Study On Algorithm Of Layered Holographic Three-dimensional Display And Image Quality Optimization

Nowadays,Moore's Law promotes the development of various industries,of which 3D display industry is included.Holographic technology has always been regarded as the ultimate solution for the 3D display.The principle of interference recording and diffraction reconstruction guarantees its ability of restoring the whole information of the target 3D scene.Nowadays,phase display devices have not only freed holography from the limitation of the traditional holographic plate but also endued holography with compact,real-time,programmable features.With the assistance of the high-performance computer,the holographic display is sure to be a phenomenal display technology soon.In this thesis,we first analyze the imaging performance of phase-only liquid crystal on silicon(LCoS)and then optimize the quality of the reconstructed image.At present,the holographic display still suffers from many drawbacks such as the sinc function distribution of light intensity,speckle noise,zero-order light,high diffraction orders,conjugate images and so on.In this work,the phases of two virtual convergence lights are innovatively introduced into the traditional iterative algorithm,adjusting the region of the reconstructed image in the radial as well as longitudinal directions at the same time.It also addresses the problems of zero-order bright spots,high diffraction orders and conjugate images.Additionally,speckle noise is suppressed.Therefore,our work guarantees a huge improvement of reconstructed image quality,which provides strong support for high-resolution holographic 3D reconstruction.Based on the phases of two virtual convergence lights,an algorithm named double-convergence light Gerchberg-Saxton(DCL-GS)algorithm has been proposed.We implement an augmented reality(AR)experimental system for virtual image reception in this thesis,achieving holographic three-dimensional reconstruction with the advantage of long zooming range,large field of view,and continuous depth cues.In this algorithm,the 3D scene is sliced into multiple parallel layers according to the depth information and these layers are calculated into the corresponding holograms.Through the combination of the second phase of convergence light and the Fraunhofer diffraction calculation kernel,a three-dimensional scene with sufficient zooming range and continuous depth clues is obtained.Our evaluation indicates the maximum depth of the scene can reach 180cm.By selecting the focal length of the lens and adjusting the parameters of the algorithm,the FOV of the reconstructed image can be modulated to a maximum of 40 degrees,solving the problem of low diffraction efficiency and small field of view of LCoS.Moreover,on the basis of the algorithm mentioned above,the idea of region partition has been proposed,successfully reconstructing complex 3D scene with high calculation speed.It takes advantage of both space-division multiplexing and layer-based method and applies them to the Fresnel diffraction domain and Fraunhofer diffraction domain respectively.Thus,partitioned Fresnel diffraction(PFreD)and partitioned Fraunhofer diffraction(PFraD)have been proposed.The arrangement of region partition and side-by-side placing avoids the superposition between layers,solving the crosstalk problem and maintaining the reconstructed imaging quality.Besides,this idea is simple in 3D modeling,faster in computation owing to the limited sampling points involved in the calculation.Therefore,the reconstruction of multi-layer complex 3D scenes is achieved,which makes it possible for real-time 3D holographic display.Aiming to mitigate the effects from the off-axis distribution of sub-holograms caused by region partition,the convolution method and the oversampling method have been proposed.And the latter works it out perfectly.Furthermore,the similarities and differences between the PFreD and PFraD are compared and their application scenes are analyzed respectively.Meanwhile,the two region-partitioned methods are compared with the traditional point-cloud-based method and layer-based method.The results demonstrate the advantages of the two proposed methods in both reconstructed image quality and calculation speed under qualitative as well as quantitative aspects.Finally,the shortcomings of holographic display have been analyzed,and the possible development prospect has been prospected,which provides the direction for the further research of holographic display technology in the future.