Median and morphological filtering of images in real time using an FPGA-based custom computing platform

Abstract

This thesis describes the design and implementation of real-time image-processing tasks on a custom computing platform called Splash-2. The tasks that have been implemented are image histogram generation, median filtering using 3X3 neighborhoods, and morphological dilation and erosion. These problems are computationally intensive, involving large amounts of data. The problems are especially difficult when the images need to be processed at real-time rates, typically 30 frames per second. Splash-2 is a reconfigurable FFPGA-based attached processor featuring several programmable processing elements and programmable communication paths. Although not designed specifically for image-processing applications, it possesses architectural properties that make it well suited for high speed computations and data transfer rates that are characteristic of this class of problems. This thesis discusses the design process which has been used to map the tasks to Splash-2, and presents results which demonstrate the effectiveness of custom computing platforms for high performance image processing.