I know this forum is mostly aimed at still photography, but I was hoping that there would be sufficient interest in video to help out with Gyroflow (GPL).
Specifically: there appears to be metadata inside the video file on a frame-by-frame basis that describes the distortion, metadata that is used by Sony’s proprietary Catalyst Browse software to perform impressive stabilization in post.
I’m an Android app developer by trade, so optical DSP is outside my skillset, but I’m also an avid hobbyist who would love this functionality to be in the libre Gyroflow rather than just the proprietary Catalyst Browse.
I saw that @Entropy512 had already taken a stab at it, but I’ve gathered that he’s a busy guy! I believe in this enough to put my money where my mouth is: if someone wants to take this on as a freelance contract, PM me and we’ll hammer out the details.
Yes, I figured that would grant us a starting point. However, like I said, image DSP is outside my area of expertise, so I’m back to feeling like a college freshman here. I did see in the GitHub thread that @Entropy512 said the metadata in the video frames didn’t precisely match up with the expected format of the still image lens distortion values. One would assume that Sony would re-use the same lens distortion formula with the same coefficients but it would seem that’s not the case…
Hey, sorry for not responding to your email… I’ve got quite a few things I need to catch up on.
In general, as far as money goes, my job is more than enough there, so for the most part it’s “whenever I get around to it”.
I did have a few discoveries I need to drop onto that issue description, and get back into messing with gyroflow. I’ve chatted with Adrian a few times, and he’s actually gotten some really useful info out of the Catalyst app (e.g. pulling shader source from somewhere). Some of that info he found was very useful for making progress.
@kmilos unfortunately, the way Sony stores distortion info in their EXIF for stills is very different from what they store for video. I’ve made some progress in getting somewhat close to being able to map from one to the other, but it’s just “somewhat close”. I keep on procrastinating on at least posting my current analysis script to a gist, along with recent plots to that issue. (I DID post a plot to Adrian on Discord a while ago…)