Using stochastic architectures for edge detection algorithms

Edge detection plays a major role in image preprocessing for object detection. However, edge detection algorithms are computation intensive, usually complex to implement, and also sensitive to the noises in the internal circuits. Considering image processing algorithms in robotic, the power limitations of their embedded systems have encouraged designers toward stochastic computing. In this paper, we propose three novel stochastic architectures for three well-known “edge detection algorithms”. Our experimental results show that the proposed architectures require less area and also they consume less power. Furthermore, to handle the long processing time of producing accurate stochastic outputs we propose novel solutions.