Alexander Dimai,
Scene Configuration Based Preattentive Similarity
Assessments for Content Based Image Retrieval.
Ph.D. Thesis, Communications Technology Laboratory, Swiss Federal
Institute of Technology, ETH, CH-8092 Zürich, Switzerland, February 1998.
This thesis addresses content based image retrieval, an important topic in information management. The rapid expansion of computer networks and the dramatically falling cost of data storage are making image databases increasingly common. The growth in number and size of image databases creates the need for new methods to access and to search images. Conventional database search is based on textual queries and provides only a partial solution to the problem, because perceptual cues contained in an image are not fully exploited. Consequently, the ultimate goal is to find image descriptors capturing perceptual information of an image and allowing similarity assessments between image scenes with a sophistication and versatility comparable to that of human. My thesis asserts that scene configuration based image descriptors and a non-linear comparison scheme indexes robustly and effectively preattentive contents of images. For supporting this thesis, this dissertation inquires into three components: 1) evaluation method, 2) descriptor extraction, and 3) comparison scheme. ...