21.03.2013 Views

Computing Visual Correspondence with Occlusions via Graph Cuts

Computing Visual Correspondence with Occlusions via Graph Cuts

Computing Visual Correspondence with Occlusions via Graph Cuts

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

in the right image. Furthermore, this assumes that each pixel in the left<br />

image is mapped into some pixel in the right image. In reality, pixels in the<br />

left image can be occluded and thus not correspond to any pixel in the right<br />

image.<br />

3 Overview of the new algorithm<br />

3.1 Problem representation<br />

Let A be the set of (unordered) pairs of pixels that may potentially corre-<br />

spond. For stereo <strong>with</strong> aligned cameras, for example, we have<br />

A = {〈p, q〉 |py = qy and 0 ≤ qx − px

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!