lec10 prelim

lec10 prelim - Lecture 10: Sept 27, 10 HW3 posted...

Info iconThis preview shows pages 1–6. Sign up to view the full content.

View Full Document Right Arrow Icon
USC CS574: Computer Vision, Fall 2010 Copyright 2010, by R. Nevatia 1 Lecture 10: Sept 27, 10 • HW3 posted – Questions? • Mid-term: Oct 13, class time, closed book, details to come • Last Class – Intro to Stereo – Triangulation – Correspondence • Sparse, followed by surface fitting • Local methods: intensity/color matching • Combine with segmentation: Tombari et al • Today’s objective – Global correspondence methods
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
USC CS574: Computer Vision, Fall 2010 Copyright 2010, by R. Nevatia 2 Use of Context • How to reduce match ambiguity by use of context? – Example only includes possibilities preserving point order • How to reduce complexity of search?
Background image of page 2
USC CS574: Computer Vision, Fall 2010 Copyright 2010, by R. Nevatia 3 Global Matching • In general, we expect object surfaces in the world to be smooth, but discontinuities to be present at the object boundaries – We don’t know where the object boundaries are prior to stereo computation, this is part of the need for stereo analysis – Edges and regions give some hints to location of the boundaries • Variety of global methods exists – Scan line optimization: dynamic programming – 2-D optimization: graph cuts, belief propagation – Integrated with use of lower level segmentation
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
USC CS574: Computer Vision, Fall 2010 Copyright 2010, by R. Nevatia 4 Scan line Optimization • Originally developed for edge matching – Individual edge matches are ambiguous, match intervals * * * x x x x – Which correspondences are best (some edges may have no matches) • Order of edges should be preserved – Common but can be violated if small thin object between cameras and another object (example in fig. 11.13) • Lengths of intervals should be as close as possible in two images • Brute force complexity is O( m n ) given m and n edges – Dynamic programming reduces complexity to O( m x n )
Background image of page 4
USC CS574: Computer Vision, Fall 2010 Copyright 2010, by R. Nevatia 5 Dynamic Programming • Define a cost function for matching two intervals, C(i,j), e.g. SSD between the two intervals
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 6
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 11/23/2010 for the course CS 574 taught by Professor Ramnevatia during the Fall '10 term at USC.

Page1 / 22

lec10 prelim - Lecture 10: Sept 27, 10 HW3 posted...

This preview shows document pages 1 - 6. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online