My Autopano Pro forum

Sharing knowledge around Autopano

You are not logged in.

Announcement

The documentation of AutoPano Pro can be found online at our Wiki. Before posting a question, please check with the manual!

#1 2008-03-24 00:29:29

richardhc
Member
Registered: 2007-02-28
Posts: 13

problems with repeating patterns - floor tiles and planking

I have been using autopano pro since the very first version - it is the very best stitcher available.
But I have noticed that - especially with fisheye shots - floor tiles and planked decking often has broken lines. I try to "fix" these using the control points editor but there are always some that remain and need to be manually repaired in photoshop.

I am using a carefully calibrated nodal-ninja pano head and the images are razor sharp - but still Autopano has glitches with tiles.

If the scene is a non-repeating pattern, like grass or branches, Autopano pro rarely misses.

In selecting control points I have tried capturing large groups of tiles and small areas of just one tile - but the glitches appear in the final product anyway.

Planked decking has the same problem, always at the seam between two images. If the tiles have a random pattern on them (like marble) this is not a problem.

It would be great if someone could help me find a solution to this as these repairs take time to correct.


Uploaded Images

Offline

 

#2 2008-03-24 02:52:59

DrSlony
Member
From: Poland, Zielona Góra
Registered: 2007-11-03
Posts: 946
Website

Re: problems with repeating patterns - floor tiles and planking

I also very often have these issues when shooting 360x180. I almost always use "customized fine tuner" to get even better rms, but i still get these problems with tiles, planks and other objects with similar patterns.

Offline

 

#3 2008-03-24 04:33:24

hankkarl
Member
From: Connecticut, USA
Registered: 2006-02-21
Posts: 1137
Website

Re: problems with repeating patterns - floor tiles and planking

I found that problem on vinyl siding - it has a very strong pattern, but each piece of siding has the same pattern, so APP can find many matches, but it only uses one of the many possible matches (and usually not the correct one.)

Floor tiles may be the same - that is, they are not random at all, but only look random becasue they have a large area between repeats.

Is the planked decking natural wood or a man-made composite (e.g. "Trex") ?

APP finds matching areas, so we can't blame it if it picks the wrong matching areas if two or more match.  But (I think) Alexandre is working on an edge detector for  this situation.

Last edited by hankkarl (2008-03-24 04:34:01)

Offline

 

#4 2008-03-24 11:47:24

GURL
Member
From: Grenoble
Registered: 2005-12-06
Posts: 1935
Website

Re: problems with repeating patterns - floor tiles and planking

richardhc wrote:

I am using a carefully calibrated nodal-ninja pano head and the images are razor sharp - but still Autopano has glitches with tiles.

The ground near to the tripod legs is often the closest and thus the most difficult part to stitch in a spherical pano. There are unavoidable parallax errors on the source images corners (rectangular format fisheye images) or near the image circle (circular fisheyes.)

One reason: fisheye lenses so called No Parallax Point is not a point but a segment ! A quick and dirty proof for that: when you are watching the lens from the side you can see that the NP"P" moves forward.

An other possible reason: "different fisheye lenses distort images differently, and the manner of distortion is referred to as their mapping function." http://en.wikipedia.org/wiki/Fisheye_le … g_function There are actually many species of fisheyes: equidistant, orthographic, equisolid (but alas no stereographic ones.) This probably precludes a perfect "lens distortion adjustment" (ask Alexandre for more details...).

Workarounds:
- uses Smartblend rather than Multiband
- take more shots
- use APP circular crop tool
- smile use viewer pitch max value to hide the nadir region.

Last edited by GURL (2008-03-24 12:01:43)

Offline

 

#5 2008-03-24 12:58:22

mediavets
Member
From: Isleham, Cambridgehire, UK.
Registered: 2007-11-14
Posts: 851
Website

Re: problems with repeating patterns - floor tiles and planking

richardhc wrote:

I have been using autopano pro since the very first version - it is the very best stitcher available.
But I have noticed that - especially with fisheye shots - floor tiles and planked decking often has broken lines. I try to "fix" these using the control points editor but there are always some that remain and need to be manually repaired in photoshop.

I am using a carefully calibrated nodal-ninja pano head and the images are razor sharp - but still Autopano has glitches with tiles.
.

I have seen this issue too - and I always use Smartblend.

Do you see the same problem if you stitch with PTgui? Just wondering.

I accept that FE lenses have no single NPP but if one stitcher can cope with this but not another...


Andrew
Nikon D40, Nikkor 10.5mm fisheye, Sigma 8mm f3.5 fisheye, 18-55mm kit lens, Nodal Ninja 5 Lite
Nikon P5100, CP5000, CP995, FC-E8, WC-E63,WC-E68, TC-E2, Kaidan  Kiwi 995, Bophoto pano bracket
Merlin/Orion panohead + Papywizard on Nokia 770 and Windows XP/2K

Offline

 

#6 2008-03-24 16:39:09

DrSlony
Member
From: Poland, Zielona Góra
Registered: 2007-11-03
Posts: 946
Website

Re: problems with repeating patterns - floor tiles and planking

I also use smartblend and this issue can happen anywhere on the image, not just close to zenith or nadir. It just so happens that most of the time the tiled objects are on the floor or ceiling :]

See the attached screenshots.

These were difficult photos beause there is no texture in the ceiling or tiles, but APP was able to correctly place, rotate and zoom each image, it just wasnt able to make the details line up nicely.


Uploaded Images

Last edited by DrSlony (2008-03-24 16:41:12)

Offline

 

#7 2008-03-24 16:49:53

digipano
Member
Registered: 2008-02-16
Posts: 124

Re: problems with repeating patterns - floor tiles and planking

Interesting to know that.

Offline

 

#8 2008-03-24 17:24:20

mediavets
Member
From: Isleham, Cambridgehire, UK.
Registered: 2007-11-14
Posts: 851
Website

Re: problems with repeating patterns - floor tiles and planking

richardhc wrote:

I am using a carefully calibrated nodal-ninja pano head and the images are razor sharp - but still Autopano has glitches with tiles.

If the scene is a non-repeating pattern, like grass or branches, Autopano pro rarely misses.

In selecting control points I have tried capturing large groups of tiles and small areas of just one tile - but the glitches appear in the final product anyway.

Planked decking has the same problem, always at the seam between two images. If the tiles have a random pattern on them (like marble) this is not a problem.

It would be great if someone could help me find a solution to this as these repairs take time to correct.

It's not clear to me whether this is a problem particularly related to tiles, planks etc. or whether it's just that easier to spot the issue where there are such features in the stitched image?


Andrew
Nikon D40, Nikkor 10.5mm fisheye, Sigma 8mm f3.5 fisheye, 18-55mm kit lens, Nodal Ninja 5 Lite
Nikon P5100, CP5000, CP995, FC-E8, WC-E63,WC-E68, TC-E2, Kaidan  Kiwi 995, Bophoto pano bracket
Merlin/Orion panohead + Papywizard on Nokia 770 and Windows XP/2K

Offline

 

#9 2008-03-24 18:01:58

DrSlony
Member
From: Poland, Zielona Góra
Registered: 2007-11-03
Posts: 946
Website

Re: problems with repeating patterns - floor tiles and planking

Its related to objects that are plain colored with no clear difference between point A and point B on that object, like that ceiling.

Offline

 

#10 2008-03-25 08:14:37

Castillonis
Member
Registered: 2008-03-06
Posts: 60

Re: problems with repeating patterns - floor tiles and planking

I thought I might use some blue painters tape to make small crosses at several points and then use the stamp tool in photoshop to remove them from the image.  I'll have to try it out.

I just rcvd my NNL5, leveler, Sigma 10mm f/2.8 fisheye, and Canon 40D so I am still figuring things out.

Offline

 

#11 2008-03-25 11:01:17

GURL
Member
From: Grenoble
Registered: 2005-12-06
Posts: 1935
Website

Re: problems with repeating patterns - floor tiles and planking

Castillonis wrote:

I thought I might use some blue painters tape to make small crosses at several points and then use the stamp tool in photoshop to remove them from the image.  I'll have to try it out.

lol A long time ago on Max Lyons'forum somebody offered small adhesive CP targets for that!

big_smile Kolor should develop and market such Disposable Adhesive CP Targets and a special Patch Remover application...

(Rather than painter tape I used  patches I cut out of a newspaper. This worked well but patches size must vary according to their distance from the camera.)

Offline

 

#12 2008-03-27 20:59:24

richardhc
Member
Registered: 2007-02-28
Posts: 13

Re: problems with repeating patterns - floor tiles and planking

I do use Smartblend rather than Multiband
Taking more shots does not correct the problem
I do use the APP circular crop tool
I know the focal point shifts on fisheye lenses but the "glitches" happen even in areas far away from the camera

The problem with CP targets is that in APP you can't pinpoint one CP. I would not need tape on the floor I can easily see the intersection of the tiles and could - if it were possible in APP - click to put a single CP on that intersection. I have always wished there was some way to put just one very important control point in the images. Sort of a very heavily weighted "this point IS the same point in both images and MUST be aligned". Then all I would need to do is add that point to a couple of intersections of the tiles and the image would align better.

This would also help with sea-level horizon glitches (which are also common in APP).

The select and grab method of CP insertion on APP is a great feature but I've never understood why a simple "insert one" CP should be excluded from the program.

With or without this APP remains my champion stitching program - I just downloaded the latest PTGui Pro, dragged in a 21 image panorama and got scrambled eggs. I dropped the same images into APP and presto - a perfect stitch and render.

But back to the tile problem, APP does not have this problem with non-repeating patterns and figures out exactly how to match branches and grass blades and palm fronds to perfection. It's the identical repeating pattern that gives it a problem.

Offline

 

#13 2008-03-27 21:27:22

richardhc
Member
Registered: 2007-02-28
Posts: 13

Re: problems with repeating patterns - floor tiles and planking

My wife had an interesting insight about tiles (she's the one who repairs these glitches).
She points out that the tiles ARE stitched correctly in most image seams - but if APP has a problem anywhere along a seam that this results in tile glitches even if the "problem" isn't with the tiles.

For example, the red tile image I included above comes from a seam where near the upper 1/3 of the image there is a ceiling fan . In one image the blade of the fan covers a ceiling beam and in the next image the fan blade does not.

So APP could not match the ceiling beam very well from one image to the next and the first "glitch" on that seam is from this problem. This problem area then disturbs the stitching along the entire seam. I tried to correct it using CPs with chairs and tiles etc on the lower section of the image but could not do anything about the fact that you could see the wood beam in one image but the fan blade covered it in the next - so that glitch had to be corrected in photoshop. But no matter what I did on the bottom the tile glitches still happened.

I have noticed that the glitches tend to propagate along a seam. It may be a difference in shadows caused by changes in fill flash, or some other "difference" in one area on the seam that causes the first problem. Then this is creates glitches along that seam, sometimes from top to bottom, sometimes just in one area.

Yet, it does not seem to be a problem in outdoor scenes (except perhaps sea horizons which often have glitches). It seems to happen mostly on straight lines and edges.

It is not a major problem as the glitches are easy enough to quickly fix in photoshop and there are never very many in a sphere - but there are always some in every sphere that has tiles.

For some reason we have not seen this as much in non-sphere panoramas done with normal lenses (although they do happen). Maybe it's just that the chances of having a mis-match between the adjacent images are greater in a sphere image because the length of the seam is greater and there are more chances of something going wrong somewhere along it.

Offline

 

Board footer

Powered by PunBB
© Copyright 2002–2005 Rickard Andersson