Hardware Image Filtering on Desktop Computers

Real-time video processing, large scale image manipulation and many other fields require much processing time. Severity of this problem can be abated in different ways. One solution is to use what is already present in a desktop personal computer as general purpose graphics hardware gets more programmable. This paper gives an overview of author's experiences and work in this field. Test results are promising: hardware filters are up to 32 times faster in perfect conditions and at least 3 times faster in common test. The following filters are tested: Gaussian and other smoothing techniques, sharpen, Roberts, Sobel, Prewitt, median, morphological corner extractor.