Experiments on Global and Local Adaptation to Illumination Conditions based on Focal-Plane Average Compuutation

This paper presents some experiments with an integrating pixel with adjustable gain for adaptive image capture. The adaptation mechanism consists in regulating the exposure time according to both global and local illumination conditions. Local adaptation is based in the concurrent processing of local information in the present and previous frames. Global adaptation corrects the average integration time towards the optimum starting on the average pixel value in the previous frame. Experimental results from a prototype chip, which contains a 16times16 array of adaptive pixels, are presented. Tests results have shown that global adaptation enlarges the interframe dynamic range up to 4 decades (80dB). This image DR is limited by test setup, as part of the control scheme is implemented off-chip. Meanwhile, local adaptation produces some histogram equalization and reaches the theoretical intraframe DR enhancement of 6dB. Both adaptive mechanisms have been incorporated into an analog parallel array processor based on a multilayer CNN for image capture based on focal-plane processing