FLIP-Q: A QCIF Resolution Focal-Plane Array for Low-Power Image Processing

Abstract

The vision processing chain is usually divided into three consecutive steps (González andWoods 2002): (i) low-level tasks, where both inputs and outputs are images, (ii) medium-level tasks, where inputs are images but outputs are attributes extracted from inputs and (iii) high-level tasks, which perform the cognitive functions associated to vision from the result of low- and medium-level tasks. The most remarkable features of low-level tasks are, as we saw in Chap. 2, their regularity and potential parallelism. All pixels are equally processed and the result of the processing over each pixel is usually independent from the result of the computations over the rest.