Gabor-Type Filtering using Transient States of Cellular Neural Networks

Abstract Gabor filtering is useful for intelligent image processing, but it requires huge computational power. Its pixel-pazallel LSI implementation is one solution for real-time image processing. This paper proposes a new Gabor filtering algorithm using adiscrete-time cellular neural network (CNN) circuit, which is suitable for pixel-parallel LSI implementation. The proposed algorithm utilizes transient states of the CNN to obtain Gabor coefficients, and it has the following advantages: (1) the amplitudes of all coefficients of the terms in the dynamics equations are on the same order; (2) the relative amplitudes of Gabor coefficients can arbitrarily be controlled; and (3) the number of calculation steps required for obtaining Gabor coefficients can be reduced compared with the algorithm using the steady state; (4) the window function of the filter is nearly Gaussian.

[1]  Bertram E. Shi 2D focal plane steerable and scalable cortical filters [image sensor] , 1999, Proceedings of the Seventh International Conference on Microelectronics for Neural, Fuzzy and Bio-Inspired Systems.

[2]  Bertram E. Shi,et al.  Gabor-type filtering in space and time with cellular neural networks , 1998 .

[3]  Joachim M. Buhmann,et al.  Distortion Invariant Object Recognition in the Dynamic Link Architecture , 1993, IEEE Trans. Computers.