Research On Three-dimensional Display Technology Based On Computer-generated Hologram

Holographic 3D display is one of the important research directions in the field of 3D display.It can not only truly reproduce the amplitude and phase information of 3D scene,but also has good viewing comfort without wearing any auxiliary devices.The traditional optical holographic display uses the principle of light interference and diffraction to "freeze" all the information of the three-dimensional scene in the form of interference fringes,and then irradiates the interference fringes by the reference light to reproduce the original three-dimensional scene.With the rapid development of computer technology,computer-generated hologram technology came into being.With its many advantages such as simple and easy operation,convenient storage and wide application,it is widely favored by researchers,especially the three-dimensional display of pure phase computer-generated hologram,which has become one of the mainstream display methods because of its high diffraction efficiency.In recent years,with the rapid development of artificial intelligence,deep learning algorithm has become one of the important means to solve research problems in various fields.In display research,deep learning plays an important role in the fields of computer-generated holographic imaging,super-resolution microscopic imaging technology,ghost imaging and so on.In the era of Internet data-driven,deep learning algorithm provides a new idea to solve the research problems in the field of computergenerated holographic display.However,in computer-generated hologram display,most of the traditional phase only holographic imaging methods rely on high-intensity iteration,which takes a long time and the imaging quality is not high.To solve this problem,this thesis uses the combination of deep learning algorithm and computer-generated hologram to carry out relevant research work.The main innovations and research contents are as follows:(1)The basic imaging principle of traditional optical holography and computer-generated holography,and the principle of amplitude type computer-generated hologram and phase type computer-generated hologram are introduced.Secondly,three main methods of CGH imaging are described: point source method,panel method and layered angular spectrum method.In this thesis,the research is mainly based on deep learning algorithm and computer-generated holography.Therefore,the basic theory of deep learning is introduced,and the method of combining deep learning and computer-generated holography to solve the problem of phase prediction,the generation of hologram and the reconstruction of three-dimensional object are described.(2)Aiming at the problems of large amount of iterative calculation and low imaging quality of traditional pure phase hologram,a pure phase hologram generation algorithm combining deep learning and layered angular spectrum is proposed,which can quickly generate hologram and improve the quality of reconstructed image of hologram.The complex amplitude information of three-dimensional scene is predicted by lenet network structure,which reduces the amount of calculation;The accurate angle spectrum algorithm is used to generate high-quality phase only holograms of three-dimensional scenes,which improves the imaging quality.The feasibility of the proposed algorithm is proved by software simulation and experiments.(3)In order to further improve the reconstruction quality of the image,the digital image processing algorithm is introduced to suppress the speckle noise on the hologram and the reconstructed image.Firstly,the basic principle of speckle noise is studied.After comparing the denoising ability of four main speckle noise suppression methods,the image denoising method based on wavelet transform and the image denoising method based on nonlocal mean filtering are selected to suppress the speckle noise in holograms and reconstructed images.The numerical values of PSNR and SSIM are used to measure and compare the image quality after denoising.The simulation results show that the selected filtering algorithm has a good suppression effect on speckle noise and further improves the imaging quality.