Stitch Images into a Panorama
Two overlapping square photographs need one wide view.
Stitch Images into a Panorama explained visually
Two overlapping square photographs need one wide view. English captions are optional and start off. The complete transcript follows the written lesson.
What you will learn
Two overlapping square photographs need one wide view.
A practical question
Our two garage photos overlap, and we want one wide view. First align their shared features with a geometrically checked mapping. Then warp one view into the other's coordinate system and decide how to combine pixels where both contribute.
The useful idea
Exposure differences can leave a seam, and a moving car can appear twice or be cut in half. Parallax is another problem: nearby objects shift differently from a distant wall, so one planar mapping cannot align every depth at once.
Work through the example
Compare the panorama with both original images. Inspect the shared painted line, the nearby car, and the border of the stitched canvas. A smooth blend might hide a bad geometric alignment; check correspondence before polishing the seam.
Limits and next step
Keep the original images, matches, transform, and blend settings. Next, we'll learn to spot a match set that looks numerically strong but is geometrically wrong.
Worked example
Check the alignment of corners before blending the seam.
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
Estimate a suitable transform, warp one image, and blend their shared region. Parallax between near and far objects prevents a single plane transform from aligning everything.
What is the main idea behind stitch images into a panorama?
Estimate a suitable transform, warp one image, and blend their shared region.
How is it used in the example?
Check the alignment of corners before blending the seam.
What limitation should be checked?
Parallax between near and far objects prevents a single plane transform from aligning everything.
Further reading
Programming Computer Vision with Python — Jan Erik Solem.
Original ToolsInVision teaching examples. Editorial approach · Suggest a correction
Accessible referenceLesson 31 transcript141 words · Video transcript · Matches the published narration and captionsOpen transcript
Our two garage photos overlap, and we want one wide view. First align their shared features with a geometrically checked mapping. Then warp one view into the other's coordinate system and decide how to combine pixels where both contribute.
Exposure differences can leave a seam, and a moving car can appear twice or be cut in half. Parallax is another problem: nearby objects shift differently from a distant wall, so one planar mapping cannot align every depth at once.
Compare the panorama with both original images. Inspect the shared painted line, the nearby car, and the border of the stitched canvas. A smooth blend might hide a bad geometric alignment; check correspondence before polishing the seam.
Keep the original images, matches, transform, and blend settings. Next, we'll learn to spot a match set that looks numerically strong but is geometrically wrong.