correcting perspective with software?

on the dpreview website (all digital), a couple of professional

architectural photographers who had switched from LF to digital

systems (canon 1DsmkII) indicated great problems using shift lenses

(specifically the 24mm TSE) due to CA which was difficult to fix in

photoshop. one of the guys said that his testing showed that he

could get much better results using a regular 24mm lens and

correcting perspective using software.

 

i want to shoot 6x7 CTs and scan them. there is a dearth of shift

lenses available for 6x7: the 75mm shift for the pentax67, and the

mamiya 75mm for the rb/rz67. 75mm is just not quite wide enough for

many shots, though i managed to do quite a bit of decent work with

the pentax setup several years ago. but i would prefer perhaps a

55mm or even a 65mm lens on 6x7. i know i can get this capability

if i select one of the monorail-based systems, like fuji 680,

silvestri, horseman, alpa (wow, the 12 swa is a nice lloking

machine, but almost $12K), etc, but i would really like a

handholdable camera so i can use it for some other purposes besides

architecture.

 

so now i am wondering if i could acutally get away with maybe a

mamiya 7II with the 65mm lens, tilt it up slightly to shoot a

building, and then scan the CT and correct perpsective using

photoshop. has anyone tried this? how good are the results? do

you wind up with digital artifacts or other problems? probably not

apparent in small prints, but what about a 20x24" print? thanks.

I will give you my unprofessional experience. I photographed a three story building from directly accross the street with my rolleiflex.

 

I used the PSE correction , and people who saw the photo assumed I used a view camera because the building looked perfectly square, top to bottom.

Not exactly an answer to what you have asked, but have you checked out the Cambo Ultima 35? Basically it's a mini-view camera platform for advanced Canon and other digital SLRs. Look at: http://www.cambo.com/capages/frameset_ult35.html.

Another thing to conisder, if doing it with PS, (or darkroom) perspective correction, is the unavoidable 'crop factor' that will occur at the bottom after you 'widen' the top. You must allow sufficient space outside your subject in the exposed frame to allow you to retain a rectangular image from the resulting 'keystone' image formed by the correction. Easy to do, but you must remember to allow for it at the time of shooting!

John, if you widen the top, the top is cropped automatically in Photoshop, and no width is lost at the bottom. If you narrow the bottom, you must manually crop to preserve the rectangular borders.

 

Simply widening the top (presuming that's the appropriate perspective correction) make the subject look shorter and fatter, unlike using a rising front. I find the proportions look better if I widen the top and narrow the bottom symmetrically, then crop the results.

 

There are no obvious artifacts, other than some loss of detail where the image is resampled and cropped. It's not noticeable at all in an 8x10 made from a 6MP camera.

Using software perspective correction is like tilting the easel on an enlarger or tilting the back on a view camera. It will correct convergence in one plane, and is a handy trick sometimes, but it doesn't always work. The only equivalent to using front rise on a view camera is to use a wide lens, level the camera and crop the excess foreground.

David, how is tilting the back to correct convergence different from using a rising front? To do so, the back would be tilted to be vertical whilst the bed of the camera is pointed upwards. The final touch is to tilt the lens forward (to a vertical position) to adjust the plane of focus. This is SOP to obtain an effective rise greater than the standards will allow.

 

In short, you can correct perspective in Photoshop the same way as with a view camera, except that you cannot correct the tilt of the plane of focus. Given the ample depth of field in miniature lenses, that's not a major concern for convergence issues. If you want toenail to horizon focus, that's a matter for tilting lenses.

Another hint, try to set up your picture so that you will only need to correct the perspective in one "direction" (for lack of proper terminology). I recently took a shot that wasn't lined up well. I then had to correct the perspecitive in more than one direction and this made it much more complicated and limiting.

<p> Doug<p>

<a href="http://home.earthlink.net/~dougfisher/holder/mainintro.html">Dougs

MF Film Holder for batch scanning of 120/220 medium format film with flatbeds</a>

I've used 4x5, MF and digital SLR for architecture and unless you need the size of the bigger cameras I reckon digital perspective correction is fine in most cases. Without a shift lens I try to get as close to the perspective I want using the wide-angle-level-camera technique outlined above then do the rest in PS. I would not use a rangefinder camera for this sort of work as I like to see the dof.

Using PS perspective correction it is a good idea to use a wider than necessary angle lens to allow plenty of room around the edge of the frame for the transformation.

My problem with perspective correction in PS is the amount of time it takes to get it right. If you're used to working with a view camera, you know that adjusting the rear standard for perspective and working the front for focus + either for rise/fall/shift is just a matter of 5 minutes or so. The same corrections in PS take at least 2-10x as long depending upon the complexity of the corrections.

 

Horseman also makes the "View Camera Converter" for 35mm and medium format cameras which may be of some interest to you. They also make the X-Act-D which is a digital 6x6 monorail optical bench system.