This preview shows pages 1–20. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: EECS 442 Computer vision Fitting methods Reading: [HZ] Chapters: 4, 11 [FP] Chapters: 16 Some slides of this lectures are courtesy of profs. S. Lazebnik & K. Grauman Problem formulation Least square methods RANSAC Hough transforms  Multimodel fitting Fitting helps matching! Fitting Goal: Choose a parametric model to fit a certain quantity from data Lines  Curves Homographic transformation Fundamental matrix Shape model Example: Computing vanishing points H Example: Estimating an homographic transformation Example: Estimating F A Example: fitting an 2D shape template Example: fitting a 3D object model Fitting Goal: Choose a parametric model to fit a certain quantity from data Critical issues: noisydata outliers missing data Critical issues: noisy data A Critical issues: noisy data (intraclass variability) H Critical issues: outliers Critical issues: missing data (occlusions) Fitting Goal: Choose a parametric model to fit a certain quantity from data Techniques: Least square methods RANSAC Hough transform EM (Expectation Maximization) [forthcoming lecture] Least squares methods fitting a line  Data: ( x 1 , y 1 ), , ( x n , y n ) Line equation: y i = mx i + b Find ( m , b ) to minimize = = n i i i b x m y E 1 2 ) ( ( x i , y i ) y=mx+b 2 2 = = Y X XB X dB dE T T [ ] 2 2 n 1 n 1 n 1 i 2 i i XB Y b m 1 x 1 x y y b m 1 x y E = = = = M M M Normal equation = = n i i i b x m y E 1 2 ) ( Y X XB X T T = Least squares methods fitting a line  ( ) Y X X X B T 1 T = ) XB ( ) XB ( Y ) XB ( 2 Y Y ) XB Y ( ) XB Y ( T T T T + = = b Ax = More equations than unknowns Look for solution which minimizes Axb = (Axb) T (Axb) Solve LS solution ) ( ) ( = i T x b Ax b Ax b A A A x T T 1 ) ( = Least squares methods fitting a line  t 1 t A ) A A ( A + = U V A 1 1 = with equal to for all nonzero singular values and zero otherwise 1 + = pseudoinverse of A Solving b A A A x t t 1 ) ( = Least squares methods fitting a line  t V U A = U V A + + = = SVD decomposition of A Least squares methods fitting a line  = = n 1 i 2 i i ) b x m y ( E ( x i , y i ) y=mx+b ( ) Y X X X B T 1 T = = b m B Not rotationinvariant Fails completely for vertical lines Limitations Distance between point ( x n , y n ) and line ax+by=d Find ( a , b , d ) to minimize the sum of squared perpendicular distances ax+by=d = + = n i i i d y b x a E 1 2...
View
Full
Document
 Fall '09
 Savarese

Click to edit the document details