摘要
首先基于相邻像素间的欧氏距离计算出轮廓曲线上每一点的支持区域及支持半径,用线性插值得到支持区域的端点;然后求得支持区域的质心,以曲线上的点和其相应的支持区域质心之间的距离作为特征响应筛选出候选特征点;最后将具有局部极大特征强度的点作为多边形的顶点。实验表明这种方法能降低曲线数字化所带来的影响,对旋转也具有较高的不变性。
For every point on contour curves, the support area and support radius were first calculated based on actual distance of adjacent pixels, end points were obtained with linear interpolation technique; Next, the distance between point and its eentroid of support area was used as feature strength to determine feature points candidate; Finally, those points with local maximum feature strength were selected as vertices of polygonal. Experiments demonstrate that this algorithm has strong errortolerant capacity for curves digitization and rotation.
出处
《计算机应用》
CSCD
北大核心
2006年第3期577-578,604,共3页
journal of Computer Applications
基金
国家自然科学基金资助项目(60372072)
关键词
多边形近似
特征点检测
支持区域
polygonal approximation
feature point detection
support area
