The rapid growth of video processing techniques has led to remarkable contributions in several applications such as compression, filtering, segmentation and object tracking. A fundamental task of video surveillance cameras is to detect and capture major moving objects (foreground). Processing video frame by frame is complex and difficult for real time applications. GPUs have led to significant advancements in the field of image/video processing especially in real time applications. In this work, we make use of the parallel computing capacity of GPUs to speed up the runtime of foreground detection algorithm. The focus of the thesis is to accelerate the runtime of the algorithm by parallelizing the time consuming portions. The final goal would then be to analyze and come up with the optimal parallelization technique(s) that give(s) the best performance.
The rapid growth of video processing techniques has led to remarkable contributions in several applications such as compression, filtering, segmentation and object tracking. A fundamental task of video surveillance cameras is to detect and capture major moving objects (foreground).
Processing video frame by frame is complex and difficult for real time applications. GPUs have led to significant advancements in the field of image/video processing especially in real time applications. In this work, we make use of the parallel computing capacity of GPUs to speed up the runtime of foreground detection algorithm.
The focus of the thesis is to accelerate the runtime of the algorithm by parallelizing the time consuming portions. The final goal would then be to analyze and come up with the optimal parallelization technique(s) that give(s) the best performance.
