GhostViewer

See Through Time

Alignment

Estimated reading: 2 minutes 47 views

Not happy with the alignment result? Try another alignment mode

GhostViewer includes three transform modes. They use the same points, but each mode handles geometry differently.

1) Similarity (recommended first)

What it allows

  • Move (X/Y shift)
  • Rotate
  • Uniform scale (bigger/smaller)

What it does not allow

  • Skew/shear
  • Perspective keystone correction

Best for

  • Most comparisons where camera angle is fairly similar
  • Cases where you want a stable, natural-looking result
  • Avoiding “leaning” buildings

If your result looks twisted or stretched

  • Switch to Similarity first

2) Affine (more flexible)

What it allows

  • Everything in Similarity, plus:
  • Shear/skew (non-uniform shape change)

Best for

  • Slight scan distortion in old photos
  • Mild viewpoint differences
  • When Similarity is close but not quite matching edges

Watch out for

  • Can introduce a subtle “lean” if points are uneven or clustered

3) Perspective (most powerful, most sensitive)

What it allows

  • Full perspective warp (keystone correction)
  • Independent corner movement

Best for

  • Strong viewpoint differences
  • Street scenes with obvious perspective change

Watch out for

  • Most sensitive to point quality
  • If points are noisy or not well spread, results can look heavily distorted

Quick mode guide

  • Start with Similarity
  • If still off, try Affine
  • Use Perspective only when you clearly need full keystone correction

Point quality matters more than mode

For all modes:

  • Use 4–8 points
  • Spread points across the full image (not just one building corner)
  • Pick fixed features (corners, window edges, rooflines, road markings)
  • Avoid trees, cars, people, shadows

Common troubleshooting

  • Image looks bent/leaning

Try Similarity and re-pick wider-spread points.

  • Center lines up but edges are off

Try Affine.

  • Top/bottom perspective still wrong

Try Perspective with better-spread points.

  • Result looks wild/distorted

Points are likely too clustered or mismatched. Re-pick points and start again with Similarity.

Leave a Reply

Your email address will not be published. Required fields are marked *

Share this Doc

Alignment

Or copy link

CONTENTS

Proudly Powered by WordPress