Semantic Image Retrieval

The number of Web images is increasing at a rapid rate, and searching them semantically presents a significant challenge. Many raw images are constantly uploaded with little meaningful direct annotation of semantic content, limiting their capacity to be searched and discovered. Unlike in a traditional database, information in an image database is in visual form, which requires more space for storage, is highly unstructured and needs state-of-the-art algorithms to determine its semantic content. As Web images tend to grow to unwieldy proportions, their retrieval systems must be able to handle multimedia annotation and retrieval on a Web scale with high efficiency and accuracy. With the exception of systems that can identify or detect music, words, faces, irises, smiles, people, pedestrians, or cars, matching is not usually directed toward object semantics. Recent research studies show a large disparity between user needs and technological supply.