Skip to content

Image Stitching

Panorama


RANSAC

RANdom SAmple Consensus

motivation

avoid outliers

How to do

  • 随机选择 s 个样本 (一般 s 为最小的可以确定一个模型的样本数量)
  • 拟合模型
  • 重复 N 次
  • 选择具有最多 inliers 的模型.
  • 使用所有的 inliers 重新拟合模型.

Pros and Cons

Pros:

  • Simple & general
  • Applicable to many problems
  • Often works well in practice

Cons:

  • Has parameters to tune
  • May fail for low inliers ratios
  • Requires too many iterations