The search algorithm for image in databases

This paper briefly discusses the existing methods for image lookup in video databases. A new method of image lookup is suggested; it is based upon active perception theory which is being developed at the Computing Systems and Technologies department of NSTU. Its key advantage is its low computing cost and thus high performance. According to the method, every image is represented as a 15-dimensional vector. Not only the original images are matched; the algorithm also handles noisy, edited, corrupted, rotated images. An algorithm for matching similar images is proposed as well. Its reliability and performance allow it to compete with existing commercially available image lookup technologies.