I'm dipping my toes into 4k UHDV for scanning purposes. I've only done a couple Hi-Def video scans, while moderately successful. I need, crave, higher resolution textures and meshes that 4K offers. Does anyone with experience in Hi-def video or UHDV have any tips or pointers?
Thanks in advance!
~MD
i could be wrong.
all video.. even hi def has blurred frames in between the sharp ones..
i noticed this since i work with both live action footage and cg (2d or 3d ) animated footage..
in order to make the cg footage match the look of the live action footage.. even 4k.. i end up adding a little blur in post.
simillarly when slowing down footage in post.. video can only go so far.. since it has all those blurred frames.. where as animated renders can be slowed down in post by almost 100% or more(it starts getting choppy.. but still crystal clear).. well unless you rendered with motion blur.
I'd suggest the following:
1) using videos of very slow movement.. to minimize the number of blurred frames.. that your photogrammetry algorithm has to deal with.
2) manually select specific frames that are crystal clear and remove the rest..
3) shoot at higher frame rates ( http://gizmodo.com/why-frame-rate-matters-1675153198 )
4) use some other tool to remove the motion blur from the inbetween frames.. http://www.mecs-press.org/ijmecs/ijmecs-v2-n1/IJMECS-V2-N1-03.pdf (will see if i find a better link)
In my experience, reconstructing from video has more issues with motion blur than resolution.
To reduce motion blur, you can set a faster shutter speed on your camera, or use the "slow motion" mode on an iPhone.
Don't forget that you will need more light to get a correct exposure. You probably don't want to have a larger aperture to keep a good depth of field.
Thanks @chaitanyak!
I believe I understand you: so like still photogrammetry, reducing motion blur and increasing depth of field are critical for getting a good scan.
Thanks for the responses!
Here's another thing about using video that I've been curious about:
Does anyone know if the amount of photos being processed have any relationship to the accuracy and clarity of a scan? For example: I'm currently running (in Agisoft Photoscan) a 4K scan of a statue. I'm using 250 frames from the video to create the scan. If I double the frames used (500 images), will I get more detail and possibly recover missing detail from shaded areas that the video codec compresses?
Yes, if possible better manually chose good sharp frames.
I use AVSPmod and Avisynth it allow fast jump every 1 sec and step forward and backward for find sharpen frame if current blurred.
It depend on your camera movement.
Try to find good gap between frames that have overlap of your scanned object for about 40-60%.
Do not try use all frames. Overlap will be about 99% and this is huge overhead that not give more quality, but only produce hugest dense cloud.
