Title :
Image reconstruction using bistable gradient neural-like networks
Author :
Chinarov, Vladimir ; Menzinger, Michael
Author_Institution :
Dept. of Chem., Toronto Univ., Ont., Canada
Abstract :
We present a novel bistable gradient neural-like network (BGN) and demonstrate its potential applications for image processing where data are corrupted by high intensity additive or multiplicative noise, or both. The goal is to develop an adaptive computational network that incorporates the physics of many-body interactions and operational rules of biological systems. We apply BGN to restore unknown images, which have been degraded by random blurring matrix and additive Gaussian white noise. The competitive advantage of BGN derives from its nice generalization capabilities, existence of the unique attractor with the lowest energy that is worked out when several patterns are stored by the network, and fast guaranteed convergence to this attractor
Keywords :
gradient methods; image reconstruction; image restoration; many-body problems; neural nets; random noise; BGN; adaptive computational network; additive Gaussian white noise; biological systems; bistable gradient neural-like networks; fast guaranteed convergence; generalization; high-intensity additive noise; high-intensity multiplicative noise; image degradation; image reconstruction; many-body interactions; random blurring matrix; unknown image restoration; Adaptive systems; Additive noise; Biological systems; Biology computing; Computer networks; Degradation; Image processing; Image reconstruction; Image restoration; Physics computing;
Conference_Titel :
Neural Networks, 2001. Proceedings. IJCNN '01. International Joint Conference on
Conference_Location :
Washington, DC
Print_ISBN :
0-7803-7044-9
DOI :
10.1109/IJCNN.2001.938838