ABSTRACT
In recent years, with the growing popularity of digital photography and multimedia systems, the sizes of many image databases have increased dramatically. To adapt to such changes, a number of international compression standards have been developed for effective multimedia information storage. Image classification in the spatial domain is performed by extracting low-level visual features and translating these features into numerical attributes for convenient comparison between images. Grouping images into meaningful classes for a large image database is essential for content-based image-retrieval applications, where the system can quickly focus on the relevant subset to reduce the retrieval time. In the spatial domain, a color histogram is constructed based on the occurrence frequencies of particular color intensities by scanning the image pixel by pixel. The adoption of compressed-domain features in image classification facilitates the efficient processing and indexing of image database entries without the need for performing the inverse discrete cosine transform (DCT).
