摘要
文中分析并回顾了几种典型的虚点提取技术,然后给出了当存在一些干扰线条时,从混有这些线条的平行线族中提取出虚点的新方法。这是一种迭代收缩思想,它避开了一个假设:小数据集内的数据分布偏差也能以足够好的精度平滑掉。实际上,这种假设常常不成立,然而,经过数次迭代收缩的远点剥离,假设反而较易成立。给出了相应的实验结果。
In this paper, several typical methods for vanishing point extraction are reviewed, then a new approach, called recursive shrinking is proposed to this issue in our case. This method avoids an assumption that the bias in a small data set can also be balanced just by LSE filtering, and after certain numbers of shrinking or extraneous datum stripping off, an estimate of enough accuracy of the model hidden in the data set can be obtained. The corresponding experimental results are also showed.
出处
《红外与激光工程》
EI
CSCD
1998年第5期11-13,18,共4页
Infrared and Laser Engineering
关键词
虚点提取
迭代收缩
图像分割
计算机视觉
Vanishing point extraction
Iteration shrinking
Image segmentation
