This report describes a circuit architecture and CMOS circuit components implementing a cellular neural network (CNN) which filters an input image with two convolution kernels similar to odd and even phase Gabor filters. The Gabor filter is a preprocessing stage used in several different trypes of computer vision and image processing algorithms. One of its primary drawbacks is that it is computationally intensive on a digital computer. Implemented in analog VLSI using the circuits described here, a CNN could decrease both the time and the power required to perform the filtering.
展开▼