一种基于梯度增强的立体匹配算法

立体匹配是深度图获取的一个关键技术,传统的立体匹配使用的图像分割方法会对图像边界造成一定的破坏,为提高深度图边缘的准确度,提出了一种基于梯度增强的立体匹配算法。首先通过梯度算子对参考图像和目标图像进行预处理,增强图像边界的鲁棒性,然后采用mean-shift图像分割算法对图像进行区域分割,对分割得到的区域使用自适应块匹配法进行立体匹配,最后采用均值滤波对图像进行后处理,得到最优的视差平面。实验结果证明该算法在图像边界的匹配精度上得到了比较满意的效果。

An image enhangcement based stereo matching algrorithm

Streo matching is one of the key techniques in depth estimatiom. Traditional stero matching methods will destroy the edge of an image partly. To improve the depth estimation precision ,an image enhangcement based stereo matching algrorithm is proposed that utilizes image enhancement on the reference image and the target image and the scene is enhanced by a gradient method which makes the latter segmentation step more robust on edges,then the disparity plane is assigned to each segment.Experiment results demonstrate the superior performance of the proposed algrithm.