Research On Super-resolution Reconstruction Based On Image Sequence

In terms of most system related with computer vision, whose function depends on the quality of the image to a great extent. However, in the process of imaging, imaging system generally cannot obtain so many informations in the original scene that the quality of theimage will be degenerated, which is limited with many factors such as imaging conditions, imaging approach and so on. Therefore the purpose of image reconstruction is to make the image be closed to theideal picture,which has not deteriorated,by the way of dealing with the degenerated image. Super-resolution image reconstruction makes use of the relatively complementary information between quantities of low resolution imagesin the same scene tomerged theminto the high resolution images,that is to say,torecorvery the high frequency informationthat is missed during the process of imaging, so as to accomplish the item that the quality of image which is higher than the resolution of imaging system will be obtained without increasing the hardware of the system at the same time.This article make a research be concerned with super-resolution image reconstruction which is based on image sequence,in the foundation of POCS,and having intelligent video safety monitoring as a background,the main contents about this article areas follows(1) For researching the method of image interpolation, the image edge will be fuzzied by the linear interpolation, so this article puts forward the interpolation method based on edge keep. Local adaptive interpolation method can remain the image edge, but whose edge direction in the consideration is not comprehensive. Based on this point, a new method by double channel local adaptive interpolation is raised to improve the image interpolation effect to some degree.(2) For block matching method in the motion estimation, two key factors, the matching criterion and the search strategy are analyzed, according to the shortage of traditional matching criterion, this article puts SAD into use to improve the constraint conditions the accuracy of block matching estimation. Combines three search with diamond search, presents the strategy with the method of three diamond search to improve the speed and accuracy of block match, in terms of searching strategies at the same time.(3) For super-resolution reconstruction based on convex setpojection algorithm are studied, including thekey factors concerning with realizing image reconstruction is the structure of the point spread function.In tradition,PSF will be supposed to Gaussian function,which will cause the phenomenon of edge concussion.AlthoughThe method can keep the edges of image up,the influence of the noise is ignored,by dealing image with bilateral filter.Above all these reasons,the idea called false median filtering,which improve the effect on the image reconstruction.