Skip to content

Latest commit

 

History

History
16 lines (14 loc) · 813 Bytes

cv_21.md

File metadata and controls

16 lines (14 loc) · 813 Bytes

Lucas-Kanade 20 years on: a unifying framework

Simon Baker, Iain Matthews (2004)

Key points

  • Usual approach to image alignment is gradient descent
  • Different ways: additive (additive increment to parameters) or compositional (estimate incremental warp that is then composed with current warp estimate)
  • Another difference: Gauss-Newton, Newton, steepest-descent or Levenberg-Marquardt approximation in each gradient descent step
    • All are orthogonal choices (any combination possible)!
  • Choice depends on:
    1. Most noise in template (a) or input (b) image
    2. Need for an efficient algorithm
  • Best choices:
    • (1a): forwards algorithm
    • (1b): inverse algorithm
    • (2): inverse-compositional Gauss-Newton or inverse-compositional Levenberg-Marquardt (probably also best options for noisy templates!)