This thesis is an investigation of methods to increase the performance of image processing techniques, while maintaining resultant image quality, through the use of programmable graphics hardware as applied to the Prokudin-Gorskii image set. To start the problems with the image set will be addressed. Following this we look at current methods, both manual and automatic, for the restoration of the Prokudin-Gorskii image set. The investigation then turns to methods for parallelization of image registration and image enhancement as they apply to the Prokudin-Gorskii set. The results of the work indicate that some image processing tasks are well suited for parallelization, such as image alignment, whereas others, such as histogram equalization, offer a performance penalty. |