Parallelization of the Mixture of Gaussians Model for Motion Detection on the GPU

Abstract

Motion detection and object tracking have many applications in various domains. The process of motion detection depends on the detailed analysis of pixels from successive frames in the given video scene. Some background subtraction techniques are commonly used for this purpose. Nowadays, even the consumer electronic devices, like cell phones, can produce high definition videos with their cameras. Efficient, real-time analysis of those videos can be performed using modern graphics processing units. In this paper, we present a GPU implementation of the mixture of Gaussians model for background subtraction. We observed speedups up to 6 times over the reference sequential implementation.