« Kodachrome Clearinghouse? | Main | Blog Note »

Friday, 15 October 2010

Comments

Feed You can follow this conversation by subscribing to the comment feed for this post.

How I wish Hasselblad would make a digital equivalent of the Xpan.

Dpan? I'd be first in line.

The stitching suggestion is a good one, and one I've been playing around with for a while, now, to the extent of purchasing a tilt-shift bellows.

I find a very useful program is this one (for windows and iMoan):
http://cvlab.epfl.ch/~brown/autostitch/autostitch.html
Autostitch periodically 'times out', but that just entails downloading of a fresh demonstration version -- which seems to do most of what I want it to. Though stitching images made using a short focal length, like 14mm (21mm-e) is problematic.

I've tried capturing a background, then subsequent 'patches' of motion for later montage on the background. Which works better with relatively stable lighting, and light to no wind.

A quick note on shifting lenses: Fotodiox sells adapters for micro 4/3 that shift. I first noticed them because many (all?) have integrated a tripod mount in the adapter: perfect for long, heavy lenses. Do check before buying if you are very interested in the tripod mount - it's not visible in that amazon link, but it is on the models listed on their site. Currently, $90 on Amazon, which is (now) on the high end for an adapter, but much, much cheaper than a real shift lens.

So, yes, the poor man's MF, on a pocket-able(?) camera.

I am hoping that someone might comment as to whether focus stacking increases the amount of real detail captured.

Will

Hey Mike,

Always a huge fan of the blog. This option for making high resolution images is totally viable and oft used, perhaps more so than you suggest in your post.

An example of mine, just on the off chance you're curious:
http://www.pbase.com/gdillon/image/44147924

It's twelve 6MP images stitched together, two vertical and six horizontal. Obviously a lot of tweaking going on. It's printed about 40" wide on metallic paper and looks fantastic.

Best,

Gabe

Medium format digital shooters figured this out already and are doing the same thing to simulate large format images. This image:
http://www.jimwitkowski.com/newWork/northCell.html

is simply three vertical shots with a base files size of over 300 mp and a native size of 28x18 @ 360 dpi.

The larger sensors found in most MF backs simply offer a greater dynamic range than found in 35mm cameras.

I use an expanded version of this technique, with a sliding back on a 4x5 camera and old Pictorialist lenses. My largest to date is shot with a D5000, 5 rows, 11 images each row, stitched with Autopano Giga software. Full size 22307 x 7803 pixels, approximately 174 megapixels: http://hemingway.cs.washington.edu/portfolio/pano/velo.html

Low cost simple sliding backs can be found for a few hundred dollars.

Mike:

Carrying one step further: imagine stitching files from a 645D.

Tom,
Check out Jim W.'s site!

Mike

Mike, I worked just the way you describe for 2 or 3 years (examples at http://www.hrichardpaul.com/pano/). using a high-end point-and-shoot and a low-end DSLR. I really enjoyed it, and enjoyed that I was able to do what I was doing without participating in the camera equipment arms race! And I got results I was very happy with from an image / print quality perspective (http://www.hrichardpaul.com/journal/new-images-exquisite-detail.html).

But there are some down sides to this approach, including:
- It's a real challenge, although practice helps, to visualize what a multiple-exposure, knitted image is going to look like when stitched and cropped. I had a tremendous number of captures where the parts looked great, but the whole was ho-hum, and I think it was partly because I couldn't really see the whole during the capture process.
- Because you're getting caught up in more complex technical details during capture, it's harder to maintain spontaneity with the subject matter. Another source of ho-hum captures...

After doing this for so long, and only this, I now find I don't want to do it at all, unless the subject matter absolutely begs for it. (Maybe the issue is that I overdid it! :-))

(By the way, even though I mostly did things with panorama aspect ratios, I also did experiment with 3x3 and 4x4 captures, just to get a higher pixel count from a small-sensor camera. Same comments apply.)

Gabe, Bruce, Jim,
Very nice--thanks. I never know when one of these little posts is going to turn up a lot of nice things to look at.

H Richard,
I think I know what you mean. A number of my attempts were not very good pictures. In some cases I just overestimated what the pano program could do. I'm also not inherently interested in panoramics--only when the subject seems to call for it. But the nice thing about that is that you don't need to carry any extra equipment with you--it's just a different way of conceiving the shot, and shooting.

The panorama in this post is a scene that's very meaningful to me personally, for instance. It's a scene I've been seeing, more or less, since I was a toddler. This was an interesting way of capturing it that's different from many other attempts and yet still feels true to the place.

Mike

Mike - I'll throw my example into this. Nature photogs that I'm familiar with have been doing this for a long time. And of course there are several kinds of pano adapters for taking care of parallax if your subject is close. This one comes in at about 100 megapixels, from a 10 megapixel camera. (Including overlaps it was 27 shots).

http://www.naturephotographers.net/imagecritique/largephoto.cgi?ref=148898

I've have a pano around here somewhere of a professional tennis match where the players are only in single frames so they came out well. The crowd shows some ghosting and movement in the various shots though.

I just did a big print of two stitched landscape shots from a gf1 with a pentax manual 135/2.5. Beautiful. With a long lens You really don't need that much stitching talent if the shots are taken on a tripod.
And really, you're not going to use a loupe on a 1 meter wide print. And there's not much you can tell about the stitching without using one.

David,
Where is that? Looks like Vermont.

Mike

Landscape master and longtime digital holdout Jack Dykinga does this with a Nikon D-SLR to mimic the results he got with his 4x5. Interesting to hear someone so pro-film for so long talk about what he loves about this technique. http://www.outdoorphotographer.com/how-to/shooting/transforming-large-format.html

My take on a two pic stitch with the GF1.

Mike my $2,000? Well I am a cheapskate, I buy used, so was thinking of a secondhand 5D MkII, but of course a 5D MkI can be had for even less. I actually use a 1Ds MkIII that I got secondhand, 1 year old and a little over half list price. I got my 17mm TS-E secondhand too, another bargain at well under $2,000.

Like you say, the post has thrown up all kinds of interesting, and beautiful, images.

I like that my two stitch images are quick to take but keep me slow and thoughtful. You are quite right in that most times you don't need a tilt shift to avoid paralax, I have a nodal/entry point rail that I can line up either way. Into the image to get the entry point so I can pan, or across the image to allow me to slide the body whilst keeping the lens stationary. But most times neither is needed. The true leap that the TS gives me is that endless and infinitely targetable depth of field, it has elevated the images from serendipitous to ones that are created, so much closer to view camera images than even my 6x9 Mamiya. The additional detail they have over the native resolution, and the control over aspect ratio, are just the icing on the cake.

With my Canon point and shoot. Eight shots and I suddenly had a fish eye lens.

http://bstinshoff.aminus3.com/image/2010-09-14.html

Personally I think it would be entertaining to figure out how to do action shots this way—you'd just need to figure out how to compose a picture of two or more frames that have all the action in just one of the frames. Could be fun.

Mike, have a look at Douglas Brown's web site Toronto Wide. He's written articles for Luminous Landscape about panoramas, including using the technique for fashion shows. His photos or concerts, plays, and other events are remarkable even without knowing that they're stitched images.

Hi Mike,

If that pano is gorgeous big, why not use it for one of the print sales on TOP?

At very least, you can get one of the prints out of the sale for yourself!

--Ned

This is exactly what I used on a recent assignment. It was a swarming p/r event with people jostling everywhere and I needed a shot to get to the client within an hour or so. This is a two shot stitch, hand held, taken in one of those 15 second windows of opportunity - and the client had it on the web and out to the media that afternoon. There are definitely "nasties" in it, but it worked for the purpose. http://williamstickney.smugmug.com/photos/1049709891_XieQk-L.jpg

Or do what George Eastman's ghost would have wanted: scanned XPan Tri-X! Action, not a problem. At 10000x4000 or so scan from a 4000 DPI scanner, it has plenty of resolution. Best of all, it's Tri-X. Need I say more? :-)

On the poor man's side, I'm surprised that no one did mention Hugin so far - a very powerful yet user-friendly, open and free stitching software. Dont acte.

My few cents of 'classical' stitched horizontal panos are at http://nikojorj.free.fr/gal/panos/ - the thinnest one looks quite nice hanged in the bedrooms' corridor, at 2.20m*25cm (should be 7'4"x10").

You want action? We have action. Sort of.
http://www.blipfoto.com/view.php?id=30193&month=5&year=2007

Hello Mike

I have also been using this stitching technique for a few years with a Canon 5D. With careful technique you can get impressive results, check out this single picture of a ceiling of the "Joanina Library" at the University of Coimbra made up of 360 frames taken with a 300mm lens (http://figaro.fis.uc.pt/joanina/gigatour/pics/sala3/tecto/9017632-991.html).
My recommended programs are AutoPanoPro and PTguiPro.

I have also been using multi frame for enhanced resolution (PhotoAcute) and Focus stacking in macro photography (HeliconFocus), and a combination of both, and you can get very impressive, and big, results.

The thing is that once you get pictures as numerical data, there's so many possibilities with mathematical manipulation that the things that scientists did beforehand as a necessity will trickle down to us consumers and pros to expand our photo tools.

Regards
Paulo M

I've done quite a number, in all sorts of configurations just as described. Plus multi-shot HDR type stuff together with stitching to increase dynamic range. Once you get comfortable with combining multiple images, the options are boundless.
My largest file: 26 stitched together. Reduced in size it prints 2m wide (about as big as I could get framed).
Maybe I ought to gather my stitched stuff into a single gallery online somewhere.

Jack Dykinga described a similar technique that he says rivals 4x5 in the January 2010 issue of Outdoor Photographer.

http://www.outdoorphotographer.com/how-to/shooting/transforming-large-format.html

Odd that noone has mentioned Hugin, so far. It's not only amazingly good at stitching hand-held shots, it will even create HDR or fusion "panos" (or call them whatever you like) from bracketed sequences. And it's free.

http://hugin.sourceforge.net/

Ralf

For Nikon shooters who own a t/s lens, this horribly named Agnos "JMBS Jumbo MultiBigShoot" adapter looks interesting to me:
http://www.agnos.com/prodotti.htm?v_lingua=ENG&v_iss_web=0000000010101613091225234003&v_categ_lista=P0000-P0025
It enables the body, and thus the sensor, to be shifted instead of the lens, maintaining the position of the entrance pupil, however I have seen reservations expressed about the strain imposed on the mount. Which seems odd as most tripod mounted long lenses tend to have the body hanging off them as well.
A word of recommendation for PTGuiPro. As a long term user I think it's hard to beat - although Autopano has many fans too. Stitching in PS I could never get on with.
Roy

Mike, I've found a very good panorama stitching program called Hugin (http://hugin.sourceforge.net/). It's a free / open source program, and there is a Mac version. It will do everything automatically (though there are extensive controls too), just given a set of jpeg images (preferably with EXIF data, so it can get lens data). Despite looking at 100%, I haven't been able to see the joins, except when I didn't keep a constant exposure! :)

I've been doing a few panos, hand held with my Pentax K20D and DA40/2.8 Ltd. Here's one of my first efforts, 9 vertical shots (warning, nearly 9MB image!) http://www.argaty.org.uk/images/LwrMiltonPano1A-I_bib_a_cright_.jpg

Alex

I, like many of the other posters, having been "doing big on the cheap" for years. Wanted to give a "heads up" on the european AutopanoGiga and AutoPanoTour products. The former is IMHO the most fantastic stitcher available - and yes, you have to pay for it. But it can now handle images shot hand-held even with modest perspective shifts. Handy when you're not as steady as you thought without a tripod. All my overseas shots are done without a tripod and Autopano "eats em up". And the "Tour" product packages them into great flash modules (look in the 'immerse' menu of my site if you are interested in what the tools deliver). Amaaaazing tools. And no - I'm not paid to say this - their stuff is just unbelievably good. (Uses your multiprocessors and GPU to the full too!) Try and Enjoy!

There is a Spanish photo artist cum digital photo guru, Jose Maria Mellado, who has been doing precisely that for ages and most of the time with cheap point and shoot cameras. He explains the technique in his books: If you can't get all the megapixels you need in one shot, keep shooting till it adds up.

I don't particularly like the results (too "beautified" for my taste) but I always thought he is very clever.

link:

http://www.mellado.info/

In Spain we refer to these manipulated, gross, landscapes as "melladismos"

For a compact my Ricoh GRD is great for these sorts of stitched "panoramas" - PASM dial, two jog wheels for dialing in aperture & shutter speed quickly, snap focus (or infinity focus) and White Balance modes in the quick menu... very easy!

Here's a 15-or-so shot "panorama" I hopped out of a car and took with it...

Stitching is great to get a wider view or to expand resolution. However, there are some limits to this techique and I think the one most limiting is arranging the picture. Because you do not have the whole picture in the viewfinder, its hard to arrange things. An additional drwawback is capturing moving objects. I'm not talking about race cars here, but even waves and moving sailboats are hard to stitch. Taking pictures of the coast will quickly show the limitations to stitching. But nevertheless, I also use this technique and think its great, specially when using a camera with a fixed focal lens (M8 with 35mm lens):
http://www.berndmargotte.com/panoramas_en.html
And yes: the Digital XPAN (D-Pan) would be great.

Click on Sam Kittner's

"Interactive Panoramas" link at


http://portfolio.kittner.com/

Microsoft ICE (Image Composition Editor -- FREE) lets you create huge panoramas effortlessly. And you can now publish them to the Photosynth web site where the user can pan and zoom from whole panoramas to minute details at high resolution. Also FREE.
http://research.microsoft.com/en-us/um/redmond/groups/ivm/ice/
http://photosynth.net/

ICE panoramas can be saved in several formats for printing. Photosynth is the best method I've seen for viewing extremely large, high resolution images on line.

"although I keep meaning to find and hire someone who could do the "merge" successfully"

Let me recommend my friend Mark who runs http://www.panoramaprinters.com/
He regularly merges 40 to 60 full frame files and makes ginourmous prints.

Currently, he's working on wet scanning and merging a two slide panorama that I shot with my Mamiya 6 and 50mm lens (about 50 x 100mm cropped). It's of a full rainbow spanning the sky at sunrise at treeline, with the clouds lit by alpenglow. Would have been impossible to take with a panorama head or using the tilt lens. And a tiny 35mm "full frame" sensor wouldn't have done this landscape justice.

I'm quite interested in the subject of merging images to simulate a larger sensor. When I was in France earlier this year, I made this image:

http://i.imgur.com/enbHx.jpg

from 27 photographs taken handheld with an inexpensive Canon 85mm f/1.8 set to f/8 attached to a similarly inexpensive Canon 450D. The final image is cropped to 10800x7200 pixels, around 78 megapixels. Here is a 100% crop from the left hand side of the cloud to give you a sense of the scale of the final image:

http://i.imgur.com/FZWy3.jpg

I've been messing around the the sweep panorama function on a Nex 3 and it just goes to show how far technology has come in image processing. You can get pretty impressive panoramas just waving the camera around. Not mega-sized images, but for the average boob like me, pretty cool stuff.

In my humble opinion, stitching is not the same as shooting one picture.
When you stitch you don't really compose the image through the viewfinder, you just take a lot of shots, assemble them later and then check if they're good. But most people get stuck at the technical side and think just because they now have a high res image it must be good. But in truth, most stitched images just don't look good.
I used to shoot a lot of stitched panoramas, but often I just wasn't happy with the results. Sure there was the occasional exception where it did work. But I usually stick to the regular 24x36mm frame these days. Content is way more important than resolution anyway.

If you want a larger format, just use a larger format.

I can recommend Hugin [1] as an open source stitching program. Been using it for several years to do my panoramas (partial [2] and spherical [3]). Also, by using Enfuse a very large captured dynamic range (via bracketing) can be effectively combined for quite realistic results by exposure blending with no need for the creation of an HDR image and tone mapping.

For good results, a panohead which allows the camera to be rotated around the point of least parallax [4] helps a lot but if the nearest object in the scene is fairly far away then handheld panos can work fairly well if a bit of care is take to rotate the camera around the front of the lens.

I thought it quite elegant that Sony turned a bug into a feature with the 3D sweep panoramic mode. In normal panoramic mode, parallax is minimised by stitching adjacent frames taken quite close to each other and only using the central vertical stripe of each.

Two create the 3D panorama, you want to record two images, each of which have significant parallax compared to the other. Assume we sweep left to right with the camera at arms length. The lens will describe an arc. Consider an object in the middle of the scene in the near-middle distance. At the start of the sweep, this object will appear at the right side of the frame but be taken from a position left of where you are standing. Conversely, at the end of the sweep, the object will appear on the left hand side of the frame by be taken from a position to the right of where you are standing.

So we can construct two panorama stitches, the "left" eye picture by stitching together a strip of each frame taken from the right hand side of the series of photos as you sweep, and the "right eye" picture taken from the left hand side of the frame [5].

References:
[1] http://hugin.sourceforge.net/
[2]http://www.dkloi.co.uk/Photos/Landscape/photos_landscape.html
[3] http://www.360cities.net/profile/daniel-oi
[4]http://michel.thoby.free.fr/Fisheye_history_short/Beyond-the-pupil.html
[5] http://groups.google.co.uk/group/hugin-ptx/browse_frm/thread/9530e4500fc36933?tvc=1&q=3d+panorama I later saw an explanation from Sony which confirmed the basic idea but can't find it again.

Mike,

Have you tried a merge with one of the latest versions of Photoshop? Perhaps I am lucky, or just not very picky (note: unlikely), but I am surprised at just how well it merges multiple frames. I'd be happy to have a go at merging some of your A900 shots for you; just drop me a line.

Andrew

Paulo M,
Very impressive. It strikes me that that must make a great study tool for art history students--they can examine the ceiling in more detail than they could in person. I remember when I was in Rome seeing students with binoculars staring at the Sistine Chapel ceiling.

Mike

Thom Hogan talked about this in 2008 http://www.bythom.com/hireztoday.htm ...

Mike - That's Shenandoah National Park from Skyline Drive.

"Sam Kittner's 'Interactive Panoramas' link"

His color is just...wow. Words fail. Painful to look at for me. Like the visual equivalent of fingernails on the chalkboard. Makes me physically feel like turning my head away from the computer.

Mike

"Need I say more? :-)"

Richard - Maybe? The issue there is that no matter how high a number of pixels you scan at, you're not adding any extra detail, just resolving finer grain/noise. The film has only captured so much detail. So your 40 MP image won't actually have the detail you'd get from a 40 MP digital stitched image.

for poor man's digital MF i'd consider shooting MF on film and scanning it. Once you get into it, it's actually quite a bit simpler than stitching digital images. Allows for shooting action and whatnot. I'd wager a modern negatives (say, ektar) should easily trump the digital when it comes to resolution and especially dynamic range... There's also a kind of wowfactor you can get from a 40 yrs old east german Sonnar 180 2.8 that really can't be matched by anything you can pull out of the so called "full frame" - and all of that without the fuss of panostitching...

Hi Mike
I regularly stitch A900 24 MP files and currently have 4 x 2 mtr prints thats about 13 feet in old money and with CS5s content aware fill it is a breeze.
I never use a tripod just take care when shooting.

David

I've been stitching for a couple of years now, primarily with landscapes, and only in single rows. I've only done panning, not shifting, so I don't know if that's any different.

I will echo some other people's comments that arranging the picture is difficult. There have been any number of times that I'll come up to a big scene, think that I have a good framing, and then come home to stitch only to find out that the picture doesn't have a great composition or... internal rhythm, for lack of a better term.

Honestly, I'm getting great resolution (enough for some truly huge prints), but I'm starting to feel like my composition has taken a step backwards. Too many of the stitches end up relatively "flat" visually. I sometimes have the sinking feeling that I was a better photographer when I was working with single-frame images.

You can see a rough framing in your head and scan back and forth to determine the vertical height, then start shooting from one position on the left and then keep shooting working to the right until you're done... but that doesn't show you where each element is going to be in the final result, or how two or three elements work with or against each other.

I have not done so yet, but just a week or two ago thought that I might start carrying some of the old cardboard L-brackets that we used to use to help visualize cropping on prints. That way I could frame different aspect ratios in the field and perhaps better see what's going to come out.

But wow, the resolution is addictive...

I have a 24mm tilt and shift but I've never stitched shots together. I used the T&S for this shot, I never thought about making two separate shots for it, didn't really need to but it's got me thinking alright

Photo & Video Sharing by SmugMug

Mike,
As others have said, there’s some really nice stuff here. Like many other posters, I've been doing this for several years now. On the personal front, one of my most popular images is Aspen Basin, Sangre de Cristo Mountains--stitched from 18 verticals (3 high by 6 wide) shot with a Nikon D2x and 200mm lens. The resulting image is ~120 Mp, and it prints very nicely at 60" wide. Knotholes are visible on aspens a quarter-mile away!

On the corporate front, I shot a sequence of an experimental aircraft on take-off for its maiden flight, panning (or course) at 9 fps. By choosing images that had overlap in the background (the aircraft was in every image), I created a long panorama that showed dozens of aircraft--actually a single aircraft in dozens of positions--against a single, continuous background. Management liked it so much they had it printed 18 feet wide, and mounted at "command central" in our shop!

As for action, I also shoot jazz performances locally, including panoramas of the performers and audiences. The way I deal with movement is to shoot several frames at each camera position, and then to choose which frames to use for stitching based on which ones have the best match in terms of peoples' locations and positions. It's not perfect, but it gives me lots of choices, and it frequently works quite well. I’m guessing this approach wouldn't work at all for something like soccer or football, though.
-gkf-

Dear Mike,

Unless you are explicitly trying to make a photograph that people will want to pixel-peep, resampling a 40"-wide photograph up to some magic number like 360 ppi is a waste of time. The improvement in rendition produced by said magic numbers is extremely minor (truthfully, I would call it entirely negligible). I never do it even for small photos; I am in no way convinced that resampling an existing photo to get it to some magic number really improves it. And when you're at 40" wide, people just won't be looking that closely!

The only time I upsample is when the resolution is going to be so low that pixelization will be visible in the print.

~~~~~~~~

Dear Bernd,

Good stitching programs, like the latest versions of Photomerge in Adobe software, are not the least bit fazed by moving water. The pano of water flowing over a dam in my last "True Fans" is a three-photo stitch. It is seamless, take my word on that.

The first demo I ever saw of this software was a Joe Holmes 3x4 (or maybe 4x5) stitch of the Oregon coast, waves and all. It's flawless. This stuff is scary good.

pax / Ctein

Re: your stitching needs. We don't usually think of Microsoft as a source of terrific photo software, but Microsoft Ice (Image Composit Editor) is some kind of stitching miricle and it's free. You can get it here in either 32 or 64 bit versions: http://research.microsoft.com/en-us/um/redmond/groups/ivm/ICE/

Will, thanks for the tip on the Fotodiox adapters! Given how small the m4/3 sensor is compared to the imaging circle of a standard lens, this sounds like a great way to make images that approximate what you'd get from a 40+ megapixel full frame sensor. It's less good for getting super-wide-angle shots, of course, but it does finally give me something to do with the OM 28mm that I got super-cheap a few months ago.

Mike, I would guess that DC group picture could have been done with a Cirkut camera either by Ed Segal or a member of the Ivey family.
As several people have pointed out, shift lenses are little advantage for stitching unless you move the camera around the nodal point of the lens. A simple Arca Swiss style clamp and rail (or L plate) could let you do that on a tripod. Or just get a nodal bracket, the Nodal Ninja seems to work well enough. And stitching software helps, I use PT Gui but there are several good ones out there I would guess.
I have shot for 30 years with a Cirkut, and Roundshots etc., and stitched digital definitely has it's place.

"resampling a 40"-wide photograph up to some magic number like 360 ppi is a waste of time"

Ctein,
As I say, it's purely a coincidence. My feeling is that the picture would look best at a bit more than 3 feet wide, and the file just happens to be 14,600 pixels wide. Nothing more to it than that.

Mike

Love the Tony Collins lawnmower pano

After reading many of the comments, I must agree with Richard. I took a wonderful shot in wide angle of Oxtongue lake near Algonquin park at sunset. (http://johnroias.smugmug.com/Landscapes/Landscape-Impressions/6106943_s6ys4#P-31-10)
When I got home I thought as Mike did, why didn't I think of shooting many single verticals and stitch the whole scene. I could have got a breathtaking pano. Well my answer is because I was caught up in the moment of the light and the cloud formations. And that's what really matters. Had I hesitated and thought to change to make all the necessary manual settings for a decent result, I probably never would have got the image as I desired it, or worse got run over on Hwy 60!

JMR

Dear Mike,

I got that. It was the last line-- "no upsizing needed" that caught my attention. Upsizing is nearly never needed, nor justified.

pax / Ctein

A handheld .3 gigapixel experiment just to see if I could get it to work.

http://www.gigapan.org/gigapans/fullscreen/48600/

It's the view from the end of my mother's driveway. 50 years of looking at it and there is vastly more in the photo that I've ever seen. The main problem is that you end up with a depth of field that goes from a half mile away to about ten miles away without stopping down too far for sharpness, and even on a cool April morning the air is kind of turbulent. The windmills in the background are between 10 and 20 miles away

Another useful technique is the concept of Vertorama. Where you get everything in focus from the foreground to the back ground.

http://www.flickr.com/groups/vertoramas/discuss/72157602953001077/

Another twist in the tail...
http://www.williamneill.com/blog/
look at sunflower and a bit further back in the archives a black and white mushroom.

Instead of shooting lots of images shoot 2 or 3 to make a square image like William who uses a tilt shift lens or shoot 3/4 vertical images to give you a massive file but normal 3 x 2 format, I have been using 85/135/200 on A900 to create panos that are of wide view but have that compression feel of tele to each section when viewed at closer distances.
One downside which I am sure someone will have mentioned is that even with a fast rammed up Mac you have to go and make a drink/meal or knit some fog whilst 6 or 8 24MP images are stitching.
David

Two points.

One: it's not just image-size/quality, or aspect-ratio, but DoF. Context and result. I stood close-up (3ft away from a scene about 8ft square), set a zoom to 40mm, took 14 frames in order to get narrow DoF (medium-format-style), and the result is the same size and at least as sharp as I would archive from a scanned Hasselblad frame (6000*6000px). And no messing around with the smell of fixer or waiting for film to dry or scanning & spotting film, in the process.

Second: there's a question of the art of the panorama. Some landscape photographers operate on a philosophy of "preserve the feel of the place". For a still photo, you can compose precisely, placing elements where you will in the output from, in-camera, and the only cropping is to remove the surplus from around your 95% viewfinder. For a panorama, by far the easiest way to operate is to shoot a ton of data and apply the crop where you will - in hugin, there's an "optimal crop" that preserves the greatest area of image - after the event. So there's a risk of taking a huge unartistic "snap"...

For those who bemoan the composition issues with stitching, I would invite you to think about it as having to use a fresh set of eyes. It is a completely different experience when trying to envision the completed capture. Much like learning any new technique or getting accustomed with any new piece of gear, there is a learning curve. It's a fresh challenge--don't give up so easily.

Dear Tim,

I really like that last paragraph. It is my feeling that most panos I see were done because, well, panos are COOL. Which there is nothing wrong with at all, to be sure. But it too often makes them unmemorable and artistically trivial.

I do relatively few panos, compared to how much I like them (I definitely think they're cool); the subject has to be one where I can find something to say with a pan that is more than merely " cool" . It is nontrivial.

pax / Ctein

I'd like to comment upon pre-visualisation of a scene. It's just an extension of what an experienced single frame photographer does, one has to use one's judgement and experience to find the right position and shoot the correct amount of the frames. I find it easier to pre-visualise a spherical panorama as there is nothing to crop, you just have to stand the camera in the right place to minimise dead spots in the view, balance the distribution of objects in the scene and take note of light sources and shadows. For partial panos, it usually helps to run through the sequence, for example using a panohead, and check the potential crop boundaries.

I find it rare that a landscape scene is exactly a 3:2 ratio so even single frame photos require some visualisation of the crop.

You could just shoot film. I don't need to 'stitch' my Mamiya 7 panoramas together, and the 'sensor' is bigger than the largest of CCDs. Plus my cheapest high res camera cost me 800 dollars and has tilt shift built in, its called a Toyo.

Never tried this but I thought I'd give it a try. This is what I got with 6 pictures handheld. Not bad. I like to use a square format and with 10 megapixels to start with I'm left with just under 7 after cropping square. Using this technique with 2 pictures could double my resolution for squares. Thanks for the inspiration Mike!

If you want a larger format, just use a larger format.

Good plan: http://stevesmithphoto.webs.com/pano612.html

Also, although you can make your pictures taller, or longer, you can't actually get the 21mp of a Canon's sensor to resolve more detail within the frame. That's really the point of MF digital in my opinion.

Hi -

Gigapan, people, gigapan. Word.

I've been using one since the beta program three years ago. As I use Olympus, I didn't need to upgrade to the Gigapan Pro, since the Epic 100 does the job. The simpler Epic was recently on sale for around $300, which is virtually identical to my beta unit and handles many P&S, I used a Canon SX 100 IS to good effect (go to gigapan.org:

http://www.gigapan.org/gigapans/26093/

400 pictures, 92313 x 21338 pixels.

NASA commissioned CMU to write code for the Mars program that allowed a relatively small sensor with a fixed lens to take very large panoramas. This is being commercialized by gigapan, which is supported as well by Google. The stitcher program (and the uploader for the gigapan site) is simplicity itself: all it does is stitch, and if you've done things right it works wonderfully with no user intervention needed AND will run multi-threaded on your PC and Mac (and is one of the reasons I will upgrade to a 6-core system in the near future...).

Works beautifully, and when you buy the tripod head you get the software. Not perfect, but nothing is in this field. The price even for the Pro is very reasonable, compared to what you pay to the tripod head makers for a manual system that takes longer and requires around 4-5 times the time. The longest gigapan I've done took around 1 hour, largely because it was a night-time picture (that one of Pittsburgh) and I was running 4 sec exposure per frame. Otherwise I reckon no more than 15-20 minutes, once I've figured out the picture...

Here are a few of mine in non-traditional format (i.e. not extremely wide): I consider it to be a large-format camera replacement for landscapes (I've done a few portraits, but that's another story for another day...), but lacking, of course, swings, tilts and all that other fun stuff.

35470 x 32368 pixels:
http://www.gigapan.org/gigapans/25686/

14428 x 23700 pixels:
http://www.gigapan.org/gigapans/45137/

and the Half-Dome at Yosemite in B&W:
16360 x 24352 pixels
http://www.gigapan.org/gigapans/45176/

More traditional panorama format:

64708 x 16740 pixels:
http://www.gigapan.org/gigapans/55449/

and this one is one of my favorites:

Yosemite from Glacier Point:
63472 x 19520 pixels

http://www.gigapan.org/gigapans/44667/

Seriously: if you are working with a tripod, this is **the** way to go. It makes it all very, very simple and lets you concentrate on composition and getting the photo done, rather than futzing about and maybe, maybe getting enough coverage.

Sure, you need batteries and all that. But a couple of sets of decent NiMH and you can do gigapans all day. My record is 17 in one day, in NYC...

Disclaimer: no connections, professionally or financial (or any other) to gigapan. Just a very, very pleased customer that desperately wishes he had more time to actually get out and get more done. :-)

Wow, thanks for the shout-out, Jonathan! I've been a long-time TOP reader, and yes, I've been doing this a lot, specifically with wide-open fast primes to get looks that are pretty much impossible otherwise. Shooting with a 6x6 frame is great, but imagine shooting a 6x6 with a 135 f/2 mounted on it. That's the basic gist of what I've been getting up to for a few years.

We also wanted an medium-format digital for aerial photography, but can't just afford it. But we did a merge of 350 shots from a Pentax K-7 + 14mm lens using PTGui to form an aerial mosaic - ran quite nicely on an Intel i7/Intel X58 box with just 3Gb DDR3 RAM. Total file size was 4Gb from 350 downsized 3MP JPEGs. Anymore resolution more than that and the computer freezes. Until we can get a computer powerful enough to handle those Pentax 645D files, there is no use shooting with it.

http://70gigapixel.cloudapp.net/index_en.html

Dear Mark,

My film-made panoramics are some of my favorite photographs, but I'm afraid I can't support your arguments. For a start, the film area in your Mamiya cropped down to panoramic formats isn't larger than what you can get out of the DSLR –– remember that several frames are being stitched together in digital!

For another... well... to use an old phrase, “it's not the meat, it's the motion.”

It isn't how big you are, it's how well you perform. Trying to compare film areas to sensor areas makes no sense whatsoever. It doesn't tell you anything.

Let me point you to a panorama I stitched together from three frames from my Fuji S100fs:

http://ctein.com/TOP/Dam_1_Falls_Pano_half.jpg

Understand that this is only half-resolution. I figured that the full resolution version would be way too much for people to want to download.

You can't do this with a 6 x 7 format camera. You can't even get close. If you're really, really good, you can get to about this level of quality with a 617 format camera. But now you're talking real bucks. (There are cheap fixed-frame panoramic cameras on the market, but their mechanical and optical performance is subpar and they don't belong in this conversation. Ditto for the swing-lens designs.) In addition, to enlarge that size film you're going to need a 5" x 7" enlarger and an appropriate lens.

Also, for a panoramic of this nature, 8" x 24" is really the smallest I'd want to print it to get the artistic effect… And that's the biggest I can print in my capacious darkroom.

This does not become a cheap alternative.

And still the panoramic I stitched together from this $500 camera with a sub-centimeter sensor is as good or better, by most objective measures of image quality, than anything I ever saw (or made) from a 617 film camera.

That's not a dig against film. I think film looks just fine. So does digital. They don't look the same. Trying to compare them makes no sense.


pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
======================================

Also, although you can make your pictures taller, or longer, you can't actually get the 21mp of a Canon's sensor to resolve more detail within the frame.

Sure you can; use a longer lens.

(This solution is not Canon-specific :-) ).

Maybe it was mentioned in one of the many posts, but I have not seen a good tutorial and using tilt-shift lens for stitching panoramas. Does anyone have any suggestions of where a tutorial can be found online.

Dear John,

Does a tilt/shift lens make a difference in the kind of pano one can produce? With the various mappings and perspective correction capabilities built into the merging programs, even Adobe's Photomerge, I'm having trouble visualizing the geometries involved, so I can't figure it out myself.

I've got a book on Pano photography from Rocky Nook, but the author doesn't appear to discuss this (at least, there's nothing in the ToC or index).

pax / Ctein

Dear Ctein,
As far as I have been able to tell the use of a shift lens interferes so much with stitching software like PT Gui and other Pan Tools (Helmut Dersch's program) derived tools that they are of little use. In theory you could use a shift lens to make a two shot stitch with the camera on a tripod, but you must shift on the lens nodal point, not with the camera fixed on a tripod. So it really turns out that a nodal bracket with stitching software gets you closer to the goal more quickly and cheaply.
I do this a lot so for a casual user the easiest path might different.
As for tilt, that's a different issue. Controlling the plain or conical section (frustrom I think it is called) is a nice thing to be able to do. My Cirkut camera gives me this control and I have used it in the vast majority of the pictures I have done with that camera. As a result I had a custom lensmount built for my Roundshot 220VR that allowed rise and fall AND tilt. But it's less essential in medium format and less important still on the Nikon D700.
But what is important to my shooting with digital is that I can change the focus from shot to shot and have the focus follow the subject. As you can imagine this is a fussy and time consuming process to actually make work smoothly, but worth the trouble to me.
Hope this makes sense and helps,
Doug Chadwick

The comments to this entry are closed.

Portals




Stats


Blog powered by Typepad
Member since 06/2007