Abstract: This paper proposes a method of discrete curvature estimation based on the growth of fuzzy segments for 8-connected curve in 2D space. The algorithm estimates digital curvature based on tangent orientation, in which tangents are approximated by the longest fuzzy segments grown from points on the curve with a given order d. Flexible choice of order d makes curvature estimation well adapting to noisy curves. Experimental results show that fuzzy segments improve the performance of curvature estimation by remarkably reducing the number of short tangents, and it can also deduce errors for digital curves with even noise.

Key words: curvature estimation, fuzzy segment, arithmetic geometry, curve analysis

摘要： 针对Z2空间中的8连通边界曲线，提出一种基于模糊线段生长的离散曲率估计方法。该方法引入“序”为d的模糊线段生长算法，将曲线上生长出的最长模糊线段作为切线的近似，并计算出了离散曲率。实验表明，该方法不仅较好地反映了曲线点的局部特性，而且增加了对离散曲线噪声的适应能力，离散曲率估计性能获得了显著的提高。

关键词: 曲率估计, 模糊线段, 算术几何, 曲线分析

CLC Number: 

• TN941.1