A New Lightweight and High Fault Tolerance Sobel Edge Detection Using Stochastic Computing Ming Ming Wong 1* , Dennis M. L. D. Wong 2 , Cishen Zhang 3 , Ismat Hijazin 3 1 Faculty of Engineering, Computing and Science, Swinburne University of Technology Sarawak, Malaysia. 2 Heriot Watt University Malaysia, Wilayah Persekutuan Putrajaya, Malaysia. 3 Faculty of Science, Engineering and Technology, Swinburne University of Technology, Hawthorn, Australia. * Corresponding author. Email: wmingming7@gmail.com Manuscript submitted March 24, 2017; accepted June 23, 2017. doi: Abstract: A new Stochastic Computing (SC) circuit design paradigm for image processing system is presented in this work. Two improved SC computational functions are derived, which are namely the stochastic scaled addition and stochastic absolute value of difference. Data correlation is also incorporated in the design for effective circuit size reduction without imposing accuracy degradation in the hardware implementations. The proposed SC functions are next employed to design the new and lightweight Sobel edge detection. Experimental results obtained from detailed test analysis have proven that new implementation has satisfactory accuracy level and higher fault tolerance capability in comparison with their conventional counterparts. The works proposed are also implemented on an Altera Cyclone V 5CGXFC7D6F31C6 FPGA for hardware complexity evaluation. Key words: Stochastic computing, scaled addition, absolute value of difference, Sobel edge detection, fault tolerance. 1. Introduction Stochastic computing (SC) [1] was first introduced several decades ago and it has recently gained a fair amount of attentions in the integrated circuit (IC) design community. This technique, which incorporates elements from probability theory, has proven to be able to handle computation uncertainties in a more effective and efficient manner [2]. Thus, it emerges as an unconventional and non-deterministic computational technique with high fault tolerance [3], [4]. Furthermore, SC is particularly attractive in IC design as it requires low complexity computation blocks. In general, SC is seen as a promising alternative in comparison to its conventional binary computing counterpart, which usually has a higher computational cost. Though SC has been known for decades, very few physical realizations have been proposed. Initially, SC applications were limited to the field of neural networks [5] and machine controls [6]. Until recent years, it was discovered that SC efficiently simplifies some mathematical functions which are computational expensive in binary computation. These functions can be efficiently approximated using stochastic logic with minimal hardware requirements and without significant accuracy degradation. Ever since, SC implementation has been extended to image processing [7]-[9], error control coding applications [10] and digital filter design [11]-[14]. The main contributions of this work are three-folds. First, we presented the improved designs for 403 Volume 9, Number 2, December 2017 International Journal of Computer Electrical Engineering 10.17706/ijcee.2017.9.2.403-420