leifs wrote:

klausesser wrote:

To get a perfect stitch you need a perfect shooting.

Kolor would soon be out of business if perfect shooting was a prerequisite.
I guess most of their algorithmes is about stitching non-perfect shot panos.

Leif

Ok - let´s say: "a very good stitch needs a very well made shooting". You can´t expect a software to compensate completely a faulty shooting fully automatically.
I realized a long time before that dealing with hires needs high precision in the setup as well as in the shooting. Wobbeling poles can´t provide this precision - so can´t handheld shooting.
Using a fisheye it´s not much of a problem - but using a 35mm and longer on FX it definitely becomes one.

best, Klaus

leifs wrote:

I have no problems stitching spheres shot with my 50mm FF-equivalent lens, except when there is a long horizon without CPs.

What many with me have been asking for is a tool to help the editing. The only small thing missing is Kolor completing what they started. You see, it's there already, but not for cylinders and spheres !
Kolor should now remove the "Planar projection only" greyout. ASAP. Users have been asking for it during the last six years !

Leif

Leif - using the vertical-line tool for straightening horizon in spherical mode works this way:

Without havin a vertical anywhere draw a vertical line somewhere. You´ll realize a yellow dotted line on the vertical line. Use this yellow dotted line to place it alongside the horizon you want to get streaightened. Use more of these lines untill the complete horizon is followed by yellow dotted lines - which touches each other. The preferablel way i found is to use many short lines instead of few long lines.

Then hit "enter". Most likely you will have the horizon perfectly straightened now. That´s all what it´s about. Always worked out fine for ma an for lots of other photographers i know - so it most unlikely wouldn´t work for you too . . . :cool:

Watch this video - near the end is a practical explanation of straightening the horizon: http://www.autopano.net/wiki-en/action/ … lines_tool

best, Klaus

Last edited by klausesser (2012-12-27 22:23:08)

leifs wrote:

klausesser wrote:

Leif - using the vertical-line tool for straightening horizon in spherical mode works this way:
best, Klaus

Thanks a lot.  But I know how to use it.
I have tried it a lot and really wanted it to do the trick. I have tried hard !
But for the horizon-problem it only makes things worse! In the editor it looks like the problem is solved. Everything looks fine.
After rendering (with anti-ghost) inspecting in 100% it is revealed that the vertical line tool produces steps ! As do the "move pano" tool. I guess its actually the same algorithm.
Rendering using the "Simple" blending preset blurs the step, makes them less pronounced. But for the bottom of the pano the shooting setup need to be "perfect" to avoid artifacts. Using "Anti-ghost" preset makes the pano fine, except for the steps. They are enhanced !

Leif

As i said before: to get a correct horizon you need to have a correct stitch first. THEN you can straighten the horizon. Your problem is to have steps in the STITCH. So you first hand have to get rid of the steps in the STITCH before getting getting the horizon geometrically straightened.

This are TWO issues: one with the control-points and one is in the geometrical construction which RESULTS from the CP issue. As i said: first you need to correct the stitch - then you can correct the geometrics.

I suggest in THIS panorama to use Photoshop to correct the steps and the horizon and come to an end finally. You most unlikely will get it corrected in APG bacause of a shooting-issue.

For the next shootings you should care for the pole resp. the whole setup. Seeing the massive issues you have again and again i suspect a fundamental problem here.
No new tool, Kolor might add - whatever it might be, horizontal straighening or whatever - will be able to solve your problem . . .

Maybe you should try PTGui . . . did you do already? Would be interesting how it handles the CPs in combination with the PapyWizard-import.

btw.: did you use the xml at all with this pano? Might be of no help because of the wobbeling pole i guess - but anyway interesting . .

best, Klaus

Last edited by klausesser (2012-12-27 23:06:28)

That was quite a lot of good advice !

1) I'm not the only one posting on horizon-trouble. In the forum there is ca 1200 posts on this issue.
2) I've doctored the steps in PS 6, even if I'm no PS-pro. It should not be neccesary.
3) I like APG and do not want to learn another stitching tool. I've got a lot of tools as is.
4) ofcourse I use the XML's from the VRdrive2. there is no way of doing 50mm spheres without a precise head that makes XML's (and use the XML's)

I won't comment the "wobbeling pole" statement. Maybe Gitzo won't like it about their expensive carbon telescopic poles.

I've seen the video case 3 several times, and I can not see any relevance to the horizon line problem. There is no line horizon without CPs there.
Totally irrelevant.

Leif

CheeseAndJamSandwich wrote:

This specific problem has frustrated me for years... And all the advice so far hasn't helped...
Did i mention that it's quite frustrating???

You did. Did i mention it´s frustrating for me also to constantly recall the same point again and again knowing that it works - if the shooting is correct.

I´m used to correct shootings  - and so i never have issues with horizons regardless how much tilted the camera stands and in which coastly and windy situation i shoot. I additionally apologize to you and Leifs for trying to convince you to think about the way of your shootings. Also i apologize to suspect there must be something wrong. Of course there CAN NOT be something wrong in the way you shoot . .

Your issues definitely are to blaim to APG only . . . Ok? Ok.

I´m pulling out of that now.

best and good luck, Klaus

Right, back on topic.

To Alexandre & gang...
About these vertical control points...  To many of us here, they seem like one of the more obvious ways to combat this specific problem, stepped horizons... And it'll make fixing wiggly horizons a breeze too...  Any chance of this in the future?
Are you sneakily already developing this ;-)

ThomasV wrote:

But, there is no limitation to that, since you can change the projection to planar, draw your horizontal lines, apply the tool, and come back to your initial projection.
Regards,
Thomas

I will try this out. It maybe a smart way of aligning the full horizon. But I don't see it as a remove-step-tool.
Below is a recent example that illustrates the problem. The overall RMS is OK, 1.36 for the image-pair is OK. Inspecting in 10% everything is OK.
Inspecting in 50%, there it is !

Maybe I'm too ambitious, wanting to shoot coastal spheres with a 50mm FF-eq lens. But the Leica 25mm f1.4 lens is the best lens in the microfourthirds universe and it gives me top-notch images. So for now I will use it and repair the steps in PS, like I did in this pano 
http://www.rundskuer.no/panotour/langen … ltour.html

Leif

Last edited by leifs (2012-12-28 14:12:37)

gkaefer wrote:

a horizontal tool did exist in past also for horizontal alignement in spherical projections. kolor did post in forum they did eliminate this feature because of false results - I would sum up it that it did cause more problems than it did solve ones, so it was disabled for some projection types.

Wrong. The horizontal lines tool ALWAYS worked in planar-mode only.

gkaefer wrote:

So again and again saying that users should resolve their errors in setup and than the horizont alignment will work out of the box...
its a valid argument, but also valid is the wish of other users to get such a feature back - and if possible a better working one than in past.

Leif´s issue isn´t a non-straight horizon. A "straight horizon" tool exists in APG and it works well.

Leif´s issue are STEPS in the horizon - NOT an issue of ALIGNING the horizon. These STEPS comes from STITCHING the images appropriately. Why ever. So ALIGNING a horizon woud be the NEXT step after STITCHING the horizon-images correctly.

Sorry, but it´s as simple as that.

best, Kaus

Last edited by klausesser (2012-12-28 14:09:02)

leifs wrote:

I guess there is enough horizon-straightening tools in APG as it is (vertical line tool, vanishing point tool, move pano tool etc)

Correct - thank you for stating that

leifs wrote:

What's needed is help to remove the steps and keep it there, so that the horizon-staightening does not generate new steps.

Regarding your issue are VERTICAL steps: how would a HORIZONTAL "straightening" tool be able to correct these?

In your case the horizontal tool MUST generate new steps - because the steps come from wrong control points in the STITCH. The tool doesn´t deal with control points but with geometrics AFTER control points are found and set - as i understand it.

The Kolor guys wrote a tutorial for you. Some months ago Alexander made a tutorial for me relating my issues in the "Opera" stitch. He showed me features i didn´t know they exist in APG. I could solve my problems by following his instructions.

So i´m quite convinced you could solve your issues also - by following the tutorial they made for you

best and good luck, Klaus

Last edited by klausesser (2012-12-28 14:43:11)

leifs wrote:

But the methodology, "move panorama" tool after editing, is not wise.

To be honest: that´s the first case in my knowledge when it doesn´t work. So i´m a bit sceptical: i can see no reason at all why it wouldn´t work . .

As i said before: i would definitely give PTGui a try - just to evaluate it´s different way of handling CPs and it´s optimization strategy.

In my eyes - i may be wrong - APGs startegy of optimization is too sensitive. Therefore i suspect - nobody explained it yet - that because of a too high sensitivity APG does deal with definitely clear structures for placing CPs but place them in somewhat funny ways sometimes.

This might be a reason for your issues also.

But my experiences so far showed that by deleting CPs and adding new ones manually such issues can be solved most of the time . . but not always

In cases i can´t solve it in APG . . i use PTGui - and in most of the cases that works out astonishingly well . . .

Another option is use in extremely wrong-going cases is: retouching in Photoshop. Because of the fact these issues usually occur clearly visible around 100% it´s a problem to look and find them all for doing a retouche in PS - but what has to be done has to be done . .

In this case http://360impressions.de/MKP_Panorama/ i retouched all (!) the ceiling and most of the walls manually in PS - took quite a time i can say . .

After havinbg it ready and published by the client i played around with it in PTGui - and got a nearly perfect stitch after running the optimizer once . . there were exactly two glitches i needed to retouch in PS. That was it.

The reason for the issue clearly was sensor-dust and the use of f11 instead of f8. The dust was clearly visible and relative sharp with the 35mm lens @f11 - APG preferred to "optimize" the dust . . . PTGui optimized the clearly visible structures on the ceiling. This told me that the optimizer in APG seems to need a feature to adjust it´s sensitivity in a wider range (if there is any adjustment-feature existent already besides of low, mid, high).

This optimization-adjustment feature might be extremely helpful with issues like yours too! I think about a variablle "catch radius" in which it looks for optimizing CPs - actually i mean the search-radius (or whatever) iss too wide . . and so it seem to "optimize" where no optimization is wanted or even needed.

Being no programmer i only can try to imagine what the issue might be - but clearly there IS an issue in some cases relating to optimization.

best, Klaus

Last edited by klausesser (2012-12-28 18:09:46)

Hi Leifs
First, I agree with Klaus that a "horizontal straightening tool" will not fix your problem. Because it isn't a problem of straightening the horizon. You need a "anti-step-tool"
Your problem based on three issues:
One is the not perfect shooting. Your pole is not fixed enougth to produce a set of images that are usable with the XML-informations.
You are really shooting in the multiple viewpoint modus. As I have previously recommended, I would not use a pole in this case or make hands-free shooting. A sturdy tripod can already help a lot to create a usefull set of images.
The second problem is that there are  at all no checkpoints to detect in the area of ​​the horizon line . There are simply no visible and assessable checkpoints there.
The third problem is that the whole scene is very dynamic, it's a stirred soup. Clouds and waves moves while shooting and make it impossible to detect meaningful checkpoints in the area. No image analysis software can ever provide useful checkpoints in this case. So there will never be a fully automated solution.

One solution has been described: Create the Panorama optimized as much as possible, then render all images to layers in a PSD, and then correct the images along the horizon line manually in Photoshop a little bit. This is no big deal.

Another approach would be to introduce a new type of manual control "points" - or control lines. The user sets them itself exactly on the horizon line in each image, APG considers these points or lines as "on horizon" and trying to make a continuous line of it without any steps.

leifs wrote:

lumelix wrote:

Another approach would be to introduce a new type of manual control "points" - or control lines. The user sets them itself exactly on the horizon line in each image, APG considers these points or lines as "on horizon" and trying to make a continuous line of it without any steps.

I agree
This is the same as I suggested in http://www.kolor.com/forum/p107908-2012 … 57#p107908

I agree too
This is the same as i suggested in http://www.kolor.com/forum/t7906-horizon-tool
Great minds think alike, and all that
But it is the most obvious way to do it.

EVERY spherical pano has a horizon.  Just like the Earth has it's equator.  Being able to use this equatorial line for defining a horizon would be ultimately powerful.

Last edited by CheeseAndJamSandwich (2013-01-04 08:51:02)