基于分割的立体匹配法提取光场视差图  

Extracting Disparity Map of Light Field Use Stereo Matching Based on Segmentation

作　　者：牛娜娜[1] 韩燮[1] 

机构地区：[1]中北大学计算机与控制工程学院,太原030051

出　　处：《科学技术与工程》2017年第1期250-254,共5页Science Technology and Engineering

基　　金：国家自然科学基金项目(61379080);国家科技支撑计划基金项目(2013BAH45F02);山西省国际科技合作计划项目(2014081012)资助

摘　　要：为了解决从光场相机采集的图像中提取物体深度信息的问题,提出一种利用Lytro光场相机设备,基于图像分割的立体匹配方法估计光场图像深度。首先对Lytro相机采集的原始灰度图进行解码和颜色校正,提取相应的子光圈图像;在Matlab环境下,结合图像颜色分割、自适应立体匹配对子图像进行处理,通过平面拟合技术和BP优化得到相应的深度估计图。实验结果表明,该方法提取到深度图在视觉精度上有了明显的提高。In order to solve the problem of extracting image depth information from light filed camera,a method of the stereo matching method based on image segmentation to estimate light field image depth was put forward,that maken use of the light field camera. First the original image that was decoded and corrected collect by Lytro camera,then sub aperture image was extracted. In Matlab environment,the sub image was processed by image color segmentation and adaptive stereo matching,got the corresponding depth estimation by plane fitting technology and BP optimization. The experimental results show that using this method to extract depth map could improve the vision precision.

关 键 词：光场 Lytro相机 图像分割 立体匹配 视差图 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]