Search

A study was conducted to show that intensity mappings can be realized by either variable sampling time or by time variable threshold. The discrete mapping was represented by the time sampling set supporting a fixed threshold level. Experimental results indicated that adaptive mapping correlated very well with the near-sensor image processing concept wherein the quantization of light intensity was inversely proportional to the intensity.

The near-sensor image processing (NSIP) approach and the 2-D global logic unit can be implemented in a single chip, based on the VSLI technology. NSIP is a novel concept applicable to low-level image processing. Powerful segmentation- and feature-extracting operations can be performed by using the regular net in the 2-D global logic unit. The segmentation operations shown by the net are hole filling, thresholding with hysteresis, smallest circumscribing rectangle and arbitrary propagation patterns.

Low-level image processing such as convolution and median filtering are performed by thresholding amplifiers that are combined with photodiodes that exhibit analog-temporal characteristics. The reference voltage and time are used as the controlling variables. Hybrid analog/digital circuits applicable for implementation of an integrated component constitutes the image processing system which exhibits analog-temporal characteristics. This image processing system also has high speed in performance.