Re: [PanoToolsNG] Re: Stereo-3D Anaglyph Panoramas
- On Fri, 01 May 2009 16:04:26 +0900, Wim Koornneef
>> You mean you used two different cameras and lenses for the stereoWell yes, I promise to be surprised! This ignores the effects of quite
>> pair? I have two cameras to use with my two different lenses, but
>> thought this would be unusable for stereo panoramas.
> At first I thought this too, people always talk about the need of two
> equal setups but I had doubt if this was true so I did a test and I
> found out that it is possible to work with 2 different cameras with
> different lenses !
> Test it for yourself:
> Take a very sturdy tripod and a precise panohead, take a pano with each
> camera+lens combination of a static and controlled scene (same light, no
> movements in the scene etc.) and take more images then usually to get
> approx. 50% overlap ( I shot 8 images with each camera+lens combination)
> Be absolutely sure that you don't move the tripod when changing cameras
> and make sure that each camera is mounted in the proper NPP. (I used a
> NodalNinja5 panohead with stoppers on the swingarm rail and my wife
> Margriet assisted me to hold the tripod in place).
> Then stitch a totally and carefully optimized pano of each set of source
> images with the same equirectangular size.
> Put them in Photoshop in layers and make a comparison at 100% pixel
> I promise you will be supprised by the small pixel shift between both
> panos !
different second-order lens distortion, so I suppose this must be
negligable. Surprising! Although come to think of it, I used to shoot
with both eyes open, one eye looking through the SLR with the zoom
set to 1:1 and the other looking at the scene itself. This was a great
help in framing for someone used to a bright-line finder, as I was,
and the bonus was that I saw everything in 3D. I'll bet there was more
difference between the two images my eyes were seeing than between
two panoramas properly optimized.
> The maximum shift between the panos I noted in my test was just 2I am encouraged. It means I can try taking some myself.
> Keep in mind that the left and right pixel shift in the anaglyphs (or
> the colorcode3D) images is far more then a few pixels and that explains
> why it is possible to use 2 different cameras with different lenses.
> BTW, the method for using 2 cameras with different lenses is copyright<Wry but appreciative grin!>
> 2009 - Wim Koornneef - dmmdh productions !!
> No just kidding, take your advantage and please feel free use the method
>> Using a bar with two cameras necessarily means that both cannot beThis is depressing. I moved to fisheyes because I hate stitching
>> revolving about their non-parallax points. Do you find that greatly
>> affects stitching, generating errors in the final panorama?
> Yes, the out of NPP shooting gives a lot of problem when shooting and
> stitching the "normal" way.
> The solution is to shoot a lot more images then usual, how many "more"
> depends on the distance to nearby objects.
> When the distance is at least 1.5 meter 25 images for each camera will
> do but when the distance is less then you have to take much more images.
> In one of my tests I had to shoot 72 images to avoid errors close to
> nadir but in this case the ceiling was only 70cm away from the camera....
> In practice I shoot 50 images around and use only 25 of them, when
> needed (to much errors in one or more seems) I put one or more extra
> images in the row to solve the error.
> To output the panos as layers is the second best option that I like
> avoid because it is a hell of a job to retouche errors in 2 panos in
> such a way that there is no visible trace of it in the 3D output.
multiple images, and four-around or six-around is about my limit. But
I can believe you about the pain of retouching two panos for stereo
vision. No thank you!
I seem to remember something said about taking many photos on the way
around the 360 degrees and using only a narrow, central, "slice" of
each to make stitching more successful. And I suppose that a rotating
slit camera (I have one, and used it exclusively for 360-degree
panoramas until I learned to stitch them) is the ultimate in this
respect. There might be a way to use a slow-panning movie to
create a good-looking panorama even if it was not actually rotating
around the NPP. The image might look quite good but of course the
parallax effects would introduce some distortion into the things at
first partially hidden and then gradually revealed as the camera
turned. I can imagine it, but I'd like to see how it looks... And
you would still need two cameras locked and turning together. You
couldn't have the same camera turning about two different centers
of rotation--the left and right images would be interchanged at
the 180-degree point! Ha, ha!!