Super-Resolution Reconstruction For Lung 4D-CT Based On Multi-Phase Similarity And Global Graph Cut

Four dimensional computer tomography(4D-CT)has been widely used in today's cancer therapy because of its abilities of comprehensive image guidance for radiotherapy.4D-CT by clearly reflecting lung target organ while reducing the motion artifact caused by respiratory movement.Despite of the potential advantages of 4D-CT,there still some disadvantages remain,such as long scanning time,high radiation dose and large slice thickness which make the 4D-CT data we collected have a lower resolution(LR)along the Z axial direction than the in-plane resolution.So if you want to obtain normal resolution lung CT images,interpolation along Z axial direction in certain proportion must be implemented.Various interpolation methods have been proposed,such as linear interpolation and cubic spline interpolation,but most of these approaches may result in image with blurred edges and details.In this paper,we want to improve lung 4D-CT data multiple planar display quality,we do some research on super resolution(SR)and reconstruct high-resolution(HR)lung 4D-CT image.SR techniques aim at producing a HR image from a set of low-resolution images that contain slightly different spatial information or a single LR image.The common SR methods are usually based on frequency space and spatial space.But the method based on frequency space is not easy to expand,it not easy to add constrain,too.So the current research mainly focuses on the super-resolution reconstruction method based on spatial research.According to lung 4D-CT image characteristics,this paper studies two lung 4D-CT image SR reconstruction method based on spatial space:multi-phase similarity based SR reconstruction method and global graph cut based SR reconstruction method.And reconstruct a better result than tradition interpolation method and POCS algorithm both in visually and quantitively.The first two chapters of this article introduce the research significance and the relevant technology status respectively.In chapter three and four,introduce the multi-phase similarity based SR reconstruction method and global graph cut based SR reconstruction method in detail respectively.In the third chapter,there expounds the nonlocal means filter,the block search based on multi-phase similarity and global constraint reconstruction.In the fourth chapter,there expounds the global graph building,energy function building and optimal solution.In the fifth chapter,there makes a summary and prospects of the full paper.