A keystone function (vertical and horizontal adjustments simultaneously) that works
Implemented
Please introduce a keystone function that works not only with vertical or horizontal adjustments individually but also with simultaneous vertical and horizontal adjustments.
I've written about keystone issues before, here for instance, as have other people. Capture One support has confirmed that the current keystone function doesn't work as it's supposed to.
(There's been much discussion of the use of keystone adjustments; it's not my intention with this request to start such discussions again, merely to request that Capture One gets a properly working keystone adjustment function.)
-
I agree that they keystone tool isn't working as one would expect and upvoted your post, but I think your are reporting a bug rather than making a feature request.
In other words, it is my understanding that this part of the forum should remain reserved for genuine feature requests.
0 -
I understand your point and do to some extent agree. Though as this issue doesn't seem to be caused by a mere bug, but rather some fundamental flaw in the keystone adjustment function that requires a "new" such function, I took the liberty of requesting it here.
0 -
Hi Thomas,
Thank you for your post.
Your comments were taken on board and added to the previous feature request regarding the Keystone tool.
0 -
Hi Lily,
Thanks for your reply.
Which request was that? (I believe there are several different keystone-related requests.)
0 -
I second the need for improvements in this area. For architectural photography these adjustments are key and not working well in C1; this is a feature that I use daily. A few requests in particular:
- Increase the maximum available amount of keystone from 75 to ?? Realizing that these significant changes degrade the image quality, I would like to be able to decide what level is acceptable without being limited here.
- Guides seem similarly limited; they do not always fully correct verticals.
- Implementing the option of a loupe on the ends of guidelines would be tremendously helpful (ala lightroom or camera raw).
Thanks! Would help my workflow tremendously not to have to jump to PS for these kinds of procedures.
1 -
The two kinds of keystone malfunction that I've come across are with simultaneous vertical and horizontal adjustments and with vertical adjustments.
In the fist case, the four-sided figure formed by the four adjustment points comes out as a parallelogram and not as a rectangle. Here's an example:
In the second case, the the lines that should come out vertical still aren't vertical (this is after raising the keystone amount from the default 80 to 100). Example:
Apart from these issues being fixed (which, obviously, is of the highest priority), I'd very much like to see the following improvements:
1) Unlocked adjustment points: two independent vertical lines (four vertical adjustment points) and two independent horizontal lines (four horizontal adjustment points), rather four adjustment points locked together.
2) Retain adjustment point positions after the application is closed.
3) Change the default amount for vertical adjustments from the current arbitrary 80 to the expected 100.
4) Allow the colour of keystone adjustment points and lines to be changed so that you can see them on all backgrounds.
0 -
Thomas’ experiences with the guided adjustments mirror my own frustrations. Separate horizontal and vertical guides would indeed be quite helpful, as would his other suggestions.
If I may, knowing that comparisons to Lightroom are a bit anathema; an ‘auto’ function would also be appreciated here, as Lightroom’s generally works quite well and is a big time saver. As a Fuji shooter I am quite enamored with C1 and will continue to use it, but as long as we’re discussing the keystone function thought I would add that request. I realize that’s likely a much larger undertaking, however.
Thanks!
0 -
Spencer,
If you have an image that requires correction in 3 dimensions do you use the Aspect ratio when attempting to correct the third dimension?
0 -
As Spencer suggested above, a loupe function would make keystone adjustments a lot easier. As it is, you need to first place the adjustment points roughly where you want them, then zoom in on each one individually to place them more precisely (a task that isn't made any easier by the often indiscernible colour of the adjustment points and the lines between them). And finally, if you're doing vertical adjustment only, you need manually raise the keystone amount from 80 to 100. This could be a great deal faster.
0 -
Perhaps the fastest option would be to use the Keystone tool, zoomed in, and then work with the grid produced when moving a slider. The grid provides the potential for multiple reference lines and some scope for greater overall precision when zoomed in.
Residual anomalies from, say, lens correction or rotation requirements should also be readily visible.
Using the Space bar invoked zoomed image current view area navigator should also help to quickly locate desired reference points, although the spot points usually discussed may not be so relevant once the grids are deployed.
Seeing the effects in real time as the sliders are moved is something I find quite useful - especially if the original shot is always going to be a compromise of adjustments in 2 dimensions when the original subject matter offers logical reference points that require 3 dimensions to describe their relationships completely.
0 -
I'm sure it's possible to get used to using the sliders, but I doubt that it's faster than using adjustment points with a properly working keystone tool. If only the actual malfunction was fixed, I wouldn't mind zooming in and out so much, though an option to change the colour of the adjustment points and lines would be a great help as far as positioning the adjustment points goes.
What I need from the keystone tool is an easy way to position two lines that are to be horizontal and/or two lines that are to be vertical, preferably independent of each other rather than locked in a four-sided figure as they currently are in Capture One.
0 -
various auto options in ACR/LR for Geometry Correction are indeed the feature that C1 is sorely lacking
0 -
Considering the current, rather rudimentary state of Capture One's keystone function, I would imagine that auto keystone is quite far off. As far as Lightroom's keystone tool is concerned, I would always choose manual (guided) adjustment over auto.
0 -
> As far as Lightroom's keystone tool is concerned, I would always choose manual (guided) adjustment over auto.
in my usage (not architectural shots) I use either of Auto options in ACR in ~4 out 5 cases as a starting point ... the argument is useless - it is like TTL vs Manual with Flash/Strobes... every TTL device can do manual, no manual can do TTL0 -
SFA:
I haven't used the aspect slider much; I can see its utility but I probably wouldn't touch it much unless a subject 'looked' wrong (i.e. a building too short). I'm a bit confused by your notion of the third dimension in an image, but for architecture the primary concern is vertical lines being vertical. Occasionally for single point perspectives etc. I will also correct for horizontal keystone. Without knowing the math, is this something that could be estimated from the other keystone values? I.e. a vertical keystone of 25 would by default adjust the aspect to -1 to produce something that 'looked' right?
Perhaps we're rehashing this a bit, but while you're right that we can figure out the fastest way to work with what we have (and your points are helpful, thank you), it seems like we are not alone in thinking that this is an area C1 could improve upon in future releases.
Thomas:
I agree that the auto adjustment is likely a long way off, but it would certainly be welcomed. When I have a number of images to select from or have the client review, editing them individually is quite time consuming and 'good enough' would be sufficient, with the caveat that manual guided adjustments could be fine tuned later on selected images.
I'm not entirely sure how the amount slider is currently working; perhaps in your tests you've established this, but is the 80 value it defaults to returning vertical lines for you, or is only 100 fully correcting the keystone? When I adjust to 100 I am sometimes seeing verticals starting to lean out, but not sure if that's just a lack of precision on my part.
Additionally, do you find that even the vertical-only guided adjustment affects the horizontal keystone slider a bit? Not sure why that would help, but perhaps it's intentional.
Would it be helpful for the movement of the keystone guides to be undo-able step by step? I sometimes find myself wanting to move them back one step.
One final ask for the C1 team:
On a mac, the first time you click on a guide point circle after in an uncorrected image (or a previously corrected image once you've exited the program, to Thomas' point), the 4-way arrow remains visible; you have to click once more to get just the precision white dot. Not a huge thing, just kind of annoying on a bunch of images.
0 -
Spencer:
I see your point regarding auto adjustment. It's a long time since I've used it in Lightroom as I rarely got vertical verticals nor horizontal horizontals, but perhaps it's been improved since then.
For vertical adjustment you need to manually raise the amount from 80 to 100 after applying corrections in order to get vertical lines. Except, as you point out, the verticals are still off sometimes. In some cases the verticals are actually parallel, only the angle needs adjusting; in other cases the verticals are converging either at the top or the bottom. (As I understand, the official reason for the default of 80 rather than 100 is that it's supposed to look more "natural", which doesn't make sense as the content isn't taken into consideration.)
I haven't noticed vertical adjustment affecting the horizontal keystone slider, but it's certainly possible.Step by step undo for keystone guide position would be useful.
You're right about the 4-way arrow remaining visible the first time you click; it would be nice to have this fixed.
0 -
Spencer:
"I'm a bit confused by your notion of the third dimension in an image, but for architecture the primary concern is vertical lines being vertical. Occasionally for single point perspectives etc. I will also correct for horizontal keystone."
The third dimension would usually be in the subject matter for architecture and perhaps more easily though of as a plane for art repro. The kind of thing that one might seek to compensate for by lens shifts when translating a geometric view onto a film or sensor. The added complication is the perspective, partly for focus concerns but also, if the amount of adjustment is significant, for the level of detail obtainable from the original data, because of a need to deal with potentially large changes.
Basically (ignoring any additional effects of built in lens distortion) the simple consideration is the equivalent of measuring a box. Length, Height, Depth.
If making an image some distance from a structure and using a telephoto lens to fill the frame the straight line distance for the sensor to the top and to the bottom of the structure , measured from a point at ground level, will be more or less the same. Verticals and horizontals, within reason, will not diverge or converge no matter whether the camera is pointing level to the ground (i.e. at the horizon) or slightly up or slightly down.
Move closer and point at the horizon using a wider angle and verticals will still be vertical, horizontals will be horizontal and only distortion from the lens design will change that. However other factors come into play, mainly the distance of different parts of the subject from the camera varying significantly in what structurally is a common plane as far as the building is concerned but not necessarily as far as the lens characteristics are concerned.
If the subject area to be corrected is, in effect, a trapezoid attempting to make it a rectangle in just 2 dimensions may still require some "correction" in the third dimension. The only available tools to assist with that in photo editors will be those offering Perspective adjustments, But to use those in many cases for architectural purposes (where the human brain expects certain appearances and compensates for them in its interpretation) will really only work if the rotation of the image is sympathetic to the changes to be made and no existing lens distortion acts against the visual perception of the correction.
If, however, one looks at pixel editors that can push and pull, cull and in particular create pixel data for the images in quite extreme ways - for example some that make a tool to stretch and bend sections of an image - the options for tweaking things,, if not the outright image quality resulting, may be more extensive than those that do not. Especially for images requiring the greatest alterations. Whether one sees that as opportunity as photographic editing to simulate the traditional lens movements of history or more of an artistic manipulation with potentially extreme options for artistic interpretations is probably a debate that will last for as long as humans exist.
0 -
I don't quite get the point about the 3rd dimension. Regardless of the appearance/illusion of 3-dimensionality in an image, the image is still 2-dimensional. Apart from barrel/pincushion distortion, once vertical and horizontal lines have been adjusted what other adjustments could you make that wouldn't throw the verticals and horizontals off again?
0 -
I think you do get the point about third dimension Thomas.
If you squeeze or stretch an image to make converging lines parallel then there will be a degree of manipulation points that are not on the two dimensional plane of focus may not necessarily be adjustable in a consistent way in both vertical and horizontal adjustment terms (i.e. as needed to interpret the 3 dimensional subject into a 2D presentation.
That part of the image may be distorted or reduced image quality may result as pixels are either binned or "stretched" (possibly in-filled?). Most obvious if the required adjustments are large.
Alternatively one ends up with the corrected image looking a little odd in its proportions (though possibly acceptable for the affect one seeks) and maybe some features, like windows, that should be the same size presenting as different sizes. Our brains will tens to adjust for that (or not really spot the anomalies) in much the same way that they automatically interpret keystoning as observed through normal 3D vision.
I have to rush out now but may be able to come back to this later.
0 -
Thanks for the elaboration, SFA.
It's obvious that there's a tradeoff, various sorts of distortion will result from keystone correction and part of the image will be lost. I don't imagine anyone would contest that. This is a different issue, though, from whether or not the keystone function works as it's supposed to or not, and what sorts of unexpected results (errors) it produces.
0 -
Another example. The flat surface of a building, one-point perspective, requiring very little correction, yet the horizontals are off.
The adjustment points are lined up with the lines in the wall of the building, the red lines are Capture One guides:
I've started using Capture One in late December, so I'm not familiar with previous versions, but I don't assume the keystone function has ever worked correctly. So it's somewhat surprising that users haven't complained more and that, by all appearances, no effort has been made to fix it long ago.
0 -
Thomas,
Two questions about the image example above.
Is the building on a slight hill? The road looks like it slopes downwards form right to left.
Were you facing the building directly or at a slight angle? I can't tell whether the road is parallel to the building or angled from it. And the Windows on the right of the sample look like they were shot with the camera perpendicular to the building (or at least more precisely perpendicular) than the windows on the left of the image.
Also - is the lens subject to any notable corrections for distortion and was the frame shot in vertical or horizontal orientation? Is it cropped away from the edges or more or less complete?
It's one of those shots that looks very simple and plain but may not be quite as simple and plain as our first visual impressions are interpreted by our brains. (and of course there is no certainty that all brains work the same way. In fact there is probably greater certainty that they do not!)
0 -
SFA,
Yes, the road slopes a little. This shouldn't make any difference though.
I believe I was facing the building directly more or less from the middle of what you see in the photo.
No distortion correction has been applied. And the photo has only been cropped to the extent required after keystone correction as you can see below.
None of this should make any difference. I've seen this behaviour again and again in Capture One, both the verticals being off and the horizontals being off. Sometimes after applying vertical and horizontal keystone correction, even the angle is off on both axes.
This photo was taken with a Canon TS-E 24MM F/3.5L II tilt-shift lens, but Capture One's keystone function produces errors no matter what lens is used. I.e. the lens is of no importance.
0 -
A TS-E with tilt and shifts applied?
0 -
"A TS-E with tilt and shifts applied?"
No tilting.
(But even if tilting had been applied, the result should still have been horizontal horizontals and vertical verticals.)
0 -
So...Capture 1 folks...
It sounded like you added Thomas' comments to one of the feature requests; can you confirm that this area of the program is something you're looking into, including the additional ideas above? Then we can maybe put this to bed for a bit. :)
Thanks!
0
Post is closed for comments.
Comments
26 comments