摘要
最小二乘渐进迭代逼近(Least Square Progressive and Iterative Approximation,LSPIA)是有效的数据拟合方法,但应用非均匀B样条在拟合低分辨率图像轮廓过程中可能产生自相交、曲线扭曲、光滑角等不合理现象。该文应用长度优先、避免自相交和避免对尖角光滑处理这3种先验知识对LSPIA解空间进行约束,并为LSPIA矢量图轮廓拟合提供合理的规范。数值实验表明:引入先验知识后能很好地解决图像矢量化过程中出现的上述3种不合理现象,得到更加合理和美观的轮廓。
Least Square Progressive and Iterative Approximation(LSPIA)is an effective data fitting method,but the application of non-uniform B-spline may produce unreasonable phenomena such as angles,self-intersection,distortion and smoothness in the process of fitting low-resolution image contours.This paper applies length first,avoiding self-intersection and avoiding smooth sharp corner to constrain the LSPIA solution space and provide reasonable specifications for LSPIA vector contour fitting.Numerical experiments show that the introduction of prior knowledge can solve the three mentioned unreasonable phenomena in the process of image vectorization,and obtain more reasonable and beautiful contour.
作者
洪庆飞
李亚娟
邓重阳
HONG Qingfei;LI Yajuan;DENG Chongyang(School of Sciences,Hangzhou Dianzi University,Hangzhou Zhejiang 310018,China)
出处
《杭州电子科技大学学报（自然科学版）》
2021年第3期58-61,共4页
Journal of Hangzhou Dianzi University：Natural Sciences
基金
国家自然科学基金资助项目(61872121,6191101102)。
关键词
非均匀B样条
自相交
曲线扭曲
光滑角
轮廓
non-uniform B-spline
self-intersection
curve distortion
smooth angle
contour
