Research On Three-dimensional Computational Reconstruction Based On The Light Field Information Of Integrated Imaging

Integral imaging is one of the most promising autostereosopic display techniques. It has the advantages such as full parallax, continuous viewpoint, true color, working with incoherent light and being observed without the requirement of special devices. As an important subject in integral imaging field, computational reconstruction not only can avoid many problems caused by device limitation of an optical integral imaging system, but also is an effective use of the light field information of integrated imaging. However, for computational reconstruction, especially three-dimensional computational reconstruction, it has the problem of low image quality, difficult to deal with complex scenes and being unable to reconstruct a complete scene. To solve the above problems, the main contents in this paper are given as follows.Firstly, according to the basic principle of integrated imaging system, the light field sampling and the reconstruction characteristics of integrated imaging is analyzed. On this basis, the two-dimensional reconstruction of light field is researched from two aspects of depth slice reconstruction and free viewpoint reconstruction, including slice reconstruction based pixel mapping and free-viewpoint reconstruction based on ray tracing. Experiments are carried out and experimental results are presented.Next, for the problem of low image quality and difficult to deal with complex scenes, the principles of integrated imaging three-dimensional computational reconstruction is discussed. Then, a reconstruction method based on depth prediction is analyzed, and the problems and their causes are analyzed according to the experimental. Based on this, another reconstruction method based on image understanding is researched in this paper. Experimental results show that this method can deal with complex scenes and can improve the quality of the reconstruction three-dimensional image significantly.Finally, to solve the problem of integrated imaging recording scene narrowly and being unable to reconstruct a complete scene, a three-dimensional computational reconstruction method using multi-frame light field images is proposed in this paper. On the analysis of multi-frame point cloud registration principle, the registration and fusion of multi-frame point cloud data based on light field image reconstruction is realized by Iterative ClosestPoint algorithm. Then the result is smoothed. The experimental result illustrates that, compared to the reconstruction based on a single frame light field image, the method can reconstruct a three-dimensional scene whit higher quality and more complete information.