An OpenCL-based SIFT Accelerator for Image Features Extraction on FPGA in Mobile Edge Computing Environment

Abstract

Mobile Edge Computing (MEC) has become a promising technology for future cloud computing. MEC enables low-latency service by extending the computation capability to the edge of the network; therefore, hardware accelerators are an essential part of MEC server. As graphics processing unit (GPU) is high-energy consumption, field programmable gate array (FPGA) can be an alternative to accelerate services in power-limited environment as MEC. In this paper, we present an OpenCL-based SIFT accelerator for image features extraction on FPGA that can be deployed as a service on MEC environment. The experimental result on an image with a size of 1024 × 1024 shows that our accelerator speeds-up the bottleneck of the SIFT algorithm up to 13.7 times compared to software version and the energy efficiency is 1.38 times better than the GPU accelerator on an high-end NVIDIA GPU.