Replacing T&S lenses w/ digital tecniques

[Tim Gray]

My personal experience is that digital blending to enhance DOF is a bit tricky.  The problem is that when you change the focus, you also seem to change the focal length slightly - at least enough so that the two shots don't align perfectly - you have to resize one of them and it's hard to get a perfect match by trial and error.  I would guess that there is a formula depending on the original FL and first focus point and the second focus point that would tell you the % difference - but it might also be dependent on the lens, brand etc.  If you have a natural boundary between the foreground and background it's a bit simpler.  BTW it was on my agenda this week to give this idea some more effort...

[ik4jqw]

I gave it a try today. The stitching option seems to work (though with tall objects in the foreground would have been tougher). Meaningless subject of course but was only for the sake of testing this tecnique.
Lense used Nikkor 18-35 3.5/4.5 AFD @18mm

[russell a]

Oh, in the light of morning, rather than the middle of the night when I posted above, it occurs to me that my technique was perhaps a little different.  I took the same exact shots from a tripod with only the focus changed.  I layered one shot over the other (let's say the near-field on top of the far field) and made the unison adjustments. Then I created a horizontal crooked selection and completed the selection around the sides and top edges.  It may be necessary, at this point, to add to the selection with the non-magnetic polygon selection tool to catch any places where the straight edges aren't straight.  Finally, I just cleared the far-field area from the near field shot and merged the layers.  If this was obvious I apolgize for belaboring the point, but I would rather err on the side of clarity.

[dbarthel]

Also consider exploiting T/S lenses with digital techniques. Using a T/S lens and shifting left to right with multiple exposures gets you the digital equivilant of an XPan. Since  nothing moves on the camera except the lens shift, stiching is a slam dunk.

Dan

[Tim Gray]

Here is an interesting technique.  When I tried this on an indoor still life it worked... sort of...  I ended up with foreground in focus and the far background in focus but the mid range was still out of focus.  Presumably I could adjust the focus points to get better coverage.  Was still difficult to get a perfect overlay of one shot with the other since the FL seemed to shift slightly with a change in focus.  (Manual focus on a tripod).  

http://www.sgi.com/grafica/depth/

Also tried a more conventional simple foreground/background layer merge on a landscape.  What worked most effectively was a "find edges" filter then (I forget which) either an expand or contract selection by 2 pixels - save the selection, copy one shot to the other, reload the selection then invert/delete as appropriate.

The problem with simply merging a "top half" with a "bottom half"  occurs when you have a foreground object (eg a tree) that extends from the bottom of the frame to the top.

[ik4jqw]

I just wonder if someone is already doing that.
Two possible scenarios:
A background at infinity with one simple object in the foreground: I was thinking of taking two shots focused differently and then blending them together. I can't see why it shouldn't work: any comments?
Another scenario: shots like Michael's Yellow Stripe at Monument Valley. The idea would be to take two horizontal shots and stitch them vertically w/ the upper one focused at infinity and the lower focused on the foreground. I just wonder if the labour needed to accomplish that is so huge to justify the use of T&S lenses or view cameras.
Any comment greatly appreciated.
Thanks

[Dean]

I had the same thought about using different focus points.  Instead of touching my lens, I just changed the focus point on my camera. I shot just two examples using film and haven't tried to blend them yet.  One sample, however, looks promising. It had a very well defined border between the near and far object.  The other sample looks like it could be a problem to blend.  That sample had a bright yellow, thin pole in the foreground.  On the frame where the pole is out of focus, the pole is of course blurred and appears to cover a larger area than when the pole was in focus. Perhaps this problem occurs in the other sample but it is just not as obvious.

[russell a]

I did a few of these a while back with a 4MP Sony.  It worked pretty well.  I brought the separate full frame images into Photoshop and applied any modifications equally (Levels, etc.) to each layer.  I used the Photoshop magnetic tool to create a crooked separation line, my theory being that it would work around the edge of micro features and not create harsh disjunctures.  An analogous selection method might work to select around a near field feature that stuck up into the far field area.

http://www.russarmstrong.com/archive....d1.html

[Dean]

My test photos were a little different than the other posts.  I tried to use various focus points to do what even a T/S lens cannot do: maintain clear focus from near to far when one object is directly in front of the other at the same level.  For example, a photo looking through a gap in a flowery bush, where the bush fills the entire frame, but another object is seen through the gap in the distance.  I don't have the requisite equipment to complete my test, but I would be interested if anyone else has tried a similar test.

[Edward]

What about shifting while tilted?  I can visualize this with a sideways shift while tilted forward to get a near-far composition in focus - an Xplan with view camera movements.  I cannot visualize shifting up and down for a near-far, however. (I do not have a TS lens so I cannot just try it.)

[dbarthel]

That, indeed is the way the T/S comes with the movements on oposite axis. So what you suggest works nicely. For the 10D and D60, 3 exposures give lots of overlap for the stiching software to work.

Dan