StatusThe thesis was presented on the 7 November, 2008
Approved by NCAA on the 22 January, 2009
Abstract– 5.01 Mb / in romanian
– 5.08 Mb / in russian
ThesisCZU 004.89 (075.8)
16.00 Mb /
This paper deals with the important and actual scientific problem of restoration of defective images of scenes. The research problem provides restoration of scenes images by technical system on the basis of reproduction of some phenomena of human intelligence used for restoration-recognition of images.
For solving the problem of image restoration, the cognitive, integrated operational and domain models of process of restoration are elaborated. The cognitive models including base elements (two objects connected by the relation of dominance), restore defective images under formal names and functions of objects, on the basis of knowledge of the identified objects of a scene, their connections, and also by control of objects parameters. Integrated operational models reproduces sequence of interaction of associative memory with operational intellectual system of a human brain in the process of decision of a problem. In domain models search of the domain of the phenomenon of an external world in which the real object (original) are stored was realized. In case of absence at any moment of time of the domain of the phenomenon with required parameters – such domain is recreated. The offered ambiguous, connected and directed associations establish associative ties between objects of various classes, providing a wide range of associative objects for restoration of scenes images.
Cognitive, integrated operational and domain models are realized by the intellectual processors constructed with use of neural networks and associative memory, which are simulated by program neural network simulator NNToolbox from MATLAB 7.0. Intellectual processors are the basic structural elements of intellectual system of images restoration (ISIR). ISIR also includes the developed original database of images, base elements, associative objects, functional domains and domains of phenomena Date Domen.
Realization of cognitive models by means of ISIR provides restoration and semantic designing of scenes images under defective images of the separate objects presented in base elements. It allows to restore the defective image of a scene within composite and semantic conformity with the true image. At the 1-st stage of restoration ISIR identifies true images and restores separate images of a scene with rather small degree of the deficiency to which restoration ISIR is trained. At the 2-nd stage of restoration ISIR applies base elements to restore defective images with high degree of deficiency. At the 3-d stage of restoration ISIR carries out the process of semantic designing of a scene. Within this process, hypotheses of possible presence of objects of base elements are considered.
ISIR effectively realizes the cognitive, integrated operational and domain
models of restoration of scenes images. It is proved by the received experimental