A hardware architecture design for real-time Gaussian filter

Abstract

The Gaussian Filter is one of the processing methods to smoothen images. Before detecting an object, this method is used as pre-processing, for eliminating useless detail, and connecting unnatural parts. We implemented a Gaussian Filter on FPGA, as that is the pre-processing system of the real time stereo vision system in this paper. The hardware implementation method saves the image in the Line Buffer for parallel processing of calculation, and sends the saved value to the Window Buffer, and then the result value of the passed Gaussian Filter is derived, through calculating the Gaussian mask and saved Window Buffer. This system receives the 640×480 image (25.4545MHz) from the CIS VCC-8350L camera, and transmits the blurred image to a monitor in real time (60fps).