Itching for Stitching

[BernardLanguillier]

I have been stitching a lot for years now, using both PTgui and Autopano Pro (1.4 being probably the best package out there at the moment).

I have found the auto stitching abilities of Autopano Pro with low contrast images containing areas with little marked features to be vastly superior to PTgui's. On the other hand I prefer PTgui's interface for control points etc...

You might be interesting in knowing that as of PTgui Pro 7.8 they have implemented their own version of the infuse algo (HDR based on pixel level brightness evaluation in 16 bit space). This works pretty well per my first tests.

One recent sample made up of 36 Nikon D3 images (resulting image is 200 megap pixel). This was shot a few weeks ago in the Himalayas, the mountan pictures is Pumori at 7150 m. This was shot with a RRS pano head.



Thanks to its speed of operation and wonderfully stable shutter, I have found the D3 to be much easier to use for stitching purpose than the Mamiya ZD I also use.

- Very high quality moderate ISO (400 or even 800 ISO) is a huge plus in various shooting conditions where the ZD's low ISO is a real pain since it induces longer exposures,
- The ability to check images sharpness on screen at pixel level in a fraction of a second is a blessing when using longer lenses in windy situations,
- Deep RAW buffer also help with panos containing more than 10 images.

Cheers,
Bernard

[BernardLanguillier]

I find that the experience with AutoPano Pro can be mixed - i've had situations where I've just machine gunned handheld (in a manner of speech) a few shots and it stitched them no problems, and other cases where I used the pano head with just 2 shots and had an alignment problem. I'm definitely doing something wrong here, I guess so perhaps it's "back to school" until I get it right.
[a href=\"index.php?act=findpost&pid=193831\"][{POST_SNAPBACK}][/a]

This is very surprising. Was the lens correctly positioned on top of its nodal point when you shot with a tripod?

Regards,
Bernard

[Graham Mitchell]

I'm surprised so many are happy with CS3. You don't appear to have any perspective and distortion control. I've never hard it work for me.

PTGui, on the other hand, works great. (And I'm sure the other PT based tools would do the same job).

[shutay]

This is very surprising. Was the lens correctly positioned on top of its nodal point when you shot with a tripod?

Regards,
Bernard
[a href=\"index.php?act=findpost&pid=193990\"][{POST_SNAPBACK}][/a]

I've only had the pano head about a week so far, and looking back at the test shots I used, the ones that were shot handheld and succeeded were the distant landscape type shots with no nearby objects. I also tried some other handheld panos, not so much (I hope) to be an idiot but to understand the limits and limitations:

1. Vintage Porsche in the street: handheld - heavy misalignment of foreground and background elements - ok, now I know, don't go there unless you have a pano head.

2. Living room 2 shot pano with the pano head - I simply couldn't get AutoPano Pro or Photoshop CS to get them to line up exactly. There were close and distant objects, so I knew that it's a challenging one, but the PTgui 7.8 trial software stitched that up wonderfully.

My key learning from the other postings and my experience is to go back and double check that the camera is properly positioned over the nodal point and be more rigorous in my technique. It's obviously not the head or the software but me!

[BernardLanguillier]

I'm surprised so many are happy with CS3. You don't appear to have any perspective and distortion control. I've never hard it work for me.
[a href=\"index.php?act=findpost&pid=193998\"][{POST_SNAPBACK}][/a]

Same thing here... not very impressed with CS3 compared to what the leading products do.

I am also at a loss as to why Adobe is wasting time working on such specialized features but is telling us that CS4 for Mac will still not be 64 bits. John, come on.

Cheers,
Bernard

[Ray]

My experience also is that Autopano Pro does a better job than CS3. It's the only stitching program I've come across that seems to stitch almost anything automatically, including HDR in the same process.

It's necessary, however, to make adjustments in 'setup' first. I found the default settings to be a bit substandard.

[Quentin]

You might be interesting in knowing that as of PTgui Pro 7.8 they have implemented their own version of the infuse algo (HDR based on pixel level brightness evaluation in 16 bit space). This works pretty well per my first tests.

One recent sample made up of 36 Nikon D3 images (resulting image is 200 megap pixel). This was shot a few weeks ago in the Himalayas, the mountan pictures is Pumori at 7150 m. This was shot with a RRS pano head.

Thanks to its speed of operation and wonderfully stable shutter, I have found the D3 to be much easier to use for stitching purpose than the Mamiya ZD I also use.

Cheers,
Bernard
[a href=\"index.php?act=findpost&pid=193983\"][{POST_SNAPBACK}][/a]

Yes, Bernard, Once you are comfortble with a stitching workflow, then for many purposes, pxel quality and kit useability becomes more important that simple pixel count, because you get the resolution you need through stitching.  So I can well understand your liking for the D3 for panos.

In what circumstances do you still use the ZD for panos?  Carrying the ZD and lenses plus D3 and lenses would add up to a lot of weight.

I have used the HDR function with PTGui, works a treat.

Nice shot of the mountain too.

cheers

Quentin

PS It has taken me a while to get over the feeling that a stitched image is not "cheating" in some way, as if its not a real photo.  Of course, that idea is nonesense, but I can't completely shake it off for some reason.

[BernardLanguillier]

In what circumstances do you still use the ZD for panos?  Carrying the ZD and lenses plus D3 and lenses would add up to a lot of weight.
[a href=\"index.php?act=findpost&pid=194123\"][{POST_SNAPBACK}][/a]

Good question. Actually I haven't used the ZD for stitching a single time since I got the D3 back in November last year. This was mostly related to the poor cold weather battery performance of the ZD.

Now that temperatures have mostly risen to a more acceptable level, I'll have to think of some criteria to decide what I use when.  I guess that I'll probably keep using the D3 for more adventurous outings and use the ZD for more urban endeavours. Although the D3 has progressed a lot compared to the D2x in terms of DR, the ZD is still ahead.

PS It has taken me a while to get over the feeling that a stitched image is not "cheating" in some way, as if its not a real photo.  Of course, that idea is nonesense, but I can't completely shake it off for some reason.
[a href=\"index.php?act=findpost&pid=194123\"][{POST_SNAPBACK}][/a]

I can relate with that. I guess that with stitching you are sort of left alone with your shooting ethics in trying to create an image that is faithful to the scene you were looking at. My view is that there is some hidden process going on whatever the shooting method. The main difference with stitching is that you get to look at some of the magic and play a more active role in litteraly putting the pieces back together. Fundamentally though the same thing is going on. Technology helps freezing in time a scene and materializing it into an image that can be seen on various media.

The only philosophical difference is in the time span over which the image is captured, but stitching is not alone in this. All the scanning devices end up compressing into a seeingly instantaneous event things that did in fact happen over the course of a few seconds or minutes. Traditional photography long exposures also do this, but they are more upfront in revealing to the viewer that the static photograph is the integral in time of some non instantaneous happenings.

The underlying question though is still always the same. Does photography really need to claim that it is a litteral representation of reality? I think not. Once you agree that reality cannot be captured in a photograph, then the importance of a stitch being truely faithful to reality diminishes somewhat I feel.

I have been thinking of writting a piece on these aspects for LL for some time, but need to do a bit more research on the topic.

Regards,
Bernard

[dkeyes]

I can relate with that. I guess that with stitching you are sort of left alone with your shooting ethics in trying to create an image that is faithful to the scene you were looking at. My view is that there is some hidden process going on whatever the shooting method. The main difference with stitching is that you get to look at some of the magic and play a more active role in litteraly putting the pieces back together. Fundamentally though the same thing is going on. Technology helps freezing in time a scene and materializing it into an image that can be seen on various media.

The only philosophical difference is in the time span over which the image is captured, but stitching is not alone in this. All the scanning devices end up compressing into a seeingly instantaneous event things that did in fact happen over the course of a few seconds or minutes. Traditional photography long exposures also do this, but they are more upfront in revealing to the viewer that the static photograph is the integral in time of some non instantaneous happenings.

The underlying question though is still always the same. Does photography really need to claim that it is a litteral representation of reality? I think not. Once you agree that reality cannot be captured in a photograph, then the importance of a stitch being truely faithful to reality diminishes somewhat I feel.

I have been thinking of writting a piece on these aspects for LL for some time, but need to do a bit more research on the topic.

Regards,
Bernard
[a href=\"index.php?act=findpost&pid=194267\"][{POST_SNAPBACK}][/a]

Not to sidetrack the OP but I've been reading this post and you bring up an interesting (and often talked about) discussion.
I think part of the "issue" with photography is related and comes through how we talk about it. As an artist who uses cameras to create images, I don't seem to have this baggage. Where many of the photographers I know, think of this "issue". Photographers use words like "take" and "capture" as though the actual thing (see truth) will be documented. Artists use words like "create", "put together", etc. I like to think "I create nice images" versus "I take nice pictures". But I digress, could be an interesting thread if done right and not get into the is it real/is it art, etc. debate.

- Doug

[marc gerritsen]

Could any of you stitchers help me on this one?
with the first two images no problem at all, minus a small adjustment.
But when I try to stitch the blue photos in CS3 I seem to have a big problem
I shot this on a normal tripod like the successful ones.
Never mind the dust, virtually impossible to change lenses in beijing and not end up
with a wollop of dust on your sensor.
much appreciated
m*

[BernardLanguillier]

Could any of you stitchers help me on this one?
with the first two images no problem at all, minus a small adjustment.
But when I try to stitch the blue photos in CS3 I seem to have a big problem
I shot this on a normal tripod like the successful ones.
[a href=\"index.php?act=findpost&pid=194313\"][{POST_SNAPBACK}][/a]

Tried a bit with PTgui with carefully hand positioned control points... doesn't like it either. There is a tremendous parallax problem with your images.

I don't know what camera you shot this with, nor how well you had positioned the lens relative to its nodal point, but this is probably beyond hope.

Cheers,
Bernard

[Kirk Gittings]

have no trouble with landscape subjects. With (contemporary) architectural subjects, the smooth surfaces, continuous lines and repeating similar patterns seem to really throw the auto-stitching type programs for a loop. PT-based stitchers with manually-placed control points are the only way I've been able to deal with this type of image. Of course this is pretty time consuming for commercial work.

FWIW, I only shoot contemporary architecture for a living and do a couple of stitches in each shoot (averaging 6 stitches a week). I only use CS3 and have no problems with it at all.

[HarperPhotos]

Gidday,

Your right this was a tricky one.

Simon

[MichaelEzra]

Using AutoPano Pro 1.4.1 it seems very straightforward... I used Highest detection quality and it stitched perfectly. In general, AutoPano requires much less overlap than these images have.

[marc gerritsen]

Thanks guys, seems like the proof is in the pudding and I will have to give Autopano
a try. A while back I saw a demonstration of a stitching program in which they stitched a row of houses that could then be individually straightened with set up guide points.
Can you do that in Auto pano?.
Seems that Simon's shows more or less the same result I had in CS3 and Michael' really looks like it should,
thanks again
m*

[MichaelEzra]

Marc, to answer your question, yes, Autopano allows to quite easily straighten buildings in the panorama.

P.S.
For that sample render I used Smartblend blender which eliminated any ghosting in the final result, color correction to smooth the transitions (though they were already pretty smooth), and global rotation to straighten the horizon + crop.

[woodrowcampbell]

I use both Autopano Pro and PTGui - in my experience you really need both.  PTGui excels at preserving straight lines and architectural relationships. Autopano has a more intuitive interface and faster workflow.  Sometimes one will simply refuse to stitch an image which the other gets right.  I use them more or less on a trial and error basis.

In most cases you don't have to be super fussy about rotating on the axis of the entrance pupil - it only matters if there are foreground and background objects in multiple frames.  You can often plan your shot around this issue.  I stitch mostly handheld.

Both Autopano and PTGui give you more intuitive perspective corrections than PS.  One reason to stitch is that the larger number of pixels gives you the capacity to do more extreme perspective corrections.

Extreme perspective correction (3 M8 images PTGui)

[attachment=6484:attachment]

Foreground isolated in one frame (2 M8 images PTGui)

[attachment=6485:attachment]

PTGui choked on this one, even after I added manual control points.  Autopano handled it ( 2 H3D 39 images)

[attachment=6486:attachment]

[dustblue]

My experience is that ptgui is faster than autopano, both in editing and final rendering. For example the yaw,pitch,roll choice in autopano should be changed by numbers(so they call it numerical transmition..)and you see your result really slow after changing the numbers. while ptgui use just the mouse dragging, left button for yaw and pitch right button for roll. much quicker and easier.

I've been encountering problems with autopano when I try to blend a 16000*10000image, use smartblend, 3layers,which unexpected shut down every time, so I change smartblend to multiband and solve the problem. Anyway with ptgui I never encountered such problems.

I use both Autopano Pro and PTGui - in my experience you really need both.  PTGui excels at preserving straight lines and architectural relationships. Autopano has a more intuitive interface and faster workflow.  Sometimes one will simply refuse to stitch an image which the other gets right.  I use them more or less on a trial and error basis.

In most cases you don't have to be super fussy about rotating on the axis of the entrance pupil - it only matters if there are foreground and background objects in multiple frames.  You can often plan your shot around this issue.  I stitch mostly handheld.

Both Autopano and PTGui give you more intuitive perspective corrections than PS.  One reason to stitch is that the larger number of pixels gives you the capacity to do more extreme perspective corrections.

Extreme perspective correction (3 M8 images PTGui)

[attachment=6484:attachment]

Foreground isolated in one frame (2 M8 images PTGui)

[attachment=6485:attachment]

PTGui choked on this one, even after I added manual control points.  Autopano handled it ( 2 H3D 39 images)

[attachment=6486:attachment]
[a href=\"index.php?act=findpost&pid=194360\"][{POST_SNAPBACK}][/a]

[NBP]

OK, I'm a bit of a stitching newbie, so excuse me for what may be a stupid question, but I'm sitting here a little stumped.

I'm using CS3 merge to put these 6 images together.

[attachment=6887:attachment]

But it appears to be missing the 1st two frames of the series in the final result

[attachment=6888:attachment]

I'm having a good fiddle around, but can't seem to get it to use them.

When taking the shots I made sure each one overlapped the other by at least 1/4.

What don't I know / am I doing wrong?

Help much appriciated.

[Dustbak]

There seems to be minimal overlap between image 3 & 4.  Maybe that is where PS drops the ball. There is also minimal overlap between 2&3.

I would have taken a bit more overlap than this on each shot. You are making it very hard for PS to set proper connection points.

Did you take this with a wide angle? It appears you did. when looking at the different angles lines have between the shots that should end-up straight with each other in the final shot. When using wide-angles it is even more adviced using the nodal point to rotate.

When not using a nodal slider, I generally take a lens that doesn't distort much (eg. 50mm on 35 & 100mm on MF).

If I do use wide angle I am also struggling with my nodal slider BTW  

I have had this email conversation with RRS about the engravings on their nodal slider. They don't see the necessity to start counting mm's from the place where the sensor/film sits


edit: on second thought (look) you did not do this with wide angle most parts line up pretty well so PS should not have difficulties doing this provided you have enough overlap.