Search

Using the concept of a transistor as a pseudoresistor, many resistive networks can be mapped to transistor based equivalents. Pseudoresistors are two-terminal circuit elements whose current is a function of the instantaneons difference in the pseudovoltage across the two terminals, where the pseudovoltage is a nonlinear function of the voltage at each terminal. A transistor with fixed gate and bulk voltages is a linear pseudoresistor between the drain and source over a wide range of currents including strong and weak inversion. Motivated by this, we extend the cocontent, defined for two-terminal resistors, to the pseudovoltage-based (PVB) cocontent, defined for two terminal pseudoresistors. We show that the PVB cocontent has similar properties as the cocontent in characterizing the dc-operating points of transistor networks and their stability, but under more restrictive conditions. As examples, we derive the total PVB cocontent for several neuromorphic circuit networks including the diffuser grid, a linear pseudoresistive network, and the winner-take-all circuit of Lazzaro, a nonlinear pseudoresistive network. This work establishes that these circuits can be viewed in terms of function minimization, which may lead to better intuition about their operation as well as the development of new circuit architectures. Index Terms--Circuit analysis, MOS analog integrated circuits, neuromorphic, nonlinear circuits, resistive circuits.

We extend previous work in orientation selective cellular neural networks to include competitive couplings between different layers tuned to different orientations and spatial frequencies. The presence of these interactions sharpens the spatial frequency tuning of the filters in two ways, when compared to a similar architecture proposed previously which lacks these interactions. The first is the introduction of nulls in the frequency response. The second is the introduction of constraints on the passbands of the coupled layers. Based on an understanding of these two effects, we propose a method for choosing spatial frequency tunings of the individual layers to enhance orientation selectivity in the coupled system. Index Terms--Cellular neural networks, competition, Gabor filters, image processing, orientation-selective filters.

We describe the design, implementation, and test results from an analog complementary metal-oxide-semiconductor very large scale integration (CMOS VLSI) chip which combines photosensing and an image filtering operation similar to Gabor filtering. The chip contains a one-dimensional (1-D) 27-pixel array and generates three outputs at every pixel; the input image intensity and the real and imaginary parts of the filter output. These results validate experimentally a recently presented theoretical framework for the design of cellular neural network (CNN) templates for Gabor-type spatial image filtering. Index Terms - Analog circuits, cellular neural network (CNN), Gabor filter, image processing, neural networks.

Gabor filters are preprocessing stages in image-processing and computer-vision applications. One drawback is that they are computationally intensive on a digital computer. This paper describes the design of cellular neural networks (CNN's) which compute the outputs of filters similar to Gabor filters. Analog VLSI implementations of these CNN's might eventually relieve the computational bottleneck associated with Gabor filtering image-processing algorithms. The CNN's compute both the real and imaginary parts of the filter outputs simultaneously, which is an important feature in applying them in algorithms utilizing the phase of the Gabor output. Index Terms - Analog circuits, cellular neural networks, filtering, image processing, neural networks.

Recently, several researchers have proposed resistive grids as primary components in analog VLSI circuit implementations of various image processing algorithms. In this paper, we use the Cellular Neural Network framework developed by Chua and Yang to analyze the image filtering operation performed by the linear resistive grid. In particular, we show first in detail how the resistive grid can be cast as a CNN and discuss the use of frequency-domain techniques to characterize the input/output behavior of resistive grids of both infinite and finite size. These results lead to a theoretical justification of one of the 'folk theorems' commonly held by researchers using resistive grids: resistive grids are robust in the presence of variations in the values of the resistors. To the best of our knowledge, this conjecture has only been verified by simulation. Finally, we suggest an application to edge detection. In particular, we show that the filtering performed by the grid is similar to the exponential filter in the edge detection algorithm proposed by Shen and Castan.