Efficient Integral Image Computation on the GPU
Dept. of Electr. Eng. & Computer Science, MIT, Cambridge, MA, USA
Processing IEEE Intelligent Vehicles Symposium (IV), Publisher: IEEE, Pages: 528-533, 2010
@conference{bilgic2010efficient,
title={Efficient integral image computation on the GPU},
author={Bilgic, B. and Horn, B.K.P. and Masaki, I.},
booktitle={Intelligent Vehicles Symposium (IV), 2010 IEEE},
pages={528–533},
issn={1931-0587},
organization={IEEE}
}
We present an integral image algorithm that can run in real-time on a Graphics Processing Unit (GPU). Our system exploits the parallelisms in computation via the NIVIDA CUDA programming model, which is a software platform for solving non-graphics problems in a massively parallel high-performance fashion. This implementation makes use of the work-efficient scan algorithm that is explicated in 5. Treating the rows and the columns of the target image as independent input arrays for the scan algorithm, our method manages to expose a second level of parallelism in the problem. We compare the performance of the parallel approach running on the GPU with the sequential CPU implementation across a range of image sizes and report a speed up by a factor of 8 for a 4 megapixel input. We further investigate the impact of using packed vector type data on the performance, as well as the effect of double precision arithmetic on the GPU.
March 14, 2011 by hgpu
