high performance computing on graphics processing units: hgpu.org

The Discrete Wavelet Transform (DWT) is applied to various signal and image processing applications. However the computation is computational expense. Therefore plenty of approaches have been proposed to accelerate the computation. Graphics processing units (GPUs) can be used as stream processor to speed up the calculation of the DWT. In this paper, we present a implementation of the translation-invariant wavelet transform using consumer level graphics hardware. As our approach was motivated by infrared image processing our implementation focuses on gray-level images, but can be also used in color image processing applications. Our experiments show, that the computation performance of the DWT could be significantly improved. However, initialisation and data transfer times are still a problem of GPU implementations. They could dramatically reduce the achievable performance, if they cannot be hided by the application. This effect was also observed integrating our implementation in wavelet-based edge detection and wavelet denoising.