Using video for 3D photos

This has been done before, of course, but with 4K cameras, the 3D pics from scanning two consecutive movies frames is getting a lot better. Here are 4 from my Tanzania, African trip. I was touring some cities in Tanzania on my way to the game drives, and grabbed these in a video editing app and passed to Stereomaker, from 4K video made of my road trip along the way. They are hyper-stereo because of the wide separation of frames as we traveled about 30 mph or more. I have lots more...just showing what can be done with a still 4K camera these days. It helps to have bright sun as the individual frames are very sharp then.

All are cross-eyed 3D:

Walking cattle to water

Goats and children--behind fence is a coffee plantation

A street market

Cornfield

Brian F Flint wrote:

You have got very good results.

In the street scene there are people walking. So there would be movement of the subject from one frame to the next. If the 4K video is 30 frames per second then at 30mph you would travel 45 mm so the stereo would be hyper. What surprises me is that the legs on the walking people appear to not have moved significantly in 1/30 second.

The maximum gap in a stereo rig with camcorders is 1/2 frame time, so 1/60s in the case of 30fps shooting.

The attached image is from 4k video pair at 25fps.

For cross eyed viewing

I was lucky on a several that didn't show people movement, but I do have some of those as well. Still viewable, though. Here's one that has a strange anomaly. The person riding the bike appears to be floating off the ground because it's pushed out toward the window. The motorcycle couple are pushed back into the scene. Both bike and motorcycle going in different directions. Van we were in was moving to the right.

cross-eyed view.

This is because your video streams are out of sync, so no proper pairs can be selected.

3D Gunner wrote:

This is because your video streams are out of sync, so no proper pairs can be selected.

This was because the Safari Van I was in was going faster than 30mph this time. The background is okay, though.

Can be a post production sync problem. Anyway, if needs something important to be done (regarding a 3D picture extraction from 3D video), there are solutions to fix this kind of problems.

This kind of problems can be fixed in 3D pairs extracted from video streams, but can not be fixed from a single pair for 3D photos which are out of sync.

3D Gunner wrote:

Can be a post production sync problem. Anyway, if needs something important to be done (regarding a 3D picture extraction from 3D video), there are solutions to fix this kind of problems.

This kind of problems can be fixed in 3D pairs extracted from video streams, but can not be fixed from a single pair for 3D photos which are out of sync.

Yup, I understand that. These were actually consecutive frames, so nothing I can do to get them to sync. I will have to look at the original video, to see what's going on there.

No need to check original videos. You can extract a few consecutive pairs from actual 3D video (7, by example), take one frame as reference (nr.4 for left eye by example), then check adjacent frames for right eye (1,2,3 - 5,6,7) to find a functional pair.

Good observation!
So, if only one camera is used, a good stereo pair cannot be obtained if subjects are also moving in all directions inside the scene.
Even if can be done for special purpose, too much work must to be done to fix all the errors inside the 3D scene.

So, everything I said in the previous posts must be ignored!

Brian F Flint wrote:

As I understand it, only one camera is used in video mode. So two successive frames from the video footage is used to make a 3d image by the CHA-CHA method. This is possible because the one camera is moving sideways in the car.

This 3d image is not made using two cameras ( which are side by side ) and the two video files are put into a video editor and from that a 3d still or 3d video image is produced.

You are correct, one camera--preferably a 4K camera that does video. The higher the resolution, and the brighter the day (increases the shutter speed),  you select two frames next two each other. The faster the car is going, the more depth you will have in the image. If the speed is too great, you will get miniaturization of the objects in the 3D image as you can see in my 3D images. It's fun to try, but you have to have a video editing program that allows you to select frames and save them as jpgs. Then you use Stereomaker to align them and create an MPO file for digital viewing or SBS for free viewing.

3D Gunner wrote:

Good observation!
So, if only one camera is used, a good stereo pair cannot be obtained if subjects are also moving in all directions inside the scene.
Even if can be done for special purpose, too much work must to be done to fix all the errors inside the 3D scene.

So, everything I said in the previous posts must be ignored!

Yes, that is correct, I was using one camera. A 3D camera would have a wonky 3D image--probably smeared looking. I had my Panasonic z100 4K camera using the 4K video and held the camera on the edge of an open window so the movement would be steady. Stereomaker does a great job of alignment even with slight rotation of the images.