Panorama techniques are useful to give a wide field of view in one direction (single row). If the resolution of the camera is not quite enough, you can do a single row horizontal panorama with the camera in the portrait orientation and stitch three or more shots together in Photoshop. You can approach 4*5 film quality with Multirow Panoramas.

If the subject distance is short, parallax is a problem, but it can be eliminated with panorama heads. However, panorama heads are not cheap. An RSS single row package is $360 and the multi-row setup is $795. If the subject distance is far, you could ignore parallax and get decent results with a simple tripod setup. Some photographers are even getting good results handheld.

My question is at what distance does parallax become negligible? Does anyone have any experience here?

Bill

Sir, at a certain distance, Sir!

It depends on many things - the amount of foreground and of recognizable details in it (ie easiness to mask the parallax errors in PP after stitching), the way foreground and background mix (view through close branches to a distant landscape being an example of the worst case), and the final resolution being a few of the prominent ones.

I'd rather see the main variable as the field of view of the lens used - the bigger the FoV, the more you'll need a pano head, with a turn around the "normal" focal length.
That said, I hate carrying a tripod and still get decent results handheld with my 10-22 (eq. 16-35mm in 24x36), most of the time without any masking trickery.

Btw, if you're on a budget, you may consider popular models such as Panosaurus or Nodal Ninja? It depends also on the camera/lens you're supporting...

Bill,
I'm amazed at the proficiency of some current stitching programs such as CS3's Photomerge and particularly Autopano Pro which I find a bit better at handling parallax errors.

My experience is that straight lines in the immediate foreground tend to be the most problematic regarding parallax.

The following scene stitched perfectly and automatically. It was a windy day and my porter, who was carrying my tripod, was miles ahead of me. I took 3 hand-held shots at F16, using the Sigma 15-30 zoom on my 5D at 15mm. The closest pebbles would be just a few feet from my feet.

I can find no errors, discontinuities or significant distortions in the stitch. Quite remarkable!

Click to view attachment Click to view attachment

This is not only the question of distance, but of the interrelation of distance and incorrect camera positioning.

You can shoot miles far away and see parallax errors if you walk around between shooting the frames. You can shoot hundred meter away from tripod without pano bracket and see no error.

It is important to understand, that close objects alone do not cause parallax errors. The problem arises, when close objects appear in several frames. The problem can be avoided often by framing: make sure, that the close objects are in single frames.

There is very much to discuss about it. I would not start out with a pano bracket. Start making panos, first avoiding difficult situation (typically, indoor is not realistic without a pano bracket). You will see, that many panos can be made even hand-held. Even in cases, where the forground is close (this occurs very often), the parallax error is often invisible due to blending. This works well with nature. A brick-like pavement is the enemy of the panomaker. I stitched a pano just a few days ago, from the Rockies, hand-held, 11 frames. The foreground was very close (within a few meters), but no problem, except the darn pavement (see http://www.panopeeper.com/panorama/Rockies...montChateau.jpg, I had to crop away the bottom part, but you can imagine, what happened there). Sometimes it is corrigable, often laborously.

ANyway, make some panos, and wait until you run into the parallax problem; then think about how to avoid it, and only later buy a bracket. It does not need to be that expensive. RRS is particularly expensive, and unreasonable: AFAIK the adapters are customized for specific lenses at specific focal lengths. Zooming ruins it. Plus, one has to know, that the entrance pupil of most lenses changes with focusing, so if you need that high precision, which is offered by RRS, then you have to look for an adjustable one.

If you are serious about it, then start out with a decent stitcher: PTGui (expensive), PT Assembler (only $39) or Hugin (free, but I don't know it). All these are using the father of stitchers, Panorama Tools.

There is an excellent forum at PTAssembler, dedicated to all issues around panoramas: http://www.tawbaware.com/forum2

I shot a pano of the Golden Gate Bridge from over a mile away at 105mm on a D200 just using a tripod and didn't break out the pano head stuff because the wind at that location causes a lot of instability for the camera. The result sucked even at that distance and the files were unusable even using Autopano.

I can not comment on that pano-candidate, but even using Autopano is upside down. Autopano is not for solving difficult panos. In other words, if Autopano can not stitch a pano properly, then a decent stitcher needs to be used (though that does not guarantee, that it can be stitched).

Hi,

I use Autopano and the "smart blend" option. I use a panorama head without parallax compensation. There is definitively parallax in the preview at short distances but it goes away in the stitching process.

I made a deliberate test with a branch something like 100 cm in the front of the lens in a pano that was made from almost a dozen pictures and could not observe any parallax related issue in the final result.

I actually have stuff for eliminating parallax but simply don't see the need to use it.

My answer in short is:

I did some tests at about 1 meter to foreground object. Parallax was non issue after stitching with Autopano Pro using the smartblend algorythm. I would definitively have problems without smartblend, howver. Smartblend is one of the blending options in AutopanoPro.

Best regards
Erik

What is that? The only real point of a panorama head (bracket) is the provision for positioning the camera around the entrance pupil. Otherwise every tripod is good (though ballheads are not very suitable for panos).

I bought a Manfrotto panohead years ago, before I'd bought my first DSLR. It's designed to accommodate MF cameras so it's unnecessarily bulky and heavy for 35mm format. It's as heavy as my 5D with 100-400 zoom. I've never actually used it. It seemed a good idea at the time because I was wasting so much time in front of the monitor trying (mostly unsuccessfully) to correct for parallax errors, restitching and repositioning flags with programs such as Image Assembler.

After experiencing the joys and sheer convenience of digital shooting, stuffing around with a cumbersome pano head, with a hundred adjustments to make, seemed a retrograde step. I've got better things to do.

Autopano Pro seems to be able to handle just about everything I give it, automatically as well. However, as Erik mentions, you need to go into set-up first and enable certain features for quality stitching, such as 'smart blend'. When I first tried Autopano, I was disappointed, not realising that the default settings were for the quickest, but not the best, result.

Ray,

Thanks for sharing your experience. The merge looks quite good. Was that PhotoshopCS3 or AutopanoPro? For most landscapes parallax is most likely negligible. Thus far, I have done only one panorama with CSPS3. It was on a tripod, but I didn't have the camera level, and excessive cropping was needed. The actual merge was quite good.

Bill

I had "smartblend" enabled; it still ended up with seams. I was using the Nikkor 70-200 VR, and what is odd about that is in my testing of the parallax on that lens the nodal point ended up being behind the camera, which was quite odd but when I asked Thom Hogan about it he said that was not unexpected -- so probably some lenses are more effective than others.

Bill,
This was an example where CS3 did an okay job but with some slight discontinuity in the horizontals at an overlap. Autopano did a better job.

I was particularly pleased that the glaring flare spot in the first image has magically disappeared in the final stitch. No retouching was necessary.

Bill, the camera does not need to be level. In fact, in many circumstances it can not be level. However, the plane of swiweling ought to be level.

A level camera could shoot only straigh horizontally. What about shooting a valley from the mountain, or shooting a tall building from the street?

The camera can be held in any position. In a multirow pano, it *has to be* held in different angles both horizontally and vertically. Swiweling on a level plane is useful for avoiding much cropping, but it is not a must.

Demo of the effect of not swiweling on a level plane: Devil's Garden in Utah. I was standing on top of one of the hoodoos and shooting a wide, single row scenery, about 265°, hand-held. When shooting wide sceneries hand-held, I swing back and forth several times looking through the viewfinder before clicking, to judge the correct hight and to avoid being led by the scenery, but this time I did not (the top of that hoodoo was not a prime place for shooting, but I needed the height).

The result is http://www.panopeeper.com/Demo/DGall_tinyDemo.jpg I had to crop away a large segment.

Panoheads need not be expensive, and can be quickly hand built for very little money by adapting existing, standard photographic equipment.




While neither of the heads above is perfect in regards to rotation centering, they both place the rotations more-than-close-enough even when nearby objects are included in the scene. Neither took more than 2 hours to buid, only tools used were a handsaw and a drill press and some sandpaper and some miscellaneous hardware bought a Lowes. The Gitzo G1270 tripod head used on the fancier one cost about $50 on ebay and offers a considerable amount of adjustment. The plastic tripod cost $4 at a garage sale.

Here's are some shots with the plastic one using a Fujifilm F31 camera. I carry the plastic tripod around in my trunk just in case the pano mood hits me unexpected.

http://unit16.net/forums/070517_dam_0189_web.jpg
http://unit16.net/forums/waiting_the_chief.jpg
http://unit16.net/forums/dscf1091_green_bldg_web800.jpg

Coming from a long time PTgui used, Autopano pro should be in that list, and near the top.

It delivers by far the best automatic results, PTgui is much much faster though (PTgui Pro 8.0beta6 is a bit unstable still, but amazingly fast).

Cheers,
Bernard

Hi,

You want your camera to be level, or at least rotate around an axis perpendicular to the horisont. For that reason a simple panorama head is most useful. Use set it level with a ball-head and rotate it around it's own vertical axis.

I can add a rail to my stuff that shifts the the entrance pupil above the rotation axis but I normally don't feel it makes any difference.

Best regards
Erik

No panorama head is required for that. Any three-way tripod head does it, only ballheads are not suitable.

Hugin is really good, I actually prefer its UI to APPP's one. The area where the current stable version is lacking is 16-bit images (I think some interesting things are in the beta pipes).

The long nikkor I find most suitable for Panos is the 70-300 VR. Its nodal point is basically localed right under the body at 200 mm...

That and VR make it perfectly suiable for hand held panos of distance subjects. It is also reasonnable sharp at f10 at 200 mm (but too soft for my taste at 300mm).

If I am not mistaken, this is how this shot was captured (this version is a little too light). Pano work done with PTgui Pro 8.0beta 4 if I recall.



Cheers,
Bernard

what are the pros and cons of Autopano Pro vs. PTGui Pro.

capability? ease of use? distortion and perspective control?

do not generally stitch a lot of images together, but results from handheld is important

the HDR capability of PTGui Pro sounds good (stupid question, but will it work on a single image?)

I should experiment a bit more with Autopano to see where its limitations are. I vow that on my next trip I shall take a lot more hand-held shots for stitching .

The following 4 images taken with a 24mm lens on a lightweight ball-head tripod look a bit of a mess. I was doubtful that Autopano could stitch them successfully. I expected that the paving in the foreground would not come out right. I was surprised.

In fact, the paving seemed at first glance to be stitched perfectly. Then I noticed a slight irregularity, but it's not a problem. I wonder if you can spot it .

These shots were 5 sec exposures at ISO 1600 using the 5D.

Click to view attachment Click to view attachment Click to view attachment

Ray,

it is a mistake to believe, that the capability of a stitcher is explained with successfuly stitched panos. It is just like when people are posting images demonstrating, how good their camera is at ISO 5,000,000, shot in excellent illumination.

The question is not what it *can* stitch but what it *can not*.

I tried the non-pro version for years ago, it sucked, and I have no reason to do anything now, for I am working with one of the decent stitchers. However, if I were to select a stitcher, I would check out following issues:

- can you specify horizontal and vertical lines?

- can you specify straight lines?

- can you specify, that the shots are with different focal lengths (or field of views) due to refocusing between frames?

- which projections are supported?

- can you intervene between stitching and blending (i.e. modify the

- and some more.

Agreed, but that is not something that you can assess only by looking at a set of features.

What you list here is indeed important for architcture, and PTgui is indeed probably still king of the hill thanks to its ability to specify vertical and horizontal lines.

There are however many images where it *can not* detect the connection between images because there are few discernable features or a low contrast. Autopano pro is significantly ahead here. The following image is an example where PTgui couldn't stitch 80% of the images in the pano, Autopano Pro did it without any problem.



There are also cases where PTgui just *can not* merge seemlessly images with moving subjects or parallax errors. Here also, Autopano Pro appears to be superior in many cases. The following image is a good example where Autopano Pro did a better job off the shelf (it wasn't perfect though):



If you are perfectly happy with PTgui, indeed no need to look elsewhere. If you see some cases where it doesn't do a perfect job, I have a hard time understanding why you would want to deprive yourself from using a tool as brilliant as Autopano Pro.

Cheers,
Bernard

Not for architecture but for everything, where there are clear vertical lines, like city scapes. It is a common error of amateur panomakers, that buildings look like build by drunkards.


Not PTGui but Panorama Tools. I am using PTAssembler, which too is based on Panorama Tools.

Althout there are neu "editions" (copycats) of Panorama Tools, they are mostly recodings. Some new projection methods too have been implemented (PTMender).


Had the clouds been moving, Autopano would have created a horrendeous result (like all other automatised processes). PTA too can create a pano automatically, though it does not pay, in my experience.


Does PTGui not allow for interwention between stitching and blending? PTA does, and that is the best way to handle this.

There is NO blender, which can solve the blending of for example lightly wavy water surfaces; only manual intervention helps.

Furthermore, panos in the usually posted sizes don't show the problems. I work on my panos in 100% (and more), and when I see an error, I reduce the size to the probable print size to see if the error remains visible or if I can ignore it.

Cityscapes is architecture in my book. There are plenty of other subjets where vertical lines don't need to stay vertical. Besides, the question is mostly how easy it is to make verticals vertical. It can be done with most packages manually.



I don't remember seeing you next to me when I was shooting on that ridge at 3900 m, what makes you think that the clouds were not moving? They were most definitely moving. Again, I am talking about Autopano Pro, a very different beast compared to Autopano. Autopano pro uses proprietary Mac implementation of Smartblend.



Well it does through the multi-layer PS export capability. The various layers are exported after having been deformed for stitching purpose. You can edit the masks any way you want, which I had to do for the second image posted above since the leaves on the water were moving, just like the clouds were.

This was complicated by the fact that this 2 row panorama has infinite DoF thanks to a preliminary DoF stacking job done by hand using Photoshop. I am saying that it complicates the task because of the longer time lag resulting from the numerous exposures (32 for this image). This is one example of an image where a sheet of 4x5 film on my Ebony would have done a similar job in a fraction of the time.

This is one weakness of Autopano pro where the multi-layer export does include non deformed layers from what I could see.



Autopano Pro does again do a very good job here, perhaps not always perfect, but better than any other stitcher I have seen.



It goes without saying that all the panos I post in small sizes here are perfect stitches when viewed at 100% on screen. I do not release any pano that isn't pefect at 100% on screen, even on smaller print sizes where defects might not show. There is always a possibility that I overlook a small defect of course.

Anyway, I am done with this topic. Fine if you prefer to keep using the tools you are used with. As far as I am concerned I prefer to use the best available combination of tool at any given moment in time, be it a new tool from a new company.

I use PTgui as my mainstream stitcher, and Autopano Pro for difficult stitching jobs. Similarly I use different raw converters based on their respective strenghts and weaknesses depending on the constraints.

Cheers,
Bernard