Content-based image retrieval in databases and Internet

The article deals with an information model of the content-based image retrieval system. Database architecture is proposed to be constructed using the algebra of completed and closed groups — elements of the theory of active perception. The algorithms can find not only original images but also noisy, edited and corrupted ones. There is a search for similar images, which can compete with the existing methods of image retrieval on the Internet on the reliability and performance. The experimental results are obtained for a random image database of about 10 million objects.