Study On Full-color Three-dimensional Optical Sectioning Imaging Based On Structured Illumination

With the continuous advancement and development of science and technology,various optical methods have been successfully applied to three-dimensional(3D) imaging in the fields of biomedicine,materials science,paleontology,etc.,which provide the most direct scientific proofs for researchers.As a wide-field microscopy,structured illumination microscopy(SIM)is developed during the past two decades,which can realize both 3D optical sectioning imaging and super-resolution imaging.With the advantages of high imaging speed,high spatial resolution,low photoxicity and the superior 3D capabilities,SIM has become the most suitable tool for super-resolution dynamic imaging of living biological tissues and true three-dimensional rapid imaging in structural observation,and has received extensive attention from many scientific research institutions at home and abroad.This thesis focuses on three-dimensional microscopic imaging based on SIM,we designed and developed a full-color SIM system.Related researches were carried out in full-color optical sectioning algorithm,high resolution 3D imaging and its digitalization with large field of view.The majority of the researches is outlined below:1.Through in-depth analysis of the traditional SIM system,a compact SIM system which has the capability of 3D high-resolution color imaging of centimeterscale samples was designed and built.The system overcomes the shortcomings of the existing system,such as small field of view,low utilization rate of light energy,complicated of structure and inability to perform color imaging.The actual size of the system is approximately 30cm×30cm×30cm(length×width×height).The compact system has improved the utilization of light energy by 3-4 times higher and the exposure time for the same sample has improved by one order of magnitude compared with the original one.Thus reaching a maximum optical sectioning imaging speed of 100fps@1024×1024 pixels,thus the data acquisition time is dramatically reduced and high-resolution imaging with centimeter-scale is realized.In addition,the system's hardware controlling,data acquisition and processing and are designed and developed by ourselves.2.Since there is crosstalk between the three primary color channels of the RGB color space,and the color information and the luminance information are not separated,the method based on multi-channel integration will generate of distortion in color restoration to a certain degree,thus the true color information of the sample cannot be accurately restored.In order to solve this problem,a color decoding algorithm based on HSV color space is proposed(HSV Color Space-Root Mean Square algorithm,HSV-RMS alogorithm),which can effectively avoid the color distortion problem caused by multi-channel integration algorithms,and obtain high-resolution full-color three-dimensional images of objects.Integrated with the self-design fast image stitching method,high-resolution 3D color optical sectioning images of insects with 2cm size are obtained,and the 3D morphological quantitative analysis of the micro-nano structure on the surface are realized.Combining 3D printing technique,3D digitalization of the images can be achieved.3.A fast 3D color SIM method based on Hilbert-transform is proposed.HSV color space based structured illumination full-color 3D optical sectioning technique can recover the full color information on the surface of the samples.However,for each optical sectioning,three raw images with fixed phase shift are required.This will dramatically enlarge the data acquisition time and image processing time,especially for a large-scale sample that needs image stitching strategy.To this end,in this paper,a fast 3D color SIM based on Hilbert-transform is proposed(Hilbert Transform based Color Opitcal Sectioning alogorithm,HT-COS algorithm).Here,only two raw images are needed to reconstruct a full-color optical sectioned image for each slice.The image acquisition data is reduced by 1/3 and the reconstruction time is saved about 28%,effectively improving the efficiency and speed of the SIM for 3D color imaging.Simulations and experiments verify the effects of noise and phase shift errors on the reconstructed optical sectioning image quality between the HSV-RMS algorithm and the HT-COS algorithm,thus prove that the HT-COS algorithm is more robust and feasible.4.Deep learning is applied to SIM imaging.Traditional SIM 3D optical section imaging requires at least three phase-shifted raw images to be acquired in each layer.For large-scale samples that require multi-view stitching,this process is not only timeconsuming,but the collected data is also huge.Using lots of wide-field images collected by existing experiments as input data,the maximum projection image(extended depth of field image)formed by the reconstructed optical sectioning images as the output target.These datas are trained in the constructed convolutional neural network,and finally the optical sectioing image with large depth of field can be directly recovered from the wide-field image.After this process,the new method can reduce the data amount of the raw images by about 21 times less.And the image quality and resolution are still unchanged.