一种微透镜光场相机的子孔径图像提取方法

Extraction Method on Sub-aperture Graph of Micro-lens Light-filed Camera

多视点图像在深度计算、3D视频合成、立体匹配等技术中有广泛的应用价值。传统相机在拍摄真实场景时往往不能获取场景多个视点的信息,而新型的微透镜光场相机和相机阵列却可以得到场景更多的信息,例如场景的多视点信息,场景的深度信息等。但是由于微透镜光场相机的软件保密和源程序不开放等问题,造成研究人员无法自主提取其多视点(子孔径)图像。对此,提出了一种由微透镜光场相机提取子孔径图像的方法。首先,进行微透镜中心像素标定,然后考虑光场原图像的有效像素区域,将各中心像素的周围像素点按次序提取并重组,从而得到子孔径图像。实验表明,所提取的多视点图像具有微小的视差,体现了所提光场相机子孔径提取方法的有效性。实现了微透镜光场相机的子孔径图像提取,将为后续恢复场景深度、重建场景三维图等研究提供有效数据。

Multi-view images have an extensive application value in technologies such as depth calculation,3 D Video synthesis,Stereo matching. For traditional cameras,multiple cameras have to be used from different angles in order to capture multi-view images,but for Array Camera or micro-lens Light-filed camera can capture several images from different angles only on one shot. Especially,the micro-lens Light-filed camera based on Light field theory can get multi-view images with the information of single-camera imaging at once to obtain multi-view images,which makes it a hot topic for researchers. Whereas,for its copy rights,the source code of micro-lens Lightfiled camera is not open,resulting that researchers cannot obtain its multi-view images（also called as sub-aperture images）. Then,this paper proposes a new approach to obtain sub-aperture images from a micro lens Light-filed camera. First of all,we calibrate the center of the micro lens pixels,then considering the effective pixel area of light field original image,we extract the pixels around the center pixel in order to rearrange them and finally obtain its sub-aperture images. Experiments show that the extracted sub-aperture images have tiny disparities,which proves the effectiveness of our extraction algorithm for sub-aperture images of the light-filed camera. Our method can extract the sub-aperture images of a micro lens Light-filed camera,consequently providing the effective data for the sequent applications,such as recovering the depth of the scenarios,reconstructing the 3-dimention scene.