Error-Tolerant Image Processing Application Based on Stochastic Logic

Aggressive scaling of semiconductor devices exacerbates issues of reliability. Stochastic logic, which operates on probabilistic signals, takes advantage of randomness resulting in a more reliable system with simple hardware. In this paper, we present our stochastic design of low-pass (smoothing) and highpass (sharpening) filters. We modeled our design in Verilog and synthesized the design. The synthesis results show that stochastic logic can have operate at a faster clock frequency and consume less silicon area in specific applications compared to the conventional implementation. In addition, the stochastic implementations are much more tolerant of soft errors. Keywords—Stochastic Logic; Error-Tolerant Computation; Image Processing; Low-pass Filter; High-pass Filter