Pages: [1] 2   Go Down

Author Topic: Stitching and "depth merge" vs. Tilt and Shift  (Read 8612 times)

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Stitching and "depth merge" vs. Tilt and Shift
« on: December 20, 2008, 03:19:04 am »

Hi,

There have been some discussions recently about the availability of TS (Tilt and Shift) lenses for the Sonya Alpha 900. At this time I'm not really sure that TS lenses are really needed for perspective correction. The reasoning I have may of course be right or wrong. Here are my suggestions:

1) Perspective correction is very well possible in Photoshop (or whatever program we use).

2) A better alternative may be to stitch multiple photos. Resolution is essentially increased by 50% when stitching as you normally turn the camera along the short side. Modern programs like Autopano Pro using the "Smart Blend" rendering option reduce the need to rotate the camera around the nodal point of the lens, but even with a nodal slider a simple "pano head" and slider fits both your pocket and your wallet, naturally assuming a voluminous pocket and wallet.

An additional advantage with stitching is that you can use a wide range of zoom lenses and that only the central part of the lens will be utilized.

The "Scheimpflug principle" cannot be duplicated the same way but several pictures having different focus can be merged into a single picture using "depth merge". I have tested this very little using "Helicon Filter" with less then impressive results but I would expect that it is workable with some practice.

I would be interested on your take on these issues.

Erik
« Last Edit: December 20, 2008, 03:20:08 am by ErikKaffehr »
Logged
Erik Kaffehr
 

Paul Roark

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 398
Stitching and "depth merge" vs. Tilt and Shift
« Reply #1 on: December 20, 2008, 11:35:43 am »

Quote from: ErikKaffehr
...

The "Scheimpflug principle" cannot be duplicated the same way but several pictures having different focus can be merged into a single picture using "depth merge". I have tested this very little using "Helicon Filter" with less then impressive results but I would expect that it is workable with some practice.

...
Erik
Logged

Paul Roark

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 398
Stitching and "depth merge" vs. Tilt and Shift
« Reply #2 on: December 20, 2008, 11:51:43 am »

For many years the Rollei SL66 was my main camera for landscape shooting due, in part, to the tilt.  For landscape work it's very useful.  However, with the advent of digital, I found that with my Bronica RF 645 and its 45mm wide angle lens I could combine a couple zone-focused frames and achieve most of what the Rollei could achieve with the tilt and the 50mm Distagon.  Moreover, I could do away with the defects of that retro-focus lens, which was really not very suitable for tilting.  But with other than wide angle lenses, duplicating what a tilt can do is very difficult.

So, moving to 35 mm DSLR and having been a Canon SLR user for years, I'm going to use the 90 TS (a terrific optic) with the 5D2 and stay with prime, non-TS wide angles.  I do have the 45 mm TS, but sent back the 24 TS due to it's lack of sharpness.  (I miss the old FD 35mm TS.)  The prime wide angle lenses are far sharper than the 24 TS, though I'm not all that thrilled with any retrofocus lens.

Paul
www.PaulRoark.com
Logged

Tony Beach

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 452
    • http://imageevent.com/tonybeach/twelveimages
Stitching and "depth merge" vs. Tilt and Shift
« Reply #3 on: December 20, 2008, 12:12:48 pm »

Quote from: ErikKaffehr
There have been some discussions recently about the availability of TS (Tilt and Shift) lenses for the Sonya Alpha 900.
Hartblei has TS lenses now -- while they are expensive, they do offer more functionality than Nikon and Canon alternatives and their optics seem really good.

Quote
At this time I'm not really sure that TS lenses are really needed for perspective correction.

The reasoning I have may of course be right or wrong. Here are my suggestions:

1) Perspective correction is very well possible in Photoshop (or whatever program we use).

2) A better alternative may be to stitch multiple photos. Resolution is essentially increased by 50% when stitching as you normally turn the camera along the short side. Modern programs like Autopano Pro using the "Smart Blend" rendering option reduce the need to rotate the camera around the nodal point of the lens, but even with a nodal slider a simple "pano head" and slider fits both your pocket and your wallet, naturally assuming a voluminous pocket and wallet.
All of this takes time, both in the field and in front of the computer, and that's a trade-off.  More importantly for me is that if I'm correcting perspective because I want to get close to the ground, then I'm compounding my efforts as I will have to also do focus stacking.

Quote
The "Scheimpflug principle" cannot be duplicated the same way but several pictures having different focus can be merged into a single picture using "depth merge". I have tested this very little using "Helicon Filter" with less then impressive results but I would expect that it is workable with some practice.
There's that time consideration thing again.  Indeed, some effects cannot be accomplished with a simple tilt and require focus stacking; however, sometimes you can make this task easier by using a tilt capable lens to get more of the image in focus so you don't have to stack as many images to get everything equally in focus.  Anyway, as you say there are things you can't do with focus stacking, such as:



This shot was of a fleeting moment and the selective focus would be hard to attain without the tilt function.
Logged

Pete Ferling

  • Guest
Stitching and "depth merge" vs. Tilt and Shift
« Reply #4 on: December 20, 2008, 09:04:01 pm »

I have an FD35 TS that was retooled for EOS.  Yes, I have fixed perspective correction in post.  However, after even ten images it starts to eat up my time (anything I can do to limit round tripping from LR to PS is a benefit).

There's nothing like composing the look and correction on the spot before taking the shot.

I especially like the 35mm as it's sharp and the range works well on cropped sensors, being more normal rather than telephoto as the EF 45mm would be.
It fits right at home on the 1Ds and I find myself saving a few stops to get the DOF I need without hitting defraction limits.  Before, I would shoot F11 to F16 and then tweak in post.  With tilt, I'm back to F5.6 - 8 and closer to the sweet spot of the lens.  This matters to me as I do print product shots on very large banners.
Logged

JeffKohn

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1668
    • http://jeffk-photo.typepad.com
Stitching and "depth merge" vs. Tilt and Shift
« Reply #5 on: December 21, 2008, 02:19:21 am »

Quote from: ErikKaffehr
1) Perspective correction is very well possible in Photoshop (or whatever program we use).
Perspective correction in Photoshop degrades image quality, it can also make compsosing more difficult and means wasting resolution because you have to frame your shots with extra space to allow for the post-correction crop. While this might be better than not having the desired perspective, IMHO it's no substitute for getting the perspective you want in-camera.

Quote
2) A better alternative may be to stitch multiple photos. Resolution is essentially increased by 50% when stitching as you normally turn the camera along the short side. Modern programs like Autopano Pro using the "Smart Blend" rendering option reduce the need to rotate the camera around the nodal point of the lens, but even with a nodal slider a simple "pano head" and slider fits both your pocket and your wallet, naturally assuming a voluminous pocket and wallet.

An additional advantage with stitching is that you can use a wide range of zoom lenses and that only the central part of the lens will be utilized.
Stitching is no panacea. For one thing there's always the concern of time, and the fact that your scene may not be completely static due to changing light, blowing wind, or moving subjects. There's also a greater margin for error - there's nothing quite like getting home and realizing that you missed the overlap on one shot and ruined the whole image. I've also found that at wider the angles of view, the results of stitching look different than if you just shot the image with a wide rectilinear lens. The projections performed by apps like PTGui and AutoPano pro seem to over-compensate and cause distortions. Finally, you lose the ability to see your composition in the viewfinder.

Quote
The "Scheimpflug principle" cannot be duplicated the same way but several pictures having different focus can be merged into a single picture using "depth merge". I have tested this very little using "Helicon Filter" with less then impressive results but I would expect that it is workable with some practice.
As far as DOF stacking, I've found the results to be less than satisfactory with today's tools. For some images it works OK, for others it doesn't do so well and you end up with artifacts or uneven focus.

Stitching has its place, I use it quite often, but in some situations it's best to get the image in a single shot and T/S are one more tool that can make that possible. And T/S lenses are also often quite good for stitching; in their centered position corner sharpness and light falloff tend to be excellent due to the larger image circle. Plus, you have additional options such as using shift for a "flat" stitch, or rotating the camera in the traditional manner while using tilts to control the DOF (only works for single-row panos though).
« Last Edit: December 21, 2008, 02:23:04 am by JeffKohn »
Logged
Jeff Kohn
[url=http://ww

Ben Rubinstein

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1822
Stitching and "depth merge" vs. Tilt and Shift
« Reply #6 on: December 21, 2008, 08:55:44 am »

Stitching doesn't help for when you need a one shot solution, i.e. when you have moving foliage or more importantly moving people! I stitch a lot and it needs careful planning to include people, i.e. you need them to be around the middle of a single frame so that you can freeze their movement without them being on a stitching overlap that will then disappear duing the stitching process. Changing light is less of a problem unless you're shooting many frames but you can adjust the shutter speed slightly between frames to compensate, I've done this often.

Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Stitching and "depth merge" vs. Tilt and Shift
« Reply #7 on: December 21, 2008, 11:18:53 am »

Hi,

I really feel that LR should have at least rudimentary controls for perspective and distorsion, especially as they are "a natural" for parametric editing.

Thanks for good comments!

Erik

Quote from: Pete Ferling
I have an FD35 TS that was retooled for EOS.  Yes, I have fixed perspective correction in post.  However, after even ten images it starts to eat up my time (anything I can do to limit round tripping from LR to PS is a benefit).

There's nothing like composing the look and correction on the spot before taking the shot.

I especially like the 35mm as it's sharp and the range works well on cropped sensors, being more normal rather than telephoto as the EF 45mm would be.
It fits right at home on the 1Ds and I find myself saving a few stops to get the DOF I need without hitting defraction limits.  Before, I would shoot F11 to F16 and then tweak in post.  With tilt, I'm back to F5.6 - 8 and closer to the sweet spot of the lens.  This matters to me as I do print product shots on very large banners.
Logged
Erik Kaffehr
 

MarkL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 475
Stitching and "depth merge" vs. Tilt and Shift
« Reply #8 on: December 22, 2008, 08:02:35 am »

I have found the perspective correction in PS is often difficult and degrades the image, I'd much prefer to use movements and get it right in camera.

As for focus blending, I use helicon focus and found it to be an excellent solution as long as you shoot enough frames and things are not moving. This gets around the problem of tall trees in the frame or issues with larger objects near the camera with lens tilt.
Logged

jani

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1624
    • Øyet
Stitching and "depth merge" vs. Tilt and Shift
« Reply #9 on: December 27, 2008, 07:40:20 pm »

Quote from: ErikKaffehr
I really feel that LR should have at least rudimentary controls for perspective and distorsion, especially as they are "a natural" for parametric editing.
As others have noted, this is a lossy process in Photoshop, and for good reasons!

When you correct your perspective in post-processing, you're stretching and/or compressing parts of the image. This means that some parts of the image will have fewer pixels left than you started out with. You're effectively performing a non-rectangular crop of your image.

This is most emphatically not "a natural" for parametric editing, quite the opposite.
Logged
Jan

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Stitching and "depth merge" vs. Tilt and Shift
« Reply #10 on: December 27, 2008, 11:43:33 pm »

Hi Jani,

I'm aware that stretching is a lossy process. The reason i feel "it's a natural for parametric edits" is that parametric edits are just transformation you apply to an unchanged original. Ideally I would prefer to work in Lightroom and avoid creating TIFF and PDS versions unnecessarily.

Another issue is that I feel that Lightroom needs is distortion correction. Distortion correction often uses the same algorithms as needed for perspective control.

With the arrival of affordable 20+ MPixel cameras the sensors may very well "outresolve" the lenses. Some of the Canons TS lenses are regarded as not very sharp. It is quite probable that much resolution is lost when a shift lens is used with shifts on a very high res camera.

The other issue I raised was the option to use stitching. With stitching we actually increase the resolution with a factor of about two, in terms of megapixels. The reason is that we normally stitch horizontal panos with the camera in the vertical position and vertical panos in the horizontal position. At least with Autopano Pro you can mark parallel structures that should stay vertical.

The reason I started this discussion is that I see these methods as economical alternatives to expensive TS-lenses. The discussion resulted in a lot of good input with photographers having insight in both techniques explaining the advantages of using TS-lenses.

Best regards
Erik

Quote from: jani
As others have noted, this is a lossy process in Photoshop, and for good reasons!

When you correct your perspective in post-processing, you're stretching and/or compressing parts of the image. This means that some parts of the image will have fewer pixels left than you started out with. You're effectively performing a non-rectangular crop of your image.

This is most emphatically not "a natural" for parametric editing, quite the opposite.
Logged
Erik Kaffehr
 

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Stitching and "depth merge" vs. Tilt and Shift
« Reply #11 on: December 27, 2008, 11:52:51 pm »

Hi,

Autopano Pro with "Smart blend" is actually very good on merging moving foliage. Moving people is a problem, tough. I would not suggest that stitching is a full replacement for TS-lenses, but I see it as an economical alternative technique. It can also be used with any lens.

Best regards
Erik


Quote from: pom
Stitching doesn't help for when you need a one shot solution, i.e. when you have moving foliage or more importantly moving people! I stitch a lot and it needs careful planning to include people, i.e. you need them to be around the middle of a single frame so that you can freeze their movement without them being on a stitching overlap that will then disappear duing the stitching process. Changing light is less of a problem unless you're shooting many frames but you can adjust the shutter speed slightly between frames to compensate, I've done this often.

Logged
Erik Kaffehr
 

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Stitching and "depth merge" vs. Tilt and Shift
« Reply #12 on: December 27, 2008, 11:59:45 pm »

Hi Tony,

A very good example of using a tilt lens! Thanks!

Erik

Quote from: Tony Beach
Hartblei has TS lenses now -- while they are expensive, they do offer more functionality than Nikon and Canon alternatives and their optics seem really good.


All of this takes time, both in the field and in front of the computer, and that's a trade-off.  More importantly for me is that if I'm correcting perspective because I want to get close to the ground, then I'm compounding my efforts as I will have to also do focus stacking.


There's that time consideration thing again.  Indeed, some effects cannot be accomplished with a simple tilt and require focus stacking; however, sometimes you can make this task easier by using a tilt capable lens to get more of the image in focus so you don't have to stack as many images to get everything equally in focus.  Anyway, as you say there are things you can't do with focus stacking, such as:



This shot was of a fleeting moment and the selective focus would be hard to attain without the tilt function.
Logged
Erik Kaffehr
 

Ben Rubinstein

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1822
Stitching and "depth merge" vs. Tilt and Shift
« Reply #13 on: December 28, 2008, 05:25:29 am »

Quote from: ErikKaffehr
Hi,

Autopano Pro with "Smart blend" is actually very good on merging moving foliage. Moving people is a problem, tough. I would not suggest that stitching is a full replacement for TS-lenses, but I see it as an economical alternative technique. It can also be used with any lens.

Best regards
Erik

I use Autopano, it is actually remarkably good at moving foliage, however when you want a one shot solution, when you need people or cars or whatever, stitching is still not a solution. Far more to the point with focus blending where anything moving is a problem. I actually use a method of focusing in tiny slices then only stitching the sharp areas, works well in a world using longer lenses for more natural perspectives but it is rather complicated. The above image had 3 points of focus, two 'slices' for the foreground and the rest at f22, gives be DOF back to front using a 100mm lens that shouldn't be able to hold that much in focus.
« Last Edit: December 28, 2008, 05:29:53 am by pom »
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Stitching and "depth merge" vs. Tilt and Shift
« Reply #14 on: December 28, 2008, 08:56:29 am »

That's part of the game, with a great picture you see the image but not all the effort behind. BTW it's a really nice image!

Best regards
Erik


Quote from: pom
I use Autopano, it is actually remarkably good at moving foliage, however when you want a one shot solution, when you need people or cars or whatever, stitching is still not a solution. Far more to the point with focus blending where anything moving is a problem. I actually use a method of focusing in tiny slices then only stitching the sharp areas, works well in a world using longer lenses for more natural perspectives but it is rather complicated. The above image had 3 points of focus, two 'slices' for the foreground and the rest at f22, gives be DOF back to front using a 100mm lens that shouldn't be able to hold that much in focus.
Logged
Erik Kaffehr
 

DanPBrown

  • Full Member
  • ***
  • Offline Offline
  • Posts: 138
    • http://danbrownphotography.com/
Stitching and "depth merge" vs. Tilt and Shift
« Reply #15 on: December 30, 2008, 05:02:33 pm »

With all this talk of focus blending and panoramas I thought I would share my greatest effort photo. This photo is made up of 151 images that were focus blended and stitched. The final image is 61 megapixels. It took me over 5 months and 50 hours to manually assemble all the shots. I consider it one of my best photos but it doesn't seem to get much attention. That is one of the problems of putting that much effort into a photo, it may end up being just an average shot.
Dan
http://www.danbrownphotography.com
« Last Edit: December 30, 2008, 06:00:05 pm by DanPBrown »
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Stitching and "depth merge" vs. Tilt and Shift
« Reply #16 on: December 30, 2008, 05:07:11 pm »

Hi!

Problem is that when you see the shot you say it's a nice sharp shot. You don't say, incredible, how did he achieve this? We take perfection for granted.

Erik

Quote from: DanPBrown
With all this talk of focus blending and panoramas I thought I would share my greatest effort photo. This photo is made up of 151 images that were focus blended and stitched. The final image is 61 megapixels. It took me over 5 months and 50 hours to manually assemble all the shots. I consider it one of my best photos but it doesn't seem to get much attention. That is one of the problems of putting that much effort into a photo, it may end up being just and average shot.
Dan
http://www.danbrownphotography.com
Logged
Erik Kaffehr
 

markhout

  • Full Member
  • ***
  • Offline Offline
  • Posts: 176
    • http://www.markhout.com
Stitching and "depth merge" vs. Tilt and Shift
« Reply #17 on: December 30, 2008, 05:08:51 pm »

Quote from: DanPBrown
That is one of the problems of putting that much effort into a photo, it may end up being just and average shot.

I think it is a beautiful and serene shot - the quote above is a keeper, and I admire your honesty!

Mark
Logged

Tony Beach

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 452
    • http://imageevent.com/tonybeach/twelveimages
Stitching and "depth merge" vs. Tilt and Shift
« Reply #18 on: December 31, 2008, 12:52:15 am »

Quote from: DanPBrown
The final image is 61 megapixels.

That is one of the problems of putting that much effort into a photo, it may end up being just an average shot.

I imagine the image's impact would be much greater if if were printed large enough to cover the wall of a typical home -- after all, that would be the purpose of all that effort and certainly not a small print that could otherwise be achieved with a P&S.
Logged

DanPBrown

  • Full Member
  • ***
  • Offline Offline
  • Posts: 138
    • http://danbrownphotography.com/
Stitching and "depth merge" vs. Tilt and Shift
« Reply #19 on: December 31, 2008, 03:56:06 am »

Quote from: Tony Beach
I imagine the image's impact would be much greater if if were printed large enough to cover the wall of a typical home -- after all, that would be the purpose of all that effort and certainly not a small print that could otherwise be achieved with a P&S.
Funny you should say that. I belong to an artist association. Every year we have a juried show in our state's largest art museum. This year we were allowed to submit two pieces of work. I printed the iris shot 20"x30" on Epson Exhibition Fiber paper and framed it with AR glass in a 29"x39" frame. I thought it was quite a striking photo, the dpi is over 300 at that size. Well it didn't make it, but at least this photo did.
Dan
www.danbrownphotography.com
Logged
Pages: [1] 2   Go Up