Need Feedback: Canon Dual Fisheye 3D Stills

VR180 produced by Canon's VR Utility:

VR180 produced by StereoPhotoMaker, with weird exif (its claim is it will look 3D VR180 in proper headset (Oculus, etc.):

Carved out sys of another shot:

Carved-out flat portion of another shot in anaglyph using StereoPhotoMaker:

Do these look correct? Auto alignment was used on both editors.

The original image straight from the camera:

I know tripod foot is in the frame; I was doing this to test the focus alignment of the camera (perfect out of the box). Real shooting will be handheld. I just want to be sure the results look ok.

Outside handheld VR180:

Btw, it is a struggle to keep one's finger out of the frame just pushing the shutter release.

3D Gunner wrote:

It depends on the outcome you expect.
1. From the start, a good 3D image with FOV at 180*, with one camera with two lenses or with two cameras in parallel in one position, is not possible.
The 3D effect is maximum in the center of the image and decreases to "zero" towards the side edges.
2. If a spherical cap is represented on a flat screen, there is also a strange curvature in the vertical plane.
3. For viewing on a flat screen, it is preferable that the image be distortion corrected and cropped so that it remains a "clean" 3D image, with carefully made corrections in terms of vertical alignment and as natural a fit as possible in the stereoscopic window.
4. Carefully selected subjects can be tracked in terms of the distribution of content in space so that the final image is easy to look at and appears natural.
5. Finally, the FOV should be less than 180* horizontally if a convincing result is desired.
6. For correct 3D at 180* or more, you need at least two pairs of images taken from two different angles, then mounted in panoramic format.
7. If the images are to be viewed with VR devices, the situation is similar, somewhat simpler if one neglects the lack of depth in the image composition at the lateral extremes, which can be partially compensated for in the stitching process.

I agree with what 3D Gunner as said.

It is interesting to see the images produced by Canon Dual Fisheye lens.

If you want a 'flat' stereo 3d image then it looks like you need to be some distance from the object you are taking a photo of - see image below - parallel view

- As done in the Street Scene. Then use the 'VR180 produced by Canon's VR Utility' to 'square up' the round image and then crop in quite a bit to the central area to eliminate the distortions in the edges.

3D Gunner wrote:

It depends on the outcome you expect.
1. From the start, a good 3D image with FOV at 180*, with one camera with two lenses or with two cameras in parallel in one position, is not possible.
The 3D effect is maximum in the center of the image and decreases to "zero" towards the side edges.
2. If a spherical cap is represented on a flat screen, there is also a strange curvature in the vertical plane.
3. For viewing on a flat screen, it is preferable that the image be distortion corrected and cropped so that it remains a "clean" 3D image, with carefully made corrections in terms of vertical alignment and as natural a fit as possible in the stereoscopic window.
4. Carefully selected subjects can be tracked in terms of the distribution of content in space so that the final image is easy to look at and appears natural.
5. Finally, the FOV should be less than 180* horizontally if a convincing result is desired.
6. For correct 3D at 180* or more, you need at least two pairs of images taken from two different angles, then mounted in panoramic format.
7. If the images are to be viewed with VR devices, the situation is similar, somewhat simpler if one neglects the lack of depth in the image composition at the lateral extremes, which can be partially compensated for in the stitching process.

So, you did not look at my images, with either a VR headset or 3D glasses to judge whether any of the known criteria are actually met by them?

3D Gunner wrote:

Nice crop, but you set the stereoscopic window to background level, which is not desirable for normal PC screens. For viewing in VR devices it doesn't bother, if you can make convergence adjustments manually.

I did  a quick crop without checking the alignment. Yes the bottom of the image has a window violation when viewed on a PC screen.

3D Gunner wrote:

I have both a large passive 3D screen and a VR device, so I've seen them in 3D very easily.🙂

..........................................

The most successful is the first image, viewed with a dedicated player, with correct settings, in a VR device with head tracking.

But, it's set with the background in the stereoscopic window, the central area is very deep (that's good) and because of this it forces an unnatural convergence of the eyes at the reference settings.

The posted images are for parallel viewing, for that they must be viewed at small sizes (here, on this forum), or they must be loaded into a player that will present them in another format at larger sizes.

They can only be viewed correctly and at their true value by people who know precisely what needs to be done.

Example for passive 3D glasses.

Thanks.

3D Gunner wrote:

How do you feel about that?:

I balanced the depth to be half behind the stereoscopic window and half between the stereoscopic window and the eye.
- 39-cross is for cross-eye view.
-360 is for 360* VR 3D viewing with VR device and appropriate player. In this case the image is quite spectacular with head tracking activated (even if it occupies only half of the sphere)!

It can be "pushed" even further back, at 3/4 back and only 1/4 between the stereoscopic window and the eye.

Very good!