摘要
为解决移动视觉系统的动态噪声问题,提出了一种可适应动态噪声的立体匹配算法。对视觉图像进行分割,利用Kalman滤波算法估计噪声对图像分割的影响,并以此动态调整分割精度,以分割边缘特征点作为基元利用置信度传播(belief propagation,BP)算法提取出边缘特征点视差,最后根据特征点视差统一对分割区域进行赋值,得出最终视差图。实验结果表明,该算法不仅符合移动视觉系统的动态实时性要求,而且能适应动态噪声影响,得出精度较高的立体匹配结果。
To solve the dynamic noise of mobile vision system,a dynamic noise adaptation algorithm for stereo matching is put forward.The algorithm first segmented image,then estimated the impact of noise on visual images using Kalman filter in order to adjust the dy-namic segmentation.Using belief propagation（BP） algorithm to extract edge parallax based-on the edge points.Finally,assign the segmented regions according to the edge parallax,and obtain the final disparity map.The results show that it can not only adapt to the dynamic system well,but also get high precision stereo matching results under dynamic noise.
出处
《计算机工程与设计》
CSCD
北大核心
2011年第4期1394-1397,共4页
Computer Engineering and Design
基金
江苏省自然科学基金项目(BK2009093)
