Hardware/software co-design system for template matching using Particle Swarm Optimization and Pearson's Correlation Coefficient

The template matching is an important technique used in pattern recognition. It aims at finding a given pattern within a frame sequence. Pearson's Correlation Coefficient (PCC) is widely used to evaluate the similarity of two images. This coefficient is computed for each image pixel, which entails a computationally very expensive process. This paper proposes an implementation of the template matching as a software/hardware co-design system taking advantage of the Particle Swarm Optimization (PSO). In the proposed system, the PSO algorithm is implemented in software while the computation of the required PCCs is implemented in hardware via a dedicated co-processor. This approach allows for a great versatility to use this kind of implementation in portable equipment. The obtained results indicate that the proposed design is up to 13.28× faster than the software-only based implementation for the used cases. We show that the co-design implementation, with PCC computation implemented in hardware, is a viable way to achieve real-time template matching, which is a pre-requisite in real-word applications.