What steps will reproduce the problem?
1) Run the "RunBundler.py" script with --maxPhotoDimension set to a small
number.
2) In the generated Bundle out file, all JPEGS are saved with lossy
compression, and compression artifacts make the results MUCH less accurate when
compared with images that are resized beforehand using imagemagick for example.
What is the expected output? What do you see instead?
Much fewer, lower quality point clouds result.
To fix, change line 260 in osmbundler/__init__.py:
- photoHandle.save(outputFileNameJpg)
+ photoHandle.save(outputFileNameJpg, quality=100)
Then, although the images get scaled down for the algorithm, the downscaling
doesn't introduce compression artifacts which significantly improves the
quality of generated point cloud.
Original issue reported on code.google.com by mghe...@gmail.com on 10 Jan 2012 at 2:46
Original issue reported on code.google.com by
mghe...@gmail.com
on 10 Jan 2012 at 2:46