This paper we put forward an orientation histogram based algorithm for estimating local orientation angles after summarizing the existing orientation estimation methods and pointing out the problems of quadra-ture filters. The method computes the vector sum of all neighboring ball voting vectors for an object image pixel to get its orientation angle's histogram and search out the peak value of the histogram as its orientation angle. These orientation angles are used for later tensor voting process as orientation information. Four groups of experiments are performed on typical contours to compare the contour inpainting effect and the precision. Experimental results on edge map of city road and noisy images show that the proposed algorithm works well as basic orientation estimation method for more intact contour, less clutter pixels and robust to noises.%在总结现有取向估计方法的基础上, 针对传统的正交滤波法出现的问题, 提出一种取向直方图法进行取向估计. 该方法通过计算图像轮廓上各点的圆形投票域投票的矢量叠加得到各点的取向直方图; 然后以直方图峰值点对应的和矢量计算取向信息, 将其作为张量投票方法的底层输入; 对4种典型轮廓进行修复, 并比较了相对于正交滤波、张量叠加和矢量叠加的修复效果和计算精度. 对城市道路的边缘图进行实验的结果表明, 文中方法作为底层取向估计时投票得到的轮廓线较为完整, 投票产生的杂散点较少, 且对噪声具有一定程度的鲁棒性.
展开▼
