An acceleration of image processing is essential technology for commercialize of software. In particular, we study about an acceleration of Particle Filter algorithm for object tracking. Generally, Number of particle in parameter space is secured enough, a tracking accuracy improves. Therefore, a computational complexity is proportional to number of particle and its observation and prediction. So fast processing procedures of these calculations is needed. In this paper, we apply a parallelization of observation and prediction by using CUDA framework. In some experiments, observation and prediction by CUDA could show fast processing time than CPU based (utilizing vector operation and multi-core operation) algorithm (maximum is 3.1 times faster).

View full abstract

Sliding window approach used in conventional pedestrian detection samples huge number of sub-windows from an input image to extract features required for classification. In this paper, we propose computationally efficient pedestrian detection based on Markov Chain Monte Carlo and GPU implementation to achieve further speed-up of the proposed method. The results of preliminary experiment using software implementation show the validity of the proposed method itself. Also, the results of GPU implementation using Tesla C1060 show that the proposed method is accelerated 25.7 to 76.1 times faster than the software implementation using Intel Core i7 CPU 975 3.33GHz.

View full abstract