基于阵列图像立体匹配的深度信息获取方法

Depth Information Acquisition Method Based on Stereo Matching of Array Images

本文实现了一种对3×3微阵列镜头获取的阵列图像进行匹配计算得到带有深度信息图像的方法.方法首先对3×3微阵列镜头进行单个镜头标定以及单个镜头分别和中心镜头进行立体标定,得到相机参数;然后建立对应的能量代价函数,构建马尔可夫随机场,设定其初始深度,在空间进行划分层次;最后运用图割算法求解能量函数最小化,选择最优匹配值,进而得到带有深度信息的匹配图.实验结果表明本文基于阵列图像立体匹配的深度信息获取方法是有效的;相比一般双目视觉系统,微阵列镜头的体积尺寸小、可以同时拍摄获取图像,能够得到比双目视觉更准确的深度信息.

In this paper,a method is implemented based on array images captured by microarray lenses which are calculated the stereo matching and then get the depth information of them.The method is,first,in order to get the intrinsic parameters and the extrinsic parameters of the microarray lenses,each sub lens of microarray lenses must be calibrated,and the center lens must be stereo calibrated with other lenses,then use the intrinsic parameters of each lens to distortion correction the image captured by each sub lens,and use the external parameters of two lenses to do space position transformation,which can be unified to a identical coordinate system of the center camera;then the Energy Cost Function is built under certain conditions,and the Markov random field is constructed,set the initial depth and level division in space;finally the energy function minimization will be solved by using graph cut algorithm,the optimal match value will be chose,and then get the matching image with depth information.The experimental results show that the depth information acquisition method based on the stereo matching of the array images is effective;compared with normal binocular vision system,microarray lenses has smaller size and obtain the images simultaneously,can get more accurate depth information than the binocular vision.