images from a database using information solely from accompanying text indices, which can not meet users'demands. Then, content-based image retrieval technique is coming into going.In this paper, techniques of color abstracting and texture abstracting of low-level visual features are discussed, and contourlet transform and its application in image retrieval are studied systematically. The main research content is summarized as follows:1. Lots of research work around the key techniques of content-based image retrieval(CBIR) has been done, and some color feature and texture feature extracting methods are realized in this paper.2. Gabor wavelet transform and contourlet transform are studied. The image feature extracting algorithms based on Gabor wavelet transform and contourlet transform are experimented. The algorithms uses Euclidean distance as the similarity messure method.The experimental results show that contourlet transform has better performance on the aspect of describing texture feature and precision ratio than Gabor wavelet transform.3. In order to overcome the insufficiency of image texure feature lacking color content, a new retrieval algorithm based on the contourlet transform and cumulative color histogram is presented. The algorithm uses energy-feature of the different subbands in contourlet domain and cumulative color histogram in HSV color space as the eigenvector of image, and Euclidean distantce as similarity measure method. Experiments indicate that this algorithm gives higher precision ratio than Gabor wavelet transform and color histogram.4. A "Contourlet and Color based image retrieval prototype system"is implemented with Visual C++6.0. All the algorithms mentioned above are evaluated on this system in standard texture image database and Corel image database.