基于极线几何的改进多约束图像立体匹配

针对计算机视觉和模式识别领域基本而重要的问题--立体匹配,提出了一种基于极线几何、结合特征匹配与区域匹配、视差梯度约束等多约束立体匹配算法,实现图像快速准确匹配.该算法将现有的基于特征和基于窗口匹配两种方法相结合,并加入视差梯度等约束条件,有效弥补了单一匹配算法的不足,同时增强了算法适应性.实验表明,该算法具有更高的求解质量和求解效率,可以满足双目立体视觉系统的需要.

Improved multi-constrained image stereo matching based on epipolar geometry

Aiming at stereo matching which is an important and basic issue in the field of computer vision and pattern recognition, a multi-constrained stereo matching algorithm has been introduced. This algorithm based on epipolar geometry meanwhile it combined with feature matching, regional matching and gradient of disparity constraint to achieve fast and accurate image matching. The algorithm combined with two current popular methods which based on feature matching and window matching, and by adding some constraints, such as disparity gradient, it effectively made up the limitation of the single matching algorithm, while enhancing the adaptability of the algorithm. Experiment results show that the algorithm has higher quality and efficiency which would satisfy the needs of binocular stereo system.