An Analysis of the Implementation of Edge Detection Operators in FPGA
Douglas Santos,Daniel Zolett,Mateus Belli,Felipe Viel,Cesar Zeferino
DOI: https://doi.org/10.5753/sbesc_estendido.2020.13107
2020-11-23
Abstract:Computer vision systems have several stages, and one of the operators used in these systems is the edge detection filter. High-performance computing is required in many applications and stages of computer vision systems, and many designs use FPGA technology to improve performance and decrease power consumption. In this context, this work presents an analysis of five edge detection filters synthesized to FPGA, including Laplacian, Roberts, Prewitt, Sobel, and Canny. In the experiments, we compared the hardware implementations with software versions to identify the impact of fixed-point representation on the quality of the output images. We have also assessed metrics regarding performance, silicon costs, and energy consumption. The results obtained show that the Laplacian filter has the lowest costs, while the Canny operator provides the best output image at the price of much higher silicon costs and energy consumption.