FPGA based image processing with R-fuctions and the curvelet transform
Abstract
In the past few years, image processing has begun to make its way into many new areas, both academic and commercial. One of the most popular areas is in computer generated animation. This includes films, video games, medical imaging, and various other multimedia systems for both entertainment and more serious applications. Two fairly recent independent developments in this field are R-functions and the curvelet transform. R-functions were developed to make it possible to represent complex objects by using a collection of simpler primitives. The curvelet transform was designed to extract specific features from complex objects. Although impressive performance can be achieved with R-functions and curvelets, the complexity of their implementation is quite a drain on standard microprocessors. It is for this reason that an FPGA implementation was developed. By offloading some of the processing work into a properly configured FPGA, speeds can be achieved in excess of one hundred times faster than current high end servers. This increase in processing speed and image representation ability combine to have some useful applications. Now, highly complex image processing can be done in small areas allowing for the design of systems that were previously not feasible to develop. By using the concepts presented in this thesis, ideas have come about for the development of a large scale Boltzman equation solver, and a satellite hyperspectral imaging system. The Boltzman equation solver has been developed before, but only by using very costly and space consuming servers. Design of the satellite hyperspectral imaging system has been hindered by the low data transmission rate of the communication system. By processing some of the data on the system itself this problem is removed. This thesis proves that R-functions and numeric transforms can be done in an FPGA to give far better performance than regular microprocessors. It also shows the power of the R-function and the curvelet and ridgelet transforms. With further development, this could yield some amazing results.
Description
Due to the character of the original source materials and the nature of batch digitization, quality control issues may be present in this document. Please report any quality issues you encounter to digital@library.tamu.edu, referencing the URI of the item.Includes bibliographical references (leaves 67-69).
Issued also on microfiche from Lange Micrographics.
Collections
Citation
Wisinger, John L. (2003). FPGA based image processing with R-fuctions and the curvelet transform. Master's thesis, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /ETD -TAMU -2003 -THESIS -W378.