Meant to be Seen

Hi Everyone,

I did a search over the internet but I didn't find anything about this. I want to know if it´s possible to do a 3D 360 degree video, where the viewer can walk around the scene.
I thought filming using some cameras around a place (Just like was filmed matrix bullet scene), then putting all videos together to make this immersive 3d 360 video.

Just like a videogame scene, where you can walk around, but I like to do this with a real video.

Do you more information?

Mohaa wrote:Hi Everyone,

I did a search over the internet but I didn't find anything about this. I want to know if it´s possible to do a 3D 360 degree video, where the viewer can walk around the scene.
I thought filming using some cameras around a place (Just like was filmed matrix bullet scene), then putting all videos together to make this immersive 3d 360 video.

Just like a videogame scene, where you can walk around, but I like to do this with a real video.

Do you more information?

You would need accurate depth mapping onto which to paint the 3D interpolated scenery from all the (hidden) cameras for the entire navigable space. Much easier to approximate this with motion capture and photographic textures and realistic CGI, but doable (just not easy or cheap).

I was thinking on something like this http://www.fastcompany.com/3042958/nbc- ... ch-cameras
Maybe a 180 degree video to start.

I was thinking in a way to not use CGI.

Regards,
www.mundovr.com.br

Mohaa wrote:I was thinking on something like this http://www.fastcompany.com/3042958/nbc- ... ch-cameras
Maybe a 180 degree video to start.

I was thinking in a way to not use CGI.

Regards,
http://www.mundovr.com.br

You can warp and interpolate multiple cameras, and even recover extra resolution from overlapping coverage. The complication is how to splice or merge the overlapping areas from multiple cameras without introducing depth-related splicing seam artifacts, which generally requires depth information. Stereoscopic viewing not only requires more accurate depth information, but also that image warping and registration not cause any vertical parallax, which can be quite uncomfortable to the viewer. Using depth cameras (structured light or time-of-flight) cameras can help reduce required post-processing computing power requirements.

That 180-degree video you linked looks like a "Bullet Time" camera setup from "The Matrix". A consecutive pair of cameras could provide stereoscopic 3D, but that limits your viewpoint and orientation.

The two main methods for practical panoramic light field recording at the moment are plenoptic cameras like Lytro and Raytrix along with cross-eyed pairs or triplets like "Camlot A" UCSD/UIC research by Harlyn Baker, Dan Sandin, Greg Dawe and Tom Defanti, separately by Samsung Project Beyond (thinktankteam R&D division) and also separately Jaunt VR.

Any single module only records 2.5D, the ".5" being the recorded distance to whatever colour. With a single module, intuitively I say that you can only reliably navigate without seeing blank spaces within the FoV of cross-eye pairs or where a point is recorded by more than one sensor. In the case of the cross-eyed pairs, I think the area is within the circumference of the physical camera module.

You can end up with a pointcloud and to then build up an entire world pointcloud you use multiple light field cameras and use common points or known sensor locations to co-register the point clouds. Doing CGI or other effects is then essentially the same as making a 3D scene in Blender or another program.

However I think it would still be very compelling to just have a mixture of fly on the wall, disco ball and first person views instead.

This exactly what I was looking for http://www.engadget.com/2015/07/29/holo ... o-capture/

Do you know which program they use?