Efficient integral image computation on the GPU

• Published in: 2010 IEEE Intelligent Vehicles Symposium pp. 528 - 533
• Main Authors: Bilgic, Berkin, Horn, Berthold K.P., Masaki, Ichiro
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.06.2010
• Subjects: Central Processing Unit; Concurrent computing; Detectors; Graphics processing unit; Parallel processing; Parallel programming; Signal processing algorithms
• ISBN: 1424478669; 9781424478668
• ISSN: 1931-0587; 2642-7214
• DOI: 10.1109/IVS.2010.5548142

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. 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.