Can you estimate a camera’s position from a cityscape photo if the buildings are known?
Asked 5/16/2019
4 views
2 answers
0
We’re digitizing historic city photos and want to estimate where a camera was located using only the image plus a database of known building positions and dimensions. In many cases we can identify buildings in the photo and know their real-world geometry.
My question is whether a single cityscape image contains enough perspective information to determine not just a viewing direction or line of possible camera positions, but an actual point or bounded area along that line.
I’m not asking for code, only whether this is geometrically possible in principle, and what kind of image features make it feasible. The eventual goal would be to automate the process as much as possible, so methods that rely on many manual comparisons are less practical.
Originally by Photography Stack Exchange contributor. Source · Licensed CC BY-SA 4.0
Photography Stack Exchange contributor
7y ago
2 Answers
12
You can often use techniques from piloting. You can find alignments in the picture, for instance the corner of a building that masks off the 2nd vertical line of windows of the building in the back. From this you determine a line on which the camera must have been. With two more such alignments you get three lines that give you a triangle in which the camera must have been, and its size gives you a degree of accuracy. Foreground elements can then give you a very precise position. GoogleEarth is your friend.
Somehow unrelated to photography, I was hunting for new housing a couple years ago and the advertisements never give the complete address, but practicing that technique using the views from windows and balcony in the advertisement pictures I was usually able to spot the building.
Originally by user75947. Source · Licensed CC BY-SA 4.0
user75947
7y ago
0
Generated from our catalog & community — verify before relying on it.
Yes—at least in principle, a single photo can contain enough perspective information to estimate the camera position, not just its direction.
The key is perspective geometry. If you know real-world building locations and dimensions, then alignments visible in the image—such as one building edge masking part of another, known separations between features, or relative spans/heights—constrain where the camera could have been. One alignment gives a line of possible positions; multiple independent alignments can narrow that to a small region or even a single point. Foreground objects usually improve accuracy a lot.
This is why a wide-angle shot taken nearby and a telephoto shot taken farther away can sometimes be distinguished: the perspective relationships between near and far structures change with camera position, not simply with focal length.
However, it is not always easy or unique. If buildings are nearly collinear, if features are obscured, or if you have too few reliable correspondences, the solution may be weak or ambiguous. In practice, success depends on having enough identifiable geometric features and a good camera model.
Recommended products
UniqueBot
AI7y ago
Your Answer
Related Questions
Can you estimate a photo’s shooting location without GPS EXIF using visible landmarks?
Why does a known-size object measure different pixel widths at different positions, and how can I estimate image magnification more accurately?
What should I look for when photographing a historic, dilapidated prison interior?
What useful scene clues can be inferred from EXIF metadata alone?
How can I measure an object's size in a photo using a ruler reference?