Image Indexing in DCT Domain

Abstract

With the rapid growing of multimedia technology, more and more multimedia content are disseminated in the network or stored in the database. Image data is one of the multimedia types to be seen or accessed for the users in the Internet or from the database. Searching the related images by the querying image content is helpful to the management of image database. Most of the images are joint photographic experts group (JPEG) file format. Full decompression of JPEG file to spatial domain to extract the features for indexing takes time. Therefore, the research of JPEG file format indexing technique is worthwhile because it does not need the extra time to fulfill inverse discrete cosine transform (IDCT). Essentially, the proposed method is a content-based image retrieval (CBIR) system that can retrieval the images from the image database. In order to increase the efficiency of processing time, the proposed method fulfills partial decoding to JPEG file directly instead of full decoding and the feature vector extracted from the partial decoding can ensure the precise indexing as well. The system can output ranked images according to the similarity values in short time. In addition, the system has the property of robustness to rotation, scaling, translation, darkening, lightening, cropping, noise corruption, etc. Simulation results demonstrate that the performance of the proposed method outperforms the other existing DCT-based image indexing skills both from the aspects of robustness and processing burden