Photoshop perspective correction vs. movements.

[Sylvester Graham]

I've been playing around with photoshop's perspective correction tool (in photoshop CS, I hear they're more advanced in CS3), and was just wondering what the drawbacks/differences are from using a camera with movements in the field?

I'm less concerned about image quality than ability to accurately correct or warp perspective, but if some horrendous degradation of image quality is involved I'll become concerned.


-Alex

[David Luttmann]

The primary drawback to using photoshop controls for perspective, etc, are that they don't fully replace what can be done with a VC....although it comes very close. As well, these adjustments are done through software interpolation instead of natural projection to the film back. Depending on how drastic the adjustments made, the effects of interpolation could be visible.

In truth, I've made software adjustments for prints from a 1Ds Mk2 at 36" and see no ill effects. YMMV.

[David A. Goldfarb]

PS perspective controls are similar to using rear swing and tilt on a view camera or tilting the enlarging easel at the printing stage. It can correct one plane.

So if you are shooting a simple building facade and want to fix the converging verticals, you could do it with Photoshop, and as long as it's not too extreme, the building won't look unattractively foreshortened (also a problem with overcorrection in the field) and there won't be noticeable image degradation.

If it's a complex scene with planes to be corrected at various distances from the camera, then it would be better to correct it in the field using rise and/or shift and the standards parallel to each other and the building. If you try correcting this type of scene with perspective controls in Photoshop (or by tilting the enlarging easel, for that matter), the lines will never look quite right.

[photographs42]

I use perspective control on almost every digital Architectural image. I don’t use the “Perspective” control but I use the “Skew” feature. Skew does the same thing but with more control. One trick that is rather important is to “balance” the correction. In other words, don’t just move the lower corner in or the upper corner out, but move the top out about equal to the amount you move the bottom in.

I only print up to about 10” x 13” but I don’t see any residual evidence of skewing.
Jerome

[65Galaxie]

I tried to adjust perspective with a 6x6 TMax 100 with an Epson 2450 scanner. The picture was of an older church with wood siding. I couldn't get the siding on the church to have a smooth look. I always got some jagged edges. I wonder if it's because of the sharpening done during the scan. I used an early version of Photoshop. I can't help but wonder about the degradation that can occur on bigger enlargements of original dgital files.

[scrichton]

aliasing will always occur with this kind of movement in PS. Mathematically orientating a square pixel off it's original line is not possible without PS generating new "calculated detail" If you can do movements do them. Otherwise scan big then downsample with BiCubic Sharper.

[Leonard Evens]

In principle, you can produce identical results using a photoeditor to what you can get using rise or fall with a view camera. But in practice it may be another matter. It has already been pointed out that the interpolation needed to do it in an editor may reduce resolution further than you would like. Another problem is that, without doing some calculations, it may be difficult to adjust the height of the subject. Photoshop can't figure that out for you just from information present in the image. To get it right, you have to input either the angle you pointed the camera up or some equivalent bit of information, Finally, when you do it with movements, you see immediately on the gg what you are going to get. Doing it afterwards in a photoeditor makes it harder to visualize the final subject when taking the picture and you may be in for some surprises.

If you want to do it accurately, you would be better off using one of the PS plugins based on Panorama Tools, which allows you to input the relevant data.

[Frank Petronio]

It's always best to try to get it as close as possible in camera, but even if you are near perfect in camera, once you overlay the Photoshop grid over the image, you can find adjustments worth doing. Especially straight lines towards the edges of the image.

Best practice is to overscan, then make the move by making the "fat" of the image smaller so Photoshop doesn't have "create" image information. Shrink the fat side rather than expand the thin side... then adjust the image size, round down in size.

And try to do it all in one move rather than a series of tweaks. Use the "Free Transform" tool rather than the more limited rotate, skew, perspective, scale tools.
Doing it all in one move is less degrading to the image than doing a series of moves.

When you want to get really rad, explore the "Liquify" filter ;-)

[Ken Lee]

"...what the drawbacks/differences are from using a camera with movements in the field?"

In this regard, Photoshop is good for repairing an image, but in the field or studio, view camera movements allow you to explore variations of the image and "get it right" the first time - including focus or lack thereof - along multiple planes.

Here's a case where a longer lens was used, and front tilt was required to maintain depth of field. Front drop was used to get more foreground - but the back of the camera remained vertical, to preserve the alignment of the trees. It's not the most compelling example, but having everything in place at the time, allows you to practice the art of... Composition.

Here's an image where the plane of focus is not parallel to the page, and the visual content of the image is dependent on what's in focus, and what's not. I wonder how easy this would have been with a "regular" camera, or in Photoshop. Again, not the most dramatic example, but I doubt I would have ever considered taking the shot, if I were using a more simple camera.

[sanking]

It's always best to try to get it as close as possible in camera, but even if you are near perfect in camera, once you overlay the Photoshop grid over the image, you can find adjustments worth doing. Especially straight lines towards the edges of the image.

Best practice is to overscan, then make the move by making the "fat" of the image smaller so Photoshop doesn't have "create" image information. Shrink the fat side rather than expand the thin side... then adjust the image size, round down in size.

And try to do it all in one move rather than a series of tweaks. Use the "Free Transform" tool rather than the more limited rotate, skew, perspective, scale tools.
Doing it all in one move is less degrading to the image than doing a series of moves.

When you want to get really rad, explore the "Liquify" filter ;-)

My practice is similar to that of Frank. With a view camera I do as much as possible to get the image right when making the negative. However, in the field there are limits to what you can do imposed by extent of movements and lens coverage. To correct the file in Photoshop I always use Distort> Free Transform. It is of course better to have the image file contain slightly more area than necessary at the end because pulling the image to correct distortion can result in some loss. And since I know that the control is possible later I usually sacrifice perspetive to sharpness on the ground glass if the two happen to be in conflict.

So far as I can see this technique does not result in any loss of image quality and when done in moderation does not create artifacts that most people would notice.

This procedure is even more useful with medium format cameras that have no perspective control. I have used it quite a bit with 6X9 negatives taken with my Fuji GW690III.

Sandy King