A couple more 'Deep Stereo' images using Tri-shots.

From two days ago:- (Corcomroe Abbey, County Clare, Ireland.) - Xeyed

Please refresh our memories of this process?

tony brown wrote:

From two days ago:- (Corcomroe Abbey, County Clare, Ireland.) - Xeyed

Wow! When I remembered to useccrossed gaze the depths in the middle-ground and distance came into their own! Nice work!!

Best t’ye,

Dave

uuglypher
"100% of the shots you don't take don't go in!"
Wayne Gretzky

threed123 wrote:

Please refresh our memories of this process?

Basically, I am taking a standard stereo pair with 90mm lens separation, then taking a further third shot typically 1metre to the right side of the first pair and at the same distant target.

Then this third image is masked, in a layered software like Photodhop, with the second (i.e. right-hand) image. The result uses the normal stereo pair for the foreground to approximately 30 metres and the third image from there to the backgound to extend the discrimination of distant objects.

In my case, the single left image of the standard stereo pair remains unaltered throughout and it is only the two right hand images which are masked together, using whichever is required in different parts of the picture.

Effectively, the standard, close lens stereo provides depth discrimination for several metres and then the remainder of the landscape becomes like a flat theatre backdrop. The approach outined is only applicable to subjects with large depth range. Just using wide spaced lens capture devices for two shots only, creates close foregrounds too separate for the eyes to reconcile them together. (Unlike the stone wall in my first example above.)

Have only been playing with stereo in the last two years and find it interesting, challenging and fun in my late retirement years, albeit 170 years after the introduction of stereo photography.

tony brown wrote:

threed123 wrote:

Please refresh our memories of this process?

Basically, I am taking a standard stereo pair with 90mm lens separation, then taking a further third shot typically 1metre to the right side of the first pair and at the same distant target.

Then this third image is masked, in a layered software like Photodhop, with the second (i.e. right-hand) image. The result uses the normal stereo pair for the foreground to approximately 30 metres and the third image from there to the backgound to extend the discrimination of distant objects.

In my case, the single left image of the standard stereo pair remains unaltered throughout and it is only the two right hand images which are masked together, using whichever is required in different parts of the picture.

Effectively, the standard, close lens stereo provides depth discrimination for several metres and then the remainder of the landscape becomes like a flat theatre backdrop. The approach outined is only applicable to subjects with large depth range. Just using wide spaced lens capture devices for two shots only, creates close foregrounds too separate for the eyes to reconcile them together. (Unlike the stone wall in my first example above.)

Have only been playing with stereo in the last two years and find it interesting, challenging and fun in my late retirement years, albeit 170 years after the introduction of stereo photography.

Interesting I idea. If you were to simply take two images far apart, you would end up with miniaturization, but this way, the 3D image "feels" normal with more depth. I will have to try that, thanks, Tony.

The images seen at small size look very good.
Viewed at medium size, alignment errors already appear in more distant areas.
Viewed on large screens you can see errors everywhere in the far away areas.
For large screens, it certainly looks better to have a single pair with slightly increased inter-axial distance compared to what is calculated to be "normal".

In order to achieve good quality images for the purpose presented here, advanced processing techniques have to be used to obtain a layered structure of the 3D image, with several layers of different 3D perspectives, finally fitted together to obtain an anticipated image depth distribution.

For example, a close-up 3D detail of a subject in the cockpit of a spacecraft (with the correct depth taken as such), on the window of which a distant scene in outer space is visible, where cosmic objects (a galaxy for example) are also distributed in depth.
........
So, you have 3 images from which you can make two stereoscopic pairs with different inter-axial distance.
-from the pair of images with smaller inter-axial distance, you can get a multi-layered depth map.
- from the second pair of images, with larger inter-axial distance, you can obtain another multi-layered depth map.
- In the first multi-layered depth map (with 10 layers for example), you can exchange the last 4 layers (10-9-8-7 for example) with the last 4 layers of the second multi-layered depth map.
- From the new multi-layered depth map you can reconstruct a pair of 3D images with depth emphasized in the remote areas of the image.

Gunner’s suggestions, as usual, make sense, labor intensive though they be.

However, from a practical point-of-view, I have, over the past decade, gained the distinct impression that the great majority of hobbiests’ 3D image pairs are produced and shared on smartphones and digital pads. As such, the majority of the products of Tony’s technique, although certainly more demanding of pp time and effort with depth map construction, would likely be quite happily viewed predominantly via the smaller displays of smartphones and pads.

Tony’s image pairs clearly overcome the single most common complaint levied at traditional 3D stereography - the flat background…kiss of death of intended 3D landscapes.  They do also provide more foreground 3D punch than offered by most 3D conversions accomplished by geometric transformations of the right eye image.

My guess is that if Tony were to put out easily followed instructions for his technique at its present level of development (self-published book or CD?) the great majority of such image pairs would be contentedly viewed on the smaller media where they shine!
Who knows? Perhaps the 30-year cycle of brief resurgences of interest in 3D might smooth out to a consistent continuum of 3D photography?
Just one guy’s attitude, anyway!

Best regards,

Dave

Dave, you are quite right in that I have no desire to output images suitable for 55" screen scrutiny, nor do I have such equipment for examination. Output is to the standard 7" stereo cards and a 24" (diagonal) PC monitor in Xeye display in which each image side is approximately 10" wide.

I am using one of three techniques to extend the rear depth where most appropriate:

1) Masking some of the Right Hand wide shot with the Right Hand narrow shot where required. Easiest but can lead to more abrupt intersections.

2) Using the depth map from the wide shot (created in StereoPhoto Maker) to displace said areas of the narrow shot. This gives far less disruption at the intersections but depends on the outcome of the SPM depth map quality. The depth map can also be emphasised and/or curve adjusted when using this method too. This method also avoids exposure or colour discrepancies where the two Right Hand images differ slightly and is my preferred method when available. I may pursue detail working with DMAG9b, depthmap creation software incorporated in StereoPhotoMaker for that pupose, to improve the frequency of usable results.

3) Hand painting a small section of depth map to emphasise the distant areas. All 6 images in post 'https://www.dpreview.com/forums/post/65785630' were normal 2D images with crudely hand painted depthmaps and displacement filtered to produce the results. (The effects on some Ansel Adams photos is amazing and lifts them to new heights.)

Naturally, the advice to take stereos with 3D Gunner's "slightly increased inter-axial distance compared to what is calculated to be "normal" was the first line of approach, months ago, with the equally 'slight' effect on the landscape backgrounds while hyperstereo was unable to include foregrounds with any validity and thus leading to 'miniaturising'.

I will humbly have to forego the approval of large screen scrutineers but still hold to the maxim that "Distance lends enchantment" in my own small way. Twenty one years of DPReview Forum contribution leads me to expect a broad spectrum of valuable responses.

Each to their own !

There is no reason to be angry. We just discuss and exchange opinions.
Like I said, at small sizes, your 3D images looks great. 🙂

3D Gunner wrote:

There is no reason to be angry. We just discuss and exchange opinions.
Like I said, at small sizes, your 3D images looks great. 🙂

I am in no way angry. I post here expecting criticism and have to cut my ambitions to suit my more limited aims.

I appreciate the truth of your comments but cannot meet such stringent targets and yet see some merit in my activities at the smaller scale. I also enjoy this way of passing the time and have attended the Zoom meetings of the New York Stereo Association for some time with great interest. There is a large, varied and knowledgeable membership therein.

rmexpress22 wrote:

3D Gunner wrote:

The images seen at small size look very good.
Viewed at medium size, alignment errors already appear in more distant areas.
Viewed on large screens you can see errors everywhere in the far away areas.
For large screens, it certainly looks better to have a single pair with slightly increased inter-axial distance compared to what is calculated to be "normal".

Ah, thanks for this. I also couldn't quite focus my eyes with the full image but great effect with the preview sizes.

This example uses the Depthmap obtained from the Wide Tri-shot pair and masked to retain the natural stereo from the Narrow (normal) pair. The resulting map is applied to the Narrow (conventional) stereo pair. The objective is to extend perceived stereo effect much further backwards.

(Xeye)

Kinvara, County Galway, Ireland.

My opinion:
- At this size, 1265 pixels wide/1000 pixels high, the results looks almost OK, with only 5-6 areas with fast visible abnormal displacements.
- After scaling at 3840 pixels wide and viewed on full 55" 3D screen (with proper 3D glasses), at least 21 areas with obvious displacements are visible.

So, if this kind of work is intended to be viewed only at forum level, is OK. If intended to be seen on large screens or on HMDs, is not good.
Standard pairs are the way. 🙂

3D Gunner wrote:

My opinion:
- At this size, 1265 pixels wide/1000 pixels high, the results looks almost OK, with only 5-6 areas with fast visible abnormal displacements.
- After scaling at 3840 pixels wide and viewed on full 55" 3D screen (with proper 3D glasses), at least 21 areas with obvious displacements are visible.

So, if this kind of work is intended to be viewed only at forum level, is OK. If intended to be seen on large screens or on HMDs, is not good.
Standard pairs are the way. 🙂

You have made your opinion perfectly clear already. I no longer need reminding.

Your comments are valuable and I am relieved to have your permission to continue making standard pairs.

Gerry Siegel wrote:

Tony, if these two pairs are what you call deep stereo then I do not know stereo. What process used to take two identical seeming images? A puzzlement as the KIng said.

Hi Gerry,

They are not identical pairs as you will see below. Were you viewing them as CrossEyed stereos as they were posted? It seems to be the preferred method of display on this Forum.

Here are the two pictures posted as Anaglyph; Side by Side; and CrossEyed stereos respectively, for viewing.

The Anaglyph versions readily display the relative differences between the Left and Right images in both cases.

Hope that helps.