The Research Of Lensfree Digital Holographic Image Super-resolution Methods

Digital holography refers to holographic methods that use solid-state detectors,and store the holograms in digital form.The main advantage of this method is that it can acquire quantitative 3D data of a specimen at a single shot,in contrast to other methods requiring multiple sequential scans.In recent years,digital holography has made progresses in the fields of object recognition,lens aberration compensation,microscopy,etc.But the poor resolution restricted its further practical applications.The resolution enhancement is a critical problem in this area.In order to adapt to the development of biomedicine and information security.Lensfree digital holographic image super-resolution methods is still an important part of the research in these fields.Through the analysis of the current super-resolution technology development,we focus on the super-resolution methods in holography in the thesis.Aiming at the limit of the reconstruction image low-resolution,we provided an extraordinarily detailed of different effective algorithms used in most research,especially in the compressive sensing based holography super-resolution and convolutional neural network holography super-resolution we proposed.The compressive sensing based reconstruction algorithm contains sparse dictionary training,Semblable patch mapping algorithm and compressive sensing reconstruction algorithm design.The convolutional neural network based algorithm contains pixels super-resolution,and convolutional neural network system training and setting.The effectiveness and performance of the presented algorithms are verified by experiments and comparative analysis.The self-learning based super-resolution method and compressive sensing method are effectively enhanced the holograms reconstruction resolution.The Peak Signal to Noise Ratio of reconstruction result based on convolutional neural network rise nearly 37.2% compared with Projection Onto Convex Sets method,and the operating speed reached 2.76 s.The Peak Signal to Noise Ratio of compressive sensing based method rise nearly 5.1% Theholography algorithms we proposed in this thesis have far-reaching meanings to the development of relevant fields.