基于物体轮廓和前向预测的自适应窗匹配

Matching with Adaptive Windows Based on Object Contours and Forward Prediction

针对用窗方法进行视差估计的问题 ,在详细分析了不同情况下窗大小对亮度差平方和(SSD)影响的基础上 ,根据立体图像内容的特点和自然界物体的性质 ,提出了基于物体轮廓和前向预测的自适应窗匹配方法 ,用于立体图像对对应像素点的视差估计 .对一维亮度信号的仿真分析清楚地展示了该算法能在图像的各像素点迅速合理地调整窗大小 ,从而达到几乎在每一像素点的匹配窗尺寸参数都是最优的目标 .与以前提出的自适应窗匹配法比较 ,该算法复杂度更低 .最后 。

Based on the comprehensive analysis of the effect of window size on sum of squared differences (SSD) in different cases, according to the property of natural objects and the features of the content of stereo images, a method to adjust window size adaptively is proposed for stereo matching based on object contours and forward prediction. In computer simulations, the processing of one-dimension intensity signals reveals clearly that the algorithm can adjust window size quickly and reasonably, and achieve the aim that window size is optimal at almost all pixels. Compared to those previous adaptive window matching algorithms, this algorithm is simpler.