A CMOS current-mode VLSI implementation of cellular neural network for an image objects area estimation

Summary form only given. The paper is concerned with the physical implementation of a CNN which realizes a specific predefined image processing operation. The circuit is intended to be realized in a 1.2 μm CMOS process. The VLSI chip is expected to be used as a smart sensor of visual information and its main function is to detect objects with an area which exceeds some user-defined threshold value. For proper operation the circuit should provide an appropriate image preprocessing to cancel the effects of non-uniform image illumination and to suppress image noise. The kernel of the proposed circuit is a reprogrammable CNN, based on the concept of a pulse-mode CNN. The paper presents an architecture of a basic element of the circuit-a reprogrammable cell, designed using the current-mode approach. The proposed cell architecture differs from the solutions presented so far since it is optimized to ensure these processing properties, which are important for proper circuit operation