A Numerically Approximate High-speed Divider for Image Processing Applications

Integer division by nine is a very common operation in image processing applications. The implementation of such division operations is normally performed through standard modules, which are not optimized for this function. In this study, we propose an approximation algorithm that can be used with any pre-existing architecture such as LUT to achieve a speed up. However this is achieved at a cost in accuracy, but this is shown to be acceptable for image processing applications. The performance of the proposed approximation is analyzed by using it in the spatial average filtering of an image corrupted by gaussian noise and by comparing the filtered image with that in which accurate division by nine is used.