Title :
A neural network for deblurring an image
Author :
Jubien, Chris M. ; Jernigan, M.E.
Author_Institution :
Dept. of Electr. & Comput. Eng., Victoria Univ., BC, Canada
Abstract :
A neural network architecture for deblurring a blurry scene without prior knowledge of the blur is proposed. Two different training algorithms are described, one a standard neural network training algorithm (employing the least mean squares (LMS) rule) and the second an original algorithm, dubbed algorithm-X. Both were successful for developing inverse blur filters to enhance a blurry picture. Algorithm-X is computationally less complex than the LMS algorithm, and in tests comparing the training times of the two algorithms, algorithm-X was found to be faster
Keywords :
filtering and prediction theory; least squares approximations; neural nets; picture processing; LMS algorithm; algorithm-X; blurry scene; image deblurring; inverse blur filters; least mean squares; neural network architecture; training algorithms; training times; Degradation; Digital filters; Filtering; Image processing; Layout; Least squares approximation; Neural networks; Neurons; Nonlinear filters; Vectors;
Conference_Titel :
Communications, Computers and Signal Processing, 1991., IEEE Pacific Rim Conference on
Conference_Location :
Victoria, BC
Print_ISBN :
0-87942-638-1
DOI :
10.1109/PACRIM.1991.160776
