With more and more visual material produced and stored in visual information systems (VIS) (ie, image databases or video servers), the need for efficient and effective methods to index, search and retrieve large images and videos collections has become critical. Large VIS users will want a more powerful method to search for images than the traditional text-based query (for example, keywords). Manually creating keywords for a huge collection of visual materials is too slow for many practical applications. Subjective descriptions based on users' input will be neither coherent nor complete. In addition, the vocabulary used in the description of the visual contents is usually domain-specific.
To overcome these shortcomings, there is a recent approach called content-based visual query, which allows users to specify the query and execute the search based on the visual content of the visual material. The term "content" refers to the semantic structure of images and videos on various levels, ranging from pixel patterns, physical objects, to spatial / temporal structures of visual material. The content-based approach is not meant to replace the keyword approach. Instead, it is considered as a complementary tool, particularly for applications that have large data collection and require a quick search response. The provision of content-based visual retrieval techniques also brings a new synergy between text-based information and visual information of the same material. The merging of different information channels (text and visual in this case) has been used to achieve performance improvements in multimedia databases as news archive. A content-based visual query system requires several key components, including extraction of visual characteristics, feature indexing data structure, distance (or similarity) measurement, fast search methods, integration of different visual characteristics, and integration of visual characteristics with indexes based on text. We focus on issues directly related to image / video processing. Our goal is to investigate automated methods using various visual features useful for content-based visual query. This is a relatively new area, in which some promising works have been published in the literature for different applications. The IBM QBIC system provides semi-automatic mechanisms to extract the color, texture, shape, and structure of the image query. Pentland et al have demonstrated imaging systems based on shape, texture and face. Other researchers have also proposed image search techniques based on shape, color or texture. I