GPU-accelerated computation for robust motion tracking using the CUDA framework

Abstract

In this paper, we discuss our implementation of a graphics hardware acceleration of the Vector Coherence Mapping vision processing algorithm. Using this algorithm as our test case, we discuss our optimization strategy for various vision processing operations using NVIDIA's new CUDA programming framework. We also demonstrate how flexibly and readily vision processing algorithms can be mapped onto massively parallelized GPU architecture. Our results and analysis show the GPU implementation exhibits a performance gain of more than 40-fold of speedup over state-of-art CPU implementation of VCM algorithm.

A Real-Time Motion-Feature-Extraction VLSI Employing Digital-Pixel-Sensor-Based Parallel Architecture

“…The traditional software-based MFE algorithms are computationally very expensive, and building high-speed systems is quiet difficult even when state-of-the-art multicore generalpurpose processors are utilized [8], [9]. By developing finetuned software exploring the single-instruction multiple-data operations on modern processors and/or graphics processing units for particular applications [10], [11], high hardware costs and large power consumptions are severely limiting such approaches to be used in portable devices, as well as in large-scale systems. As a result, high-speed and low-power architectures for MFE are now highly demanded.…”