Corner Detection

Corner: Point where 2 edges meet. i.e., rapid changes of image intensity in 2 directions

• Taking gradients
• distribution of gradients

→ Classify the distribution of gradients to detect corners

• Moravec corner detector
• Harris Corner Detector

Harris Corner Detector

• Fitting an Ellipse to the distribution
• the length of the axes become the parameters
  • if both axis are small → flat region
  • if λ₁ >> λ₂ → edge region
  • if both are large → corner region
    • response function maps axis values to a number R
      • R = λ₁ λ₂ - k (λ₁ + λ₂) ²
      • where 0.04 ≤ k ≤ 0.06
    • if R > threshold T → both axes are sufficiently large → Corner region
    • 
    • 
    • 
• Finding axes values
• 
• invariant to image rotation
• not scale invariant
  • Use Harris-Laplace
    • multiscale harris corner detection
    • scale = Laplacian of Gaussian, approximated by Difference of Gaussians

Non-Maximal Suppression

• The detector is likely to produce large responses not only at the exact location of the feature but also close to it
  • find the exact locations of the corners → detect the peak of each of these clusters → find local maximas
• Slide a window of size k over the image
  • if the pixel at the center is the maximum value within the window, label it as positive (retain it).
  • Else label it as negative (suppress it → reduced values, or eliminate it → set to 0)