RANSAC Interactive Demo

RANdom SAmple Consensus — visualizing line fitting and homography estimation

Hanbyul Joo | Visual Computing Lab, Seoul National University | More Demos ↗

# of points 150

Outlier ratio 50%

Noise (σ, px) 3

Inlier threshold (px) 8

Max iterations 500

Iter

Cur Inliers

Best Inliers

Total Points

data point

minimal sample (this iter)

inliers of current hypothesis

ground-truth line

current hypothesis

best so far

Click "Generate New Data" to start.

# of matches 80

Outlier ratio 40%

Noise (σ, px) 2

Warp strength 30

Inlier threshold (px) 6

Max iterations 600

Iter

Cur Inliers

Best Inliers

Total Matches

Source image (point A coordinates)

Target image (point B coordinates)

match

4-point sample

current inliers

GT-warped quad

quad warped by best Ĥ

quad warped by current Ĥ (this iter)

Click "Generate New Data" to start.