Study On Image Pre-processing Of TDI-fMOST

The brain research is one of important problems in all life science subjects,involving neural information,high performance computing,medical informatics and other fields.TDI-fMOST,Fluorescence micro-optical sectioning tomography based on time delay integration camera,combined with the chemical sectioning property of fluorescent protein can realize high-throughput and high-precision imaging of biological tissue,providing a wealth of resources for the brain research.While,the raw images generated by the TDI-fMOST have some problems which may lead to the failure of subsequent analysis about the brain.In view of the seriousness of the image quality degradation and dual-channel image registration problems,specific algorithms have been studied to make the necessary preparation for the quantification research of the data set.Using background extraction technology and seamless mosaic technology in the case of non-overlapping regions between images to homogenize the background within single frame image;Using the phase correlation method and mutual information to obtain parameters for dual-channel image registration precisely and efficiently.The algorithm can overcome the big data problem and high variation problem of dual-channel image,which conventional methods are difficult to solve;On the basis of the above,multi-thread parallel technology is implemented to accelerate the algorithm.The pre-processing algorithms presented in this paper have successfully solved problems mentioned above.The resolution of uneven background problem on single frame image and seam mosaic process improve the quality of TDI-fMOST and reduce error probability of quantitative analysis;The algorithm for dual-channel image registration have simplified registration operation and ensured the accuracy of registration.The preprocessing speed of the system data set is accelerated,which saved time cost.A good result has been achieved by practical application of pre-processing algorithms of TDI-fMOST.