Ad
Ad
Ad
Pages: [1]   Bottom of Page
Print
Author Topic: Stitching D2X (HDR) Files  (Read 7080 times)
Bakester
Newbie
*
Offline Offline

Posts: 1


« on: April 22, 2006, 02:04:51 PM »
ReplyReply

Hi All, My first post on this forum and hoping it's not one of those "not again" topics.

 A little background. After years of specializing in jewelry and shooting 4X5, I find myself wanting to transition into more editorial type work.  

I'd still like to service my current jewelry clients but with an upcoming major relocation, the film based workflow will be an issue.  

I would prefer not spending the $15K - $30K on a MF back if I can help it (I'm still not entirely convinced the MF would work unless it's the P45 or similar res).

I have a D2X and am wondering about the possibility of mounting it on the back of my Sinar P and doing multi shots (2 to 4) then stitching it all together.  I picked up a Nikon adapter for the back so thats not an issue (the least of the problems I'm sure). Has anyone tried this? Used HDR with any success?  Thanks.
Logged
TimothyFarrar
Newbie
*
Offline Offline

Posts: 49


« Reply #1 on: April 23, 2006, 11:08:32 PM »
ReplyReply

Stitching can be quite a large amount of work, both at the time of the shoot taking many shots (you might need up to 30% overlap which will be much more than 4 shots) and later in post processing. Adding dynamic range blending will complicate the process, each series of exposures for each photo used in the stitching will need to have the same fstop and series of shutter times.

Automating the process of blending all the exposures into single extended dynamic range 16bit source exposures for use in stitching can be done (I have a photoshop script for this). After stitching (perhaps using an outside program), you will have a huge 16bit source image which will need to be brought back in Photoshop for development (brightness, contrast, etc). Then you can make a downsampled image to try various developments of. Then when satisfied, apply the same development to the full resolution image.

The other option is to try and develop your source shots prior to stitching, which risks stitching artifacts if the development is not consistent across all the photos.

I am not sure if any external stitching software (used to speed up the process) supports transparency in tiff files (so you can erase the shadow on the shots away from the center axis). If not you might have to cut up those shots into smaller rectangles and crop out the shadows before loading them in a stitcher.

You could try the development process out on a series of say 50mm shots from your Nikon alone mounted on a tripod (instead of mounting the Nikon back to your 4x5). There is a very good , but sometimes buggy, free stitcher (based on Panorama Tools) which can also correct for lens distortions,

http://hugin.sourceforge.net/

In terms of resolution (you might have already seen this), Max Lyons did a gigapixel image stitching from a 6MP camera (using a different process than moving a nikon back mounted on a 4x5),

http://www.tawbaware.com/maxlyons/gigapixel.htm

This might give you a better idea of the time involved in the stitching process.

Hope that helps,

- TImothy Farrar
Logged

Timothy Farrar
Farrar Focus Digital Darkroom
www.farrarfocus.com/ffdd
BernardLanguillier
Sr. Member
****
Offline Offline

Posts: 7522



WWW
« Reply #2 on: April 25, 2006, 12:54:22 AM »
ReplyReply

The Horseman LD might be a better option if you intend to do flat stitching with a D2x.

I am not sure to understand why you also need to do high dynamic range stuff on top of that for product shots. My feeling is that a D2x has more DR than slides.

Regards,
Bernard
Logged

A few images online here!
ivyrob
Newbie
*
Offline Offline

Posts: 1


WWW
« Reply #3 on: May 13, 2006, 11:36:46 PM »
ReplyReply

My experience with D2X and stitching:

I´ve been taken panoramas with a D100 for several years, and with a D2X for half and a year.
Usually I first take a photo with a wide-angle lens to check for the metering in what is something like an aproach to the final photo.
I use manfrotto´s pano head, to keep the camera rotating on the lens nodal point, fix shutter and aperture as previously metered, and fix white balance.
The lens is always above 50mm. Photogramas have to be overlaped between 20 to 40 % for better results. (Depending on the lens, the subject, etc)
I stitch photograms with Panavue Image Assembler, setting manual joining points, and keeping layers for final retouching on Photoshop.
This method is very "time - consuming", sometimes getting good results, and sometimes not.

I also have used the D2X on a Cambo Ultima 4x5 Camera. My idea was to get the images I needed with more resolution, without investing on a MF digital back.
So I got the 35mm accessory for the Cambo, as well as the 23 front standard, to have the Nikon working right on axis.
You can take several shots (usually 4 to 5) shifting the cambo´s back with the nikon mounted with the vertical adaptor.
This allows for an image between 12x3 to 12x4 megapixels, which can be quickly and perfectly stitched on Photoshop, usually without any king of warping. (The Cambo keeps the focal plane for the shots as you shift the back of the camera laterally). (You need to set manual aperture / shutter/ white balance...)

The main problem I have encountered using this method is the poor quality of the lenses. (And so the photographs)
There´s a highly noticeable difference between shots taken with good nikon optics and the ones used on the Cambo. (I use several Schneider Digitar lenses, and the 28mm is specially dissapointing.)
Also, depending on how much you swing or tilt the camera, you are not able to shift the back the full 25mm left / right
Because of the way the sensor is kept inside the body, and its size, I personally cannot recommend to use the Nikon for shooting this way. (Another drawback is that there are just a few lenses to choose which can focus on the sensor area.)

Same problems should apply on shooting on a Sinar P. (Also the camera is not kept on axis and focus has to be checked after any swing - tilt movement)

The Vcc option at horseman keeps the MoreResolution/LessLensQuality vs LessResolution/HighLensQuality subject, and also the disadvantage(?) of being limited to lenses over 80mm.

As it seems you have all the "ingredients" to check the results on your Sinar, you can try and see the results, and share them on the web.
Hope my post helps,

Regards,

Robert
Logged
Dave Carter
Full Member
***
Offline Offline

Posts: 108


WWW
« Reply #4 on: May 14, 2006, 08:45:36 AM »
ReplyReply

Bakester, I am also doing stiching with a D2x.  I started with a D1x a number of years ago, than moved to the D2x and a home built pano head.  Most of my panos are 9+ pictures.  It is a lot of labor to get the final picture just right.

I have considered putting my camera on the back of a 4x5 or 8x10 and moving it around the field to take the pictures.   Certainly this would help in stiching.  However, I am concerned about lens quality.

If all our work is static objects, you should look at a scanning back for you 4x5.  I think you can either rent one or talk to Betterlight about a trial.  Plus they do not cost as much as the new MF Backs.
Logged
BernardLanguillier
Sr. Member
****
Offline Offline

Posts: 7522



WWW
« Reply #5 on: May 14, 2006, 09:39:42 AM »
ReplyReply

I have been doing panoramas with a d2x myself for quite some time now.

My mainstream setup is a RRS spherical head, and PTgui, it works like a charm and the workflow is very efficient.

Some recent samples:

http://static.flickr.com/56/145374901_07e65db3b7_o.jpg
http://static.flickr.com/49/144844176_034b80dfd9_o.jpg

I have startet to do flat stitching with a Horseman LD recently. Together with a Mamiya 645 45 mm f2.8 lens, you can get reasonnably wide angles pano made up of 6 horizontal D2x images. The result is typically 10.000 * 4500 pixels and is basically perfect.

The Nikkor lenses are probably a bit sharper, but the Mamiya is sharp enough at f11 - f16.

The Schneider 110 XL 5.6 is another lens that I have used also, and I find it to be sharp enough also.

The problem is the extra weight compares to a basic spherical pano kit. The Horseman is a solution for those locations you can access from a car, but not a credible trekking option IMHO.

Regards,
Bernard
« Last Edit: May 14, 2006, 09:47:19 AM by BernardLanguillier » Logged

A few images online here!
Blendenteufel
Newbie
*
Offline Offline

Posts: 19


WWW
« Reply #6 on: May 15, 2006, 04:25:49 AM »
ReplyReply

Quote
Hi All, My first post on this forum and hoping it's not one of those "not again" topics.

 A little background. After years of specializing in jewelry and shooting 4X5, I find myself wanting to transition into more editorial type work. 

I'd still like to service my current jewelry clients but with an upcoming major relocation, the film based workflow will be an issue. 

I would prefer not spending the $15K - $30K on a MF back if I can help it (I'm still not entirely convinced the MF would work unless it's the P45 or similar res).

I have a D2X and am wondering about the possibility of mounting it on the back of my Sinar P and doing multi shots (2 to 4) then stitching it all together.  I picked up a Nikon adapter for the back so thats not an issue (the least of the problems I'm sure). Has anyone tried this? Used HDR with any success?  Thanks.
[a href=\"index.php?act=findpost&pid=63389\"][{POST_SNAPBACK}][/a]

Hi. I have been experimenting a bit with the combination of stitching and HDR. There are two possible paths to go in the "development phase"

1) stitch first all of the images into one big image with a pano stitching tool (e.g. PTGUI allows for stitching of 100% overlapping images) and then save several different stitched images (visible images only for one EV) for each EV value and then use DRI/HDR on those;

2) create 32bit HDR images, transform them into 16bit and then stitch the resulting 16bit images.

Method 1) is a lot of stitching work (especally setting control points so that all images of one EV are perfectly aligned with images of EV-1, EV -2, EV +1, EV+2). So for a 4x4 picture with 5 EV there is already a lot of stitching to do.

Method 2) seems very workable, BUT I seem not to be able to apply exactly the same curve at the mode transformation from 32-bit into 16-bit for the images that will be stitched. Hence, the have different luminosity levels, which can be annoying when blending.

So, it is possible and it is cheap ... (assuming CS2 and a pano stitcher).
Logged

BernardLanguillier
Sr. Member
****
Offline Offline

Posts: 7522



WWW
« Reply #7 on: May 15, 2006, 06:01:29 AM »
ReplyReply

Quote
Hi. I have been experimenting a bit with the combination of stitching and HDR. There are two possible paths to go in the "development phase"

1) stitch first all of the images into one big image with a pano stitching tool (e.g. PTGUI allows for stitching of 100% overlapping images) and then save several different stitched images (visible images only for one EV) for each EV value and then use DRI/HDR on those;

Method 1) is a lot of stitching work (especally setting control points so that all images of one EV are perfectly aligned with images of EV-1, EV -2, EV +1, EV+2). So for a 4x4 picture with 5 EV there is already a lot of stitching to do.
[a href=\"index.php?act=findpost&pid=65487\"][{POST_SNAPBACK}][/a]

Regarding method 1, I haven't tried yet, but I was under the impression that it was probably possible with PTGui to resuse a set of control points with a different set of images.

Am I wrong to think so?

If I were right, method 1 would become a lot more attractive, wouldn't it?

Cheers,
Bernard
Logged

A few images online here!
nineinone
Newbie
*
Offline Offline

Posts: 11


« Reply #8 on: May 17, 2006, 11:09:56 AM »
ReplyReply

re: software:  I've also used PtGui, but am now using Real Viz's Stitcher 5.1, and I find it to be the better program. yes, its a tad expensive, but I like the 3d working space and many of the automated functions. I've even been able to stitch hand held shots with it that needed some serious warping, and they came out seamless.

re: the lenses, I can imagine this being a problem, as the smaller nikon lenses are sharper, but wouldnt the fact that you're stitching say a 5O MP file together make up for this? ie 4 stitched shots of an admitedly softer digitar should still come out pretty good, no? I guess the best way to test for this would to be to do it both ways, ie firstly the flat stich method where only the sensor moves (with the bigger lens) and secondly, where you use the smaller nikon lens and move it along with the sensor. this is obviously the less perfect way to stich, but with what Ive seen from Real Viz, even doing everything completely wrong without even a panoramic head seems to come out perfectly usable, so who knows, maybe doing it the "wrong" way actually maybe the best way, ie where the actual shots are sharper, but need more processing from the software to be put together.
Logged
Pages: [1]   Top of Page
Print
Jump to:  

Ad
Ad
Ad