The Research On Large Field Of View Holographic Image Mosaic Captured Through DMD Scanning

The advantage of optical holography and modern computer technology are combined in digital holography technology.The holographic image could be recorded by CMOS and easily processed through the computer.Compared with the traditional imaging technology,the phase information is contained in holographic image,so the three-dimension shape of the measured object can be reconstructed.It has proved to be attractive in the field of biomedicine and precision measurement because of the above advantages.However,limited by the size and resolution of the CMOS sensor,the field of view is very small,which cannot meet the demand of large-scale measurement.The traditional scheme for increasing the field of view is dependent on the mechanical motion structure,leading to the introduction of unavoidable errors.And the control system of high-precision motion structure increases the complexity of the system design.In this thesis DMD is used to replace the mechanical structure,which reduces the costs and increases the stability.In this thesis we discussed the research in the fields of digital holography and image stitching.Firstly,the principle of digital holographic recording and rebuilding are described in detail.Based on the previous research results,the digital holographic image is recorded and the recorded object is rebuilt through numeric calculate.As for the noise interference in the recorded digital holographic image,we analyze the process of noise introduction.The experimental light path is improved to filter background stray light,and the zero-order image is eliminated by recording the background light.For the random noise that couldn't filter by improving the experiment system,image filter technology is introduced.After the process of reconstructing,the conjugate image appeared in the object area.An iterative algorithm is introduced to eliminate the conjugate image interference.For the problem that the calculation takes a long time,the GPU is used to speed up the algorithm.For the limitation of the field of view,we improve the previous light source scanning scheme and replace the mechanical motion structure with DMD.Experimental results show that the scheme is effective to eliminate the random jitter introduced by mechanical components and easily to achieve.When the light source scans,the local information of the measured object is captured.Image stitching technology is introduced to synthesize the images into the same image.The image registration and image fusion are laid special stress.In general,image registration is based on image feature information.Firstly,the SIFT operator is introduced and the effectiveness in holographic image registration is validated.However,due to the high dimension of SIFT,it's difficult to achieve real-time processing.In addition,SIFT uses only local features and has more mismatched points when the hologram texture is similar.In this thesis,global information is added in the SIFT feature.After searched the matching points,RANSAC algorithm is used to remove the mismatched feature point pairs.The GPU is used to accelerate the above feature extraction process in order to save computing time.How to maximize the registration of overlapping information in the same scene is the task to be achieved in image fusion.Due to the large amount of data to be processed during digital image fusion,in this thesis the effects of wavelet image fusion strategy and the strategy based on guided image filtering are compared.Based on the theoretical analysis above,the holographic imaging system is built.