Find Corresponding Points Across Views
Many bricks look alike in the two camera images.
Find Corresponding Points Across Views explained visually
Many bricks look alike in the two camera images. English captions are optional and start off. The complete transcript follows the written lesson.
What you will learn
Many bricks look alike in the two camera images.
A practical question
In the left camera, we can see one distinctive tail-light corner. Which pixel in the right camera shows that very same physical point? Stereo correspondence tries to answer that question for selected points or for many pixels.
The useful idea
Compare local appearance along the epipolar search line. Watch for surfaces with little texture, shiny reflections, and parts visible to only one camera. A plausible brightness match on a different car is not a true correspondence. Consistent calibration and rectification can make the search easier, but they do not solve occlusion.
Work through the example
Overlay a few verified point pairs on both raw images. Inspect a blank car door and a repeated parking stripe as hard cases. If the match isn't defensible, mark it uncertain instead of passing it to a depth calculator.
Limits and next step
Good correspondences let us combine two viewing rays. Next, we'll see how their intersection gives an estimated three-dimensional point.
Worked example
Compare candidate brick matches along the relevant epipolar line and reject inconsistent neighbors.
Quick check: test your understanding
Try it yourself: In the situation described above, explain the method and one reason its output might be misleading.
Check the reasoning
Correspondence combines appearance with geometric and visibility constraints. Occlusion means some pixels have no valid match at all.
What is the main idea behind find corresponding points across views?
Correspondence combines appearance with geometric and visibility constraints.
How is it used in the example?
Compare candidate brick matches along the relevant epipolar line and reject inconsistent neighbors.
What limitation should be checked?
Occlusion means some pixels have no valid match at all.
Further reading
Programming Computer Vision with Python — Jan Erik Solem.
Original ToolsInVision teaching examples. Editorial approach · Suggest a correction
Accessible referenceLesson 34 transcript146 words · Video transcript · Matches the published narration and captionsOpen transcript
In the left camera, we can see one distinctive tail-light corner. Which pixel in the right camera shows that very same physical point? Stereo correspondence tries to answer that question for selected points or for many pixels.
Compare local appearance along the epipolar search line. Watch for surfaces with little texture, shiny reflections, and parts visible to only one camera. A plausible brightness match on a different car is not a true correspondence. Consistent calibration and rectification can make the search easier, but they do not solve occlusion.
Overlay a few verified point pairs on both raw images. Inspect a blank car door and a repeated parking stripe as hard cases. If the match isn't defensible, mark it uncertain instead of passing it to a depth calculator.
Good correspondences let us combine two viewing rays. Next, we'll see how their intersection gives an estimated three-dimensional point.