Merge conversion improvements from custom branch. #5
Conversation
This avoids unnecessary quality loss from several jpg-to-jpg conversions. Internal conversions are done with PNG to prevent quality loss. Final POVRay output is converted from PNG to JPG at Q90. Ideally, a flag could be added to use the original PNG output as the final image for users who want more control over the process. This re-unifies @sjuxax and @khufkens master. Some comments and minor tweaks made during the squash. See sjuxax/less-lossy for further commit history. Thanks to https://stackoverflow.com/a/12797512/68573 for working inline comments in bash ;)
|
Hi @sjuxax thanks for the additions. Good point in not stacking the jpg to jpg conversions! I'll merge it into master as others can indeed benefit from this. |
|
Thanks for the merge. I have a MacOS system that I'll test on soon, though with the correction from shellcheck and the previous work from @gerlos I'd expect it to work without issue. I have a side question here re: the Theta V. Recent photos I've taken are coming straight from the camera with zero values in the Pitch or Roll EXIF tags. It's not in either the normal tags nor the RicohPitch/RicohRoll tags. If I load these into the official desktop app, the option to "Write with top/bottom correction" is grayed out. Since this basically makes the POVRay output a no-op, would it be prudent to prompt, warn, or fail? Is anyone else seeing this on semi-recent firmwares? I'll try to find the last image I took that had them and pinpoint the version in which they stopped being recorded soon. I can't seem to find an option in either the smartphone app or the official desktop app to toggle Pitch/Roll recording. |
|
A warning on fail would be nice indeed. I don't have a V to test things on. As you mention, this seems more like a bug in firmware than a "feature". I hope this gets fixed soon as this is rather crucial information for lots of applications I think. Thanks for the additions! |
|
I did some tests with a Theta V running firmware 3.10.1 (released on 2019-10-24). I confirm that It seems to me that now the orientation correction is done in-camera, perhaps before saving the picture to internal storage (so now is I tried taking a few shots with camera upside down or laying on it side, and I always got them rectified straight out the camera regardless if I download the images from the smartphone app or via PTP from my laptop. However, this change is not mentioned in Ricoh changelog. |
|
Ha, interesting. I'll update my S and see what happens. It would make life easy for sure. |
|
Glad to know it's not just me. @gerlos If you have developer mode unlocked, can you get some logcat output when you take a photo? It gives me a lot of errors including a complaint about missing metadata fields. Probably just pieces of the app that were not updated to anticipate in-camera correction, but would be nice to confirm. |
|
I don't have enabled developer mode so far. I'll give a try this week end. In the meanwhile, I posted a question related to this topic on the official Theta Developer community. |
This prevents several lossy conversions in the pipeline and saves out the rectified images with substantially less quality loss. Someday, someone should add options to allow the user finer control over the output level/quality.
Most of these changes were made in June 2018 and I never got around to making a PR for them, so figured I might as well do it now. Changes from upstream have been merged in and commits squashed. See https://github.com/sjuxax/theta_rectify/tree/less-lossy for commit history/details.
Thanks!