基于立体像素匹配的图像重构技术研究

Imaging Reconstruction Technique Based on Tridimensional Pixel Mapping

为了解决目前全景成像技术中分辨率低的问题,提出了一种新的基于3D场景立体像素光线映射的全景图像计算机重构技术.在全景成像技术中,3D场景的每个立体像素点经全景成像系统的编码系统分别映射在一定区域的多个体元素图像的不同像素点上.在计算机重构全景图像时,根据逆光学路径原理,提出了从立体像素映射到的体元素图像区域中提取对应立体像素的多个2D像素点来重构全景图像,使重构的全景图像最大分辨率可达到传统成像方法图像分辨率的N倍(N为映射区域面积).提出的立体像素的匹配技术大大提高了重构的计算机全景图像分辨率.

A novel CIIR technique was proposed to dramatically improve the image resolution of the computational reconstructed object images. According to integral imaging system principle,one pixel of the 3D scene was recorded by multiple elemental images. Contrary optic-ray pickup method was introduced in the CIIR process to calculate each position of the pixel mapped by tridimensional pixel on the elemental images. In computational integral imaging system, each mapping pixel of elemental imaging was contributed to the reconstruction image. As the number of mapping pixel was increased, the reconstruction image resolution would be raised. The maximal resolution is N times of conventional reconstruction image（N is the number of elemental image mapped by tridimensional pixel）. To show the feasibility of proposed method, some experiments were performed, and the results are compared and discussed with those of the conventional method. The reconstruction image resolution can be dramatically improved in the proposed method.