Ad
Ad
Ad
Pages: [1] 2 »   Bottom of Page
Print
Author Topic: Digital back stitching on a 4x5" analog system - is it feasible?  (Read 2472 times)
torger
Sr. Member
****
Offline Offline

Posts: 1400


« on: February 19, 2012, 10:09:45 AM »
ReplyReply

I'm an amateur landscape photographer with some artistic ambition, but also obsessed with high technical image quality and large high resolution prints. So far I've worked with 35mm digital and done mosaic stitching to compensate for the lack of resolution, it works but lacks the attractive compositional on site workflow as a real view camera has. I've always kept an eye on MF digital, but when 35mm is as good as it is now (especially with the new D800 coming) entry level MF has lost its attraction in my eyes, and the attractive MF like P65+ on a tech cam is just way too expensive for my amateur budget.

So I have started to look into LF film. Film has some romance and a value by itself, but my main goal is to be able to work with a view camera and get high resolution within budget. I do realize that LF film with quality lenses and precision view camera won't be cheap either, but within reach.

But to the point - I'd like to have a digital option to complement the film. Obviously it won't be a one-shot solution, but used to stitching and shooting mostly static scenes I can live with it. I've in another thread got help with looking at scanning backs, and I've from that concluded that stitching use a large pixel MF back may be better for me. I've especially looked at 9 um pixel pitch backs with 36x48mm sensors like P25/P25+/DM22 etc, which now and then are sold used to fairly amateur-friendly prices.

I've noted that there are several "stiching backs" for 4x5" on the market (sliding back with pre-sets to quickly move the MFDB between the stiching positions), but the problem is that these are large and heavy (not intended for field use) and cover a smaller area than I want to.

My idea is to cover the full 4x5" effective film area, that is 94x120mm, which using a P25+ or corresponding means 9 images with +/- 29mm horizontal shift (58mm total) and +/- 36mm vertical shift (72mm total), and use the tech cam's native back movements for this rather than a stiching back. I'm looking at an Arca Swiss F-Metric 4x5" camera, which seems to provide a geared 100mm range vertical and 60mm range geared horizontal. By using most range for stitching movements, compositional image circle movements have to be mostly done with the front (with a 36mm offset - Arca F-line allows for rise only), but I think I can live with that.

There are a few unknowns around this:

  • How bad will lens color cast be, beyond correction? I don't like ultra-wides, so 90mm Rodenstock Apo Grandagon-N is the widest I plan to go. Stitching full 4x5" area is however more than stitching backs do, so I guess it will be a rather low angle of the incoming light so I do expect to have to correct LCC inpost, but if it is so bad that most dynamic range is lost in a channel then it won't be acceptable. High DR is an important aspect to me.
  • Will the analog lenses disappoint in some way I don't know about, like having gross chromatic abberations which does not show on film, or similar? I've read both that analog lenses perform well and that they perform really bad, hard to find information. Sure for 4 um pixel pitch sensors it won't be great due to resolving power limitations, but here I'm aiming at 9 um. But perhaps there's more too it than just resolving power and lens color cast?
  • Will the precision in the quite large movements be adequate, depth of focus will be say 0.5mm, so with those fairly large shifts the camera must be precise enough to keep the digital back in the film plane - is it unrealistic to get that from say an Arca Swiss F-metric support and frame? I'm aiming for high resolution here, so throwing away everything due to unprecise shifting would not be good.
  • Switching from ground glass (focusing) to digital back, can that be made with sufficent precision? With the focus techniques I use and near 100% large DOF photography I think I can deal with some slight focus shifts, but it must be reasonable.


Concerning lenses I'm currently aiming for Rodenstock which seems to have the focal lengths, image circles, quality and prices in the right range. 90mm Grandagon-N, 135mm Apo Sironar S (most important focal length), 180mm Apo Sironar S.

Stitching 9 frames like this will yield about 140 megapixels. To be satisfied, I'd want 100+ of these to be effective, that is not blurred by limitations in lenses.

Digital medium format really requires extreme precision to keep in the film plane etc, but one property of large format is that precision is traded with format size, that is lower precision requirements but still gain some resolution due to the much larger format. I'm hoping that those large 9um pixels (corresponding to 2800 ppi) and smaller apertures that increase depth of focus will make this feasible.

Has anyone any experience of similar setups or other valuable information?
Logged
theguywitha645d
Sr. Member
****
Offline Offline

Posts: 970


« Reply #1 on: February 19, 2012, 10:44:39 AM »
ReplyReply

I use a P25+ on a Linhof C679 view camera with Rodenstock lenses--essentially rebadged 90mm and 55mm film lenses to digital. Focus is easy enough on a ground glass. No superhuman task. I think the emphasis on how difficult to focus MFD is overblown.

With Capture One, it is day to get rid of lens cast--simply shoot LCC frames for the back position and aperture and apply them to the taken frames. I do get lens cast with both of my lenses when using a single frame.

I would try single frames with a P25+ before you start stitching. 125mp images sound really great until you think about the processing power and storage requirements you will need for the data (a single 40MP image opened to 16-bit is 226MB). I don't know how big you need to print (not that print size has any relation to pixel resolution), but I print p25+ image files off 44" large format printers. I see no reason to stitch the files for more resolution--a single frame is beautiful and there is more to image quality than the number of dots you put in an image and, no, a 35mm 24MP image does not compare to the P25+. I guess what I am trying to say is before you buy all your glass, first find out what the P25+ is going to do for you. It sounds like your stitching workflow is about pixel envy more than the realities of imaging nor experience with MFD. You may find the P25+ os an amazing back and you will want to rethink the glass you want.
Logged
Doug Peterson
Sr. Member
****
Offline Offline

Posts: 2767


WWW
« Reply #2 on: February 19, 2012, 11:09:25 AM »
ReplyReply

A Leaf Aptus 22 and a Cambo Wide DS with a one of the less expensive lenses and stitching 4-6 shots makes a lot more sense to me for what you're trying to end up with. You'd have to talk to one of our salesman to get specific numbers (I don't keep track of them) but this is the most affordable option that makes sense to me. Stitching inside the image circle with a system like that would give you 40-80mp with minimal effort.

I REALLY think you need to get your hands on a system like that to see what it's like in real life. If you do I strongly suspect you'll drop your quests to shoe-horn such functionality into an analog system that was not designed for it.

UNLESS you also are strongly interested in tabletop/product photography in which case the view camera could kill two birds with one stone.

Doug Peterson (e-mail Me)
__________________

Head of Technical Services, Capture Integration
Phase One Partner of the Year
Mamiya Leaf, Leica, Arca Swiss, Cambo, Profoto, LaCie, Canon, TTI, Broncolor & More

National: 877.217.9870  |  Cell: 740.707.2183
Newsletter | RSS Feed
Buy Capture One 6 at 10% off
« Last Edit: February 19, 2012, 11:11:53 AM by dougpetersonci » Logged

DOUG PETERSON (dep@digitaltransitions.com), Digital Transitions
Dealer for Phase One, Mamiya Leaf, Arca-Swiss, Cambo, Profoto
Office: 877.367.8537
Cell: 740.707.2183
Phase One IQ250 FAQ
torger
Sr. Member
****
Offline Offline

Posts: 1400


« Reply #3 on: February 19, 2012, 01:48:13 PM »
ReplyReply

Yes I'm probably a bit crazy Smiley

The idea to cover the 4x5" area is not only about resolution, it is also about using the same lenses, the same field of view etc. A typical scenario would be that I would first take a film exposure and then make a digital one of the same scene (if time allows). There's always a risk with stitching that due to movement etc it will fail. I also think that a successful film exposure will be more worth to the eyes of the audience than a digital stitch, oh well I know my own ego would grow when looking at a nice 4x5" transparency on the light table, if I ever succeed with one Smiley

I'm fully aware of the mammoth file sizes, I manage those already from my stitching work, and I have a quite impressive work-station to do the post-processing. My output is from a disk space perspective fortunately quite low. I throw away stuff that does not really cut it. I'm not that good and have limited time for photography. Last year I made about 8 images which I think had show-to-the-public-quality, and 3 which I made prints of and 2 which ended up on a wall.

A portfolio is built over many years though and I'm just starting. With eight images a year I only need 10 years or so to reach David Fokos production, in numbers that is... Smiley

Concerning print resolution, if I'm into "good enough", I'd surely get away with a 35mm system. Very few of those that view my pictures would notice and appreciate technical quality I do myself, it is all about content as we know. I've printed quite large from an APS-C camera and it looks good. I've never even heard someone complain about the overexposed cloud in the corner I see on that picture every time I look at it.

But 35mm does not (without stitching) give me resolution to the vanishing point. If it is not 400 sharp ppis to the printer, I want more Smiley. I'm thinking that one part of my brand/style should be to have many of those insanely sharp pictures at large sizes. If I get hooked on film, I probably try 8x10" too, one thing I find attractive about the Arca Swiss is that it is modular, an 8x10" frame can be put on it. An advantage of being an amateur is that I have the luxury of only doing things that I like myself and don't really need to care if a picture is economical or not, as long as my day job brings in enough cash to finance the hobby. Well, perhaps I'll grow up and get realistic about it, but now I'd really like to find out if this is possible. If I start compromising I'm very soon back to shooting 35mm digital.

Concerning "pixel envy", yes, to be honest there is some of that too. It would not feel great to do all this effort if an IQ180 on a digital tech cam would easily get higher technical image quality. It is some element of "if it is not on par with the best, why care?", the alternative to keep shooting landscape on 35mm is after all a quite economical solution.

But anyway you're probably right, this really needs testing before being sure if it works.
Logged
Ben Rubinstein
Sr. Member
****
Offline Offline

Posts: 1733


« Reply #4 on: February 19, 2012, 01:52:39 PM »
ReplyReply

http://www.camerafusion.com/

I had one, the stitched files were quite unlike anything I've ever seen before in my life. Does take a while to shoot with though, not for outdoor use with changing light.
Logged

BernardLanguillier
Sr. Member
****
Offline Offline

Posts: 7874



WWW
« Reply #5 on: February 19, 2012, 04:25:06 PM »
ReplyReply

What are the advantages you expect to gain with 4x5 + back stitch vs spherical stitch with a D800(E)?

Cheers,
Bernard
Logged

A few images online here!
torger
Sr. Member
****
Offline Offline

Posts: 1400


« Reply #6 on: February 19, 2012, 05:15:11 PM »
ReplyReply

For the record, I have looked at P25+ raw, and yes it looks nice, and yes they are superior to my 5Dmk2 mainly in DR and probably in colour. To my eyes it is not a very large difference though, with proper post-processing and the best lenses the 5D does well. I would not dare to tell the difference between a 5D and P25+ image in a finished print both executed with excellent technique. But I have no problem seeing the difference between 150 and a 400 ppi print, assuming it is a detailed image and I get to look up close as I always do when I like a picture.

I think the "MF superiority regardless resolution" reputation comes from excellent default color handling for typical professional work (portrait, fashion) and 35mm history of poor DR (which especially Nikon via Sony Exmor has fixed), I do respect that for some users MF format is vastly superior, but from my point of view low resolution MF is not really a strong case.

But I have still looked at it (I think I've looked at any possible combination of systems) and there's another problem - the lens problem. With that small 36x48mm sensor I would need digital complex lenses that are considerably more expensive than the simple low-element-count designs that works for large format. Instead of 90mm for about 1800 I would need a 32mm for 5100. Instead of 135mm for about 1000 I need a 50mm for 3000. In the "starter-kit" the 90 and the 135 are exactly the lenses I would get.

The 300mm lens is a bit painful to get for LF (3000) but in the range 90 - 200, which is the most important range for my shooting style, lenses are very affordable -- an important advantage of 4x5" format as I see it. To be able to use those lenses with my desired FOVs I need to stitch the 4x5" area though.

I guess what I'm trying to say is that there really are reasons why this seemingly strange/extreme way of stitching can be a good idea. And of course, I'm not planning to do only digital in this system, but mainly film (when shooting seriously). Had it been digital only it would have to be one-shot.
Logged
torger
Sr. Member
****
Offline Offline

Posts: 1400


« Reply #7 on: February 19, 2012, 05:42:52 PM »
ReplyReply

What are the advantages you expect to gain with 4x5 + back stitch vs spherical stitch with a D800(E)?

It is an excellent question.  The short answer would be that I expect to gain a more pleasing experience of photography in the field.

I'm used to work with stitching on 35mm so I know very well how to produce these images. Concerning result only pano head stitching is truly competitive. It is the main method I would recommend someone that likes to get into high res photography. But it also has some disadvantages.

It is possible to develop further, but one problem is that pano heads are generally not very suited for mosaic stitching, they are generally designed to make ultrawide panoramas or spherical stitching. I don't want that. I want normal FOVs but high resolution. I have a pano head with tight clickstops on the horizontal axis, but it is worse on the vertical, and you need two rows, sometimes three (portrait format). It works, but you can't really make the true composition at the scene, instead you gather an overkill amount of data and crop down to your final composition in post-processing.

Working with a view camera is a very interesting process from a compositional standpoint. For example one studies the scene to see if there is any merging of foreground and background that doesn't look good and make very minor adjustments in position/shift to make a better composition. That kind of overview of the composition and possibility to adjust it you don't get with a pano head workflow. It is not absolutely necessary either, it depends on what shooting style and preferences one have. I value those things high. The same way I always use a geared head, I just don't like the imprecise feeling of a ball head.

I use tilt-shift a lot today through the Canon TS-E 24mm, and I'd love to have it for all my landscape focal lengths, which I get with a view camera.

And then it is the side effect of getting into film, something I try to make into a bonus rather than a punishment (looking at drum scan costs it feels like a punishment though), had there been an excellent digital only option though I would have aimed for that.

Another advantage which is about feelings more than rational, is that digitizing what I have composed in the view camera at the scene feels more "real photography" than taking a rough sweep with the pano head and crop at home.
Logged
BernardLanguillier
Sr. Member
****
Offline Offline

Posts: 7874



WWW
« Reply #8 on: February 19, 2012, 10:01:16 PM »
ReplyReply

It is possible to develop further, but one problem is that pano heads are generally not very suited for mosaic stitching, they are generally designed to make ultrawide panoramas or spherical stitching. I don't want that. I want normal FOVs but high resolution. I have a pano head with tight clickstops on the horizontal axis, but it is worse on the vertical, and you need two rows, sometimes three (portrait format). It works, but you can't really make the true composition at the scene, instead you gather an overkill amount of data and crop down to your final composition in post-processing.

I have used my share or 4x5 and like the experience too, but I find flat stitching with back movement to be a lot slower than spherical stitching with a pano head.

As far as deciding on composition and crop is concerned, I personally feel that I don't need a viewfinder to do it. It is fairly easy to:

1. Imagine a crop of a scene just by looking at it,
2. Decide on the resolution you need and therefore roughly the focal lenght you need to stitch with (I only use 100 and 50 mm in fact),
3. Identify the opposite corners you need,
4. Locate the vertical and horizontal limits on your head
5. Decide what should be the peak of focus in the scene and focus there with live view
6. Stitch

Cheers,
Bernard
« Last Edit: February 19, 2012, 10:04:18 PM by BernardLanguillier » Logged

A few images online here!
torger
Sr. Member
****
Offline Offline

Posts: 1400


« Reply #9 on: February 20, 2012, 02:51:57 AM »
ReplyReply

I have used my share or 4x5 and like the experience too, but I find flat stitching with back movement to be a lot slower than spherical stitching with a pano head.

As far as deciding on composition and crop is concerned, I personally feel that I don't need a viewfinder to do it. It is fairly easy to:

1. Imagine a crop of a scene just by looking at it,
2. Decide on the resolution you need and therefore roughly the focal lenght you need to stitch with (I only use 100 and 50 mm in fact),
3. Identify the opposite corners you need,
4. Locate the vertical and horizontal limits on your head
5. Decide what should be the peak of focus in the scene and focus there with live view
6. Stitch

I may actually end up with that, if all just becomes too messy. It seems like I'm kind of moving the wrong direction, while everyone is leaving film and large format, I'm considering moving to it... hmm.... Smiley. Visualization of the crop is the hardest part I think of this, but with training one will get better of course. In multirow stitches it can be useful to have multiple focus points, similar effect to tilting the focal plane, I use that from time to time. The stitcher used needs to handle focus breathing though (which Hugin does).
Logged
BernardLanguillier
Sr. Member
****
Offline Offline

Posts: 7874



WWW
« Reply #10 on: February 20, 2012, 04:04:53 AM »
ReplyReply

I may actually end up with that, if all just becomes too messy. It seems like I'm kind of moving the wrong direction, while everyone is leaving film and large format, I'm considering moving to it... hmm.... Smiley

Well, I thought I was going backwards when I started 4x5 6 years ago... Smiley

The #1 pain is scanning and within scanning keeping slides clean of dust... or cleaning up scans.

I still own an excellent Imacon III which I finally managed to have work on the Mac Pro, but this is just so much of a pain compared to the workflow you get with stitching.

The image below took exactly 70 seconds to capture handheld, 150 seconds end to end to stitch with Auto Pano giga... and then 10 minutes in PS for processing... and it is in my view better than 4x5 quality (I only kept a low res 6,000 pixel image here):

http://www.flickr.com/photos/bernardlanguillier/6902261393/sizes/o/in/photostream/

Cheers,
Bernard
Logged

A few images online here!
torger
Sr. Member
****
Offline Offline

Posts: 1400


« Reply #11 on: February 20, 2012, 05:25:58 AM »
ReplyReply

With Capture One, it is day to get rid of lens cast--simply shoot LCC frames for the back position and aperture and apply them to the taken frames. I do get lens cast with both of my lenses when using a single frame.

I realize that this part may be a bit tricky, when shift has been used and one does not really remember what the shift was set at. Writing down what settings that was used is going to be important Smiley.

My concern here is though that color cast comes from reduced sensitivity of channels, so some dynamic range is lost. If we're talking about fractions of a stop, I'm not worried, but say if you'd lose 2 stops of the green channel towards the corners or something it would be no good.
Logged
torger
Sr. Member
****
Offline Offline

Posts: 1400


« Reply #12 on: February 20, 2012, 05:30:26 AM »
ReplyReply

The #1 pain is scanning and within scanning keeping slides clean of dust... or cleaning up scans.

I'm digitizing some old transparencies so I know about dust. It is indeed a pain. But I'm prepared Wink
Logged
BernardLanguillier
Sr. Member
****
Offline Offline

Posts: 7874



WWW
« Reply #13 on: February 20, 2012, 06:06:15 AM »
ReplyReply

I'm digitizing some old transparencies so I know about dust. It is indeed a pain. But I'm prepared Wink

Yep... some quick figures:

- surface of 24X36: 864 mm2
- surface of 4x5: 12,375 mm 2
------------
ratio: 14.3

Typical scan times on a good scanner:
- 24x36: 1 min
- 4x5: 5 min

Enough said... there is a lot more dust on a 4x5 nega at the end of a scan unless you wet mount.

Cleaning up one sheet perfectly can take 30 mins... multiply this by 3 if you intend to stitch 4x5...

Cheers,
Bernard
Logged

A few images online here!
torger
Sr. Member
****
Offline Offline

Posts: 1400


« Reply #14 on: February 20, 2012, 06:32:07 AM »
ReplyReply

Yep... some quick figures:

- surface of 24X36: 864 mm2
- surface of 4x5: 12,375 mm 2
------------
ratio: 14.3

Typical scan times on a good scanner:
- 24x36: 1 min
- 4x5: 5 min

Enough said... there is a lot more dust on a 4x5 nega at the end of a scan unless you wet mount.

Cleaning up one sheet perfectly can take 30 mins... multiply this by 3 if you intend to stitch 4x5...

The digital stitching part here is not of a film sheet, it is mounted directly in the camera. When I need film digitized I plan to use a professional drum scan service which work in a fairly dust-free environment, there's still dust of course, but not too bad. The drawback is that it costs 80 per frame, but since my production of prints is something like 4 - 8 per year it is manageable. For post-processing a file for a print I spend something like 12 - 16 hours, including test prints and re-adjustments, thinking it over, wait a few days, getting regrets, redoing it, so adding 1-2 hours of dust retouching won't be too much. Yes I'm an incredibly slow photographer with low output, commercially I would fail miserably Smiley. But the mess of film and high cost per frame probably causes less suffering than it would to most photographers...
Logged
elliot_n
Full Member
***
Offline Offline

Posts: 165


« Reply #15 on: February 20, 2012, 08:02:33 AM »
ReplyReply

Visualization of the crop is the hardest part I think of this, but with training one will get better of course...

I understand your concerns. A lot of stitched images look aimless in their composition.

When I stitch I always shot a 3x3 grid that corresponds exactly to an original single master frame. This allows me to compose the shot with great precision.

My technique relies on zoom lenses, a Nodal Ninja head, and an L-plate on camera. Stitches are shot in portrait or landscape mode, depending on the orientation of the single master frame.

I shoot the master frame on a D700 at either 20, 24, 28 or 35mm focal lengths. I then refer to my notes to get the focal length, and upper and lower rotations for my 3x3 stitch. (The stitch is performed mechanically - I don't look through the viewfinder, I just follow my notes, starting at top left, shooting row by row.)

I have my NN5 on top of a Manfrotto geared head - this gives me fine control of composition. (Sadly, it's not the most rigid set-up, so I have to be careful of vibration - mirror-up, waiting a couple of seconds between frames, shielding from wind etc)

All this to emulate a large format film approach.

I get 80 MP files from a D700. (It will be 240 MP from the D800.)
« Last Edit: February 20, 2012, 08:04:42 AM by elliot_n » Logged
benmar
Newbie
*
Offline Offline

Posts: 19


« Reply #16 on: February 20, 2012, 09:42:10 AM »
ReplyReply

I've done some experimentation similar to what you're describing. My reason is that I have an antique lens with certain focal length/vignette/soft focus characteristics that worked well on 4x5 film, but are lost using only the center "sweet spot" when doing single captures with a medium format digital back.

Using a Cambo 4x5 I framed the shots on the 4x5 ground glass, then took off the 4x5 back, replacing it with a Kapture Group sliding adapter back with with an H-25 attached. The sliding back is only for moving the H-25 to the side and moving a small ground glass into place in order to re-focus (different lens to ground glass distance between the two backs). Once focused for  the Digital back, I could capture to double-check focus, then make captures, bracket for extra highlight/shadow detail, and do a capture with the Phase One lens cast correction plexi in front of the lens.

The stitching captures were done using the view camera rear shift and rise and fall to get nine capture zones which covered the full 4x5 area, bracketing and doing LCC captures in each position. I had to always keep the swing and tilt of the rear camera standard at zero so the direction of movement in the shift etc would move in the same plane as the face of the H-25.

Granted the lens I have worked with is very forgiving for this process (it is soft enough that "focus" is only vague at best) but the focal length is somewhere around 90 mm and I was easily able to handle the LCC color shift in C1 by applying an LCC correction to the captures from each zone.

In general, I think for more exacting work the more sturdy and precise the camera body and its movements the better. For example, my Cambo 4x5 has a geared rear shift but only a manual "rapid" rise and fall, which can introduce a little bit of rotation when using that movement.
Logged
yaya
Sr. Member
****
Offline Offline

Posts: 1124



WWW
« Reply #17 on: February 20, 2012, 10:02:28 AM »
ReplyReply

There's a lot that can be done if you're willing to put time and effort into it....

Record-breaking image of Gas Street Basin and http://www.britishwaterways.co.uk/scotland/newsroom/display/id/2073
Logged

Yair Shahar | Product Manager | Mamiya Leaf |
e: ysh@leaf-photography.com | m: +44(0)77 8992 8199 | www.mamiyaleaf.com | yaya's blog
torger
Sr. Member
****
Offline Offline

Posts: 1400


« Reply #18 on: February 20, 2012, 10:30:52 AM »
ReplyReply

I've done some experimentation similar to what you're describing.

Great feedback. It does seem like this is possible to do. Now I just wonder if Arca Swiss F-Metric support and frame is rigid and precise enough to keep the MFDB in the focal plane...
Logged
BernardLanguillier
Sr. Member
****
Offline Offline

Posts: 7874



WWW
« Reply #19 on: February 21, 2012, 04:36:19 PM »
ReplyReply

Granted the lens I have worked with is very forgiving for this process (it is soft enough that "focus" is only vague at best) but the focal length is somewhere around 90 mm and I was easily able to handle the LCC color shift in C1 by applying an LCC correction to the captures from each zone.

I understand the fun of doing this, but is there any value in terms of results compared to spherical stitching that will generate a tack sharp image from corner to corner, will be much faster to shoot, will not have any color cast issues,...?

Cheers,
Bernard
Logged

A few images online here!
Pages: [1] 2 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad