【摘要】 随着计算机图形学的发展,立体匹配技术已经成为三维场景恢复中一项重要的手段,视差估计是立体匹配中的关键基础技术。为了能够更好进行三维场景恢复,改善视差的工作便迫在眉睫。主要研究了在基于模板的可信传播立体匹配算法中改善视差初始值的算法:通过引入梯度差算子与绝对差和算子加权的匹配代价,运用交叉检验估计及WTA优化初始视差矩阵,进而提高初始视差值的准确性,从而改善最终的视差结果。经实验证明,本方法能够很有效地去除噪点,进而获得较高质量的视差结果。更多还原

【Abstract】 With the development of computer vision,stereo matching has become a very significant tool in three-dimensional scene recovery.Disparity estimation is fundamental technique in stereo vision.In order to improve the three-dimensional scene recovery,raising the accuracy of the disparity is very necessary.It mainly studies a method to improve the initial value of the parallax in stereo matching using belief propagation.Focus on improving the accuracy of the initial disparity by introducing a new matching cost which is weighted by absolute differences of gradient and sum of squared differences,applying the cross checking method to estimate the disparity matrix and optimizing it with the WTA.Experimental results show that the approach reduces noises efficiently and gets refined disparity map.更多还原