Efficient implementation of video processing algorithms on FPGA
The work contained in this portfolio thesis was carried out as part of an Engineering Doctorate (Eng.D) programme from the Institute for System Level Integration. The work was sponsored by Thales Optronics, and focuses on issues surrounding the implementation of video processing algorithms on field programmable gate arrays (FPGA). A description is given of FPGA technology and the currently dominant methods of designing and verifying firmware. The problems of translating a description of behaviour into one of structure are discussed, and some of the latest methodologies for tackling this problem are introduced. A number of algorithms are then looked at, including methods of contrast enhancement, deconvolution, and image fusion. Algorithms are characterised according to the nature of their execution flow, and this is used as justification for some of the design choices that are made. An efficient method of performing large two-dimensional convolutions is also described. The portfolio also contains a discussion of an FPGA implementation of a PID control algorithm, an overview of FPGA dynamic reconfigurability, and the development of a demonstration platform for rapid deployment of video processing algorithms in FPGA hardware.