Research On Parallel Algorithms Of Image Matching

The image matching technologies get more and more extensive applications in the field of modern information processing, so the researches of image matching algorithm are not only widespread, but more deep. on basis of maintaining the matching accuracy, how to improve the matching speed is the focus of the matching algorithm research. with the popularity of multi-core computers and the rapid development of parallel computing technologies, designing parallel algorithms of image matching on multi-core platforms to improve matching speed has important practical value.In this thesis, the main work:(1) Review the development of image matching and parallel computing technologies, and introduce the methods of parallel computing based on Matlab.(2) To design a parallel image matching algorithm based on hausdorff distance. Firstly to introduce image matching method based on Hausdorff distance, after analyzing the bottleneck of algorithm efficiency, to design and achieve parallel algorithm based on Hausdorff distance using Matlab language, finally apply it to the grayscale image to show its performances in faces of matching speed, distortion and noise. The experimental results show that this parallel algorithm not only can significantly improve matching speed on multi-core platforms, but also resist to distortion and noise in a certain.(3) To design a parallel image matching algorithm based on SSDA . Firstly to introduce image matching method based on SSDA, after analyzing the bottleneck of algorithm efficiency, to design and achieve parallel algorithm based on SSDA using Matlab language, finally to verify the correctness and validity of the algorithm apply it to grayscale images. Experimental results show that the parallel algorithm designed can significantly improve the matching speed on multi-core platforms.The designing thinking of the parallel algorithm in this thesis is universal,can be extended to other design of image matching parallel algorithm.