基于多尺度几何分析的图像超分辨率重建技术研究-理学专业论文.docxVIP

摘要 摘要 关键词：图像超分辨率重建；轮廓波变换；曲波变换；图像配准；图像融合 本研究得到以下基金资助 国家自然科学基金资助项目（No； 陕西省教育厅科学研究项目（No.11JK0900）。 II Abst Abstract Title: RESEARCH ON MULTI-SCALE GEOMETRIC ANALYSIS BASED IMAGE SUPER-RESOLUTION RECONSTRUCTION TECHNOLOGE Major: Applied Mathematics Name: Yang ZHOU Signature: Supervisor: Prof. Xiaofeng WANG Signature: Abstract In the process of image acquisition, there are many factors such as optical distortion, aliasing, motion blur and noise, to result in the decrease of image resolution. The requirement for high resolution images promotes the continuous development of image super-resolution (SR) reconstruction technology. In this dissertation, we aim at degraded images, use the method of multi-scale geometric analysis, and study image super-resolution reconstruction technology. The main works are as follows: We studied the corresponding relationship between observation model and degenerate model of image, and improved an image super-resolution reconstruction method based on Contourlet transform. Our method utilizes the advantage that Contourlet transform can capture the contour information and edge information of images. In improved method, SIFT feature point extraction and matching technology were used for image registration, and Contourlet transform was used to the redistricted image. Then we used the weighted average method to the low frequency part that come from the Contourlet transform; About the high frequency part that come from Contourlet transform, we comparing the variances of the relevant high frequency coefficient on the same direction and the same scale, and let the coefficient that have the smallest variance as the high frequency coefficient of inverse Contourlet transformation. Then we take an inverse Contourlet transformation by using the new low frequency coefficients and the new high frequency coefficient selected. Next, we use Bicubic interpolation and Bilateral Filtering to the image that obtained last step. Finally, a new higher resol