Optical Flow Estimation Based Super-Resolution Reconstruction For Lung 4D-CT Image

Four-dimensional computed tomography(4D-CT)imaging technique obtains different phases of CT images during the whole respiratory cycle by scanning human tissues in normal breathing state several times.The final 4D-CT image is obtained by classifying and sorting the raw CT data obtained at Scanning stage.4D-CT image not only has spatial resolution ability,but also has time resolution ability,which can reflect the location of chest and abdomen organs well.The shape and volume of lung cancer are widely used in the diagnosis and treatment of lung cancer due to the movement of respiration.However,due to the high dose characteristics of CT images,the clinical practice can only be done by reducing the sampling accuracy and scanning time.In order to reduce the amount of radiation,the inter-slice resolution of 4D-CT image is smaller than that of inter-slice resolution,which results in the pulmonary crown reconstructed from clinical 4D-CT data,and the sagittal image does not match the true ratio of lung.In recent years,in order to get a normal ratio of lung images,4D-CT image super resolution algorithm based on reconstruction has been widely studied.This algorithm regards different phase multi-plane(coronal,sagittal)images obtained from 4D-CT data as a series of low-resolution images in the same scene.The redundant information between different phase images is fused well by reconstruction algorithms,which improves the image quality of coronal and sagittal planes.And how to retain as much high-frequency detail information as possible,in the reconstruction process,is also an urgent problem.In this paper,a variational optical flow model based on local and global combination is first used to obtain more accurate sub-pixel displacement between different phase images,and the fast solution of optical flow field model is realized by applying alternating direction method of multipliers algorithm.Based on the displacement information of different phase images,two algorithms are proposed to improve the image quality of lung 4D-CT multi-planar display:the improved non-local iterative back-projection algorithm and the maximum a posterior estimation based super-resolution reconstruction for lung 4D-CT image.We present an improved non-local iterative back-projection algorithm to improve the quality of coronal or sagittal images of lung.First,one image is selected,as the reference image,from the input low-resolution image sequence to get the high-resolution initial estimation image by applying traditional bi-cubic interpolation algorithm and non-local filtering operation.Then,the displacement information between the low resolution imagesequences is obtained using the local and global variational optical flow estimation model.Finally,the error image which is filtered by the non-local mean value is inversely projected to the high-resolution estimation image.The algorithm makes use of complementary information between different phase images to preserve and enhance the details of the image.The experimental results show that,compared with the existing algorithms,the super-resolution method proposed in this paper not only enhances the texture of the image,but also preserves the contour of the image.In the maximum a posteriori estimation based lung 4D-CT image super-resolution reconstruction algorithm,firstly,based on the inherent self-similarity of medical images,we use non-local low-rank priori to model high-resolution images.In order to weaken the influence of abnormal noise on noise estimation,this algorithm uses likelihood exponential distribution to model the noise term.Finally,the alternating direction multiplier method is used to solve the model.The experimental results show that the reconstructed image is clearer.The comparison of image average gradient and other indexes can also prove that the algorithm can improve the image quality.