Research On Technology Of Computed Tomography Based On GPU

Computed tomography (CT) is a Nondestructive Detecting technology that basis outer projection data to reconstruct the interior picture of an object. At first it applied to medical detecting,with the fast development of the microcomputer and image processing systems, the medical used CT formed an industry during very short time. The fast development of the medical used CT has received much attention from the industry field. At present, the industrial CT develop rather fast, as a Nondestructive Detecting technology, it has been widely applied in fields such as agriculture petroleum, mechanics, minerals, automobile, aeronautics, astronautics and military equipment etc.The theory of CT image reconstruction was introduced and ART, FBP, FDK were introduced in this thesis. FBP and FDK were very popular in this field. The design of filter function is very important in the CT reconstruction. In order to improve the image quality, a new method of filter function design was presented based on the FBP theory, a new filter model was built and a new filter was design based on this filter model. Based on the simulated data of Shepp-Logan and the projection data of conch, the compared result of CT reconstruction using R-L filter function, S-L filter function and a new filter function based on the new design method was given. The reconstruction result of the new design filter function was the best in overall performance. In local areas, the density resolution had increased. So this paper provided a novel method for filter function designing in CT reconstruction. The simulated software of CT reconstruction based on MATLAB was design and the program based on C++ was given in this thesis.The amount of data needs to be processed in CT image reconstruction, so the slow speed of CT image reconstruction is unbearable in some situation. The parallel processing based on GPU is a great option to accelerate the CT image reconstruction. Compute Unified Device Architecture (CUDA) which was introduced by NIVIDIA company in November 2006 is a general purpose parallel computing architecture with a new parallel programming model, and instruction set architecture that leverages the parallel compute engine in NVIDIA GPU to solve many complex scientific computational problems in more efficient way than on CPU. A parallel processing based CUDA for CT image FDK reconstruction was proposed in this thesis. According to experiment results, under the condition that the precision was assured, the speed of CT image reconstruction has been accelerated about 11 times in the overall performance. Also a the processing software based on LabVIEW was given. This software could show the DR figure in different perspectives and the different section of sample. Three-dimensional display technology was discussed in this thesis.At last, according to learning from the success of the CT industry, the possibility to apply the Nondestructive Detecting technology——CT to the plants roots detecting was discussed in this thesis. The experimental platform has been designed and a series prophase experiments has been finished.