Research On High-speed Method For Generating Elemental Image Array Of Integral Imaging

Integrated imaging is a real three-imensional display technology that has a simple optical structure and can be viewed with the naked eye.It uses a lens array to realize spatial and angular sampling of a three-dimensional light field,and then it can be converted into a three-dimensional reconstruction image in a certain depth range by optical or computer reconstruction at the display side.However,due to the limitations of viewing angle,resolution and other factors,integrated imaging still remains in the theoretical research stage and there is still a certain distance from the actual application.In order to achieve real-time acquisition of high-quality integrated imaging image arrays,this thesis uses sparse camera array acquisition system to study a real-time image array generation method based on CPU-GPU heterogeneous computing system,including calibration and correction of sparse camera arrays,depth real-time acquisition based on stereo vision,real-time generation of virtual multi-view arrays based on virtual viewpoint generation,and real-time generation of EIA based on multi-viewpoint ray tracing.The specific research content is as follows:(1)Research on integrated imaging 3D scene acquisition method based on sparse camera array.Firstly,the calibration and correction methods of the sparse camera array acquisition system are studied.Based on Zhang's dual-target algorithm,the sparse camera array is divided into left and right vertical camera pairs,and the left and right vertical camera pairs are respectively calibrated within the group.Based on the above camera pairs,the left and right vertical camera pairs are stereoscopically calibrated,all the cameras are unified into the same spatial acquisition coordinate system,and the collected sparse image arrays are calibrated using the calibration results.Then,the depth acquisition technology based on the stereoscopic vision technology is studied,and the disparity map between adjacent image pairs in the sparse image array is calculated by stereo matching,and the depth information of the three-dimensional scene is obtained.(2)Research on parameter matching in 3D display of integrated imaging based on multiviewpoint generation.Firstly,using the dense disparity map generated by the stereoscopic vision technology,and the sparse image array,the virtual multi-viewpoints are divided into four different areas: top left,top right,bottom left,and bottom right,respectively.Image consists of a virtual multi-view image array.Then,according to the multi-view ray tracing method,generate a high-quality EIA that matches any display device and can form a clear three-dimensional image in a specific depth range.(3)Research on a real-time image array generation method based on CPU-GPU Heterogeneous computing system.Based on the proposed parameter matching model for integrated imaging 3D scene acquisition and display,aiming at the characteristics of high parallelism among pixels in digital image processing,how to effectively use the multithreaded computing capability of GPU is studied to study the depth real-time acquisition of three-dimensional scenes based on stereo vision.real-time generation of virtual multiview image arrays based on DIBR technology,and real-time generation of EIA based on multi-viewpoint ray tracing,improves the efficiency of EIA generation.