Move along? What are you referring to? What software? The PS5 I was referring to can't even do it with a simple flat stitch of two images if there is not sufficient detail in the overlap areas. I deal with this on commercial jobs every week. This will be a issue. On more sophisticated software you have to manually do control points-a PITA. So what do you know that you haven't said?
Thanks,
Kirk
at age 73:
"The woods are lovely, dark and deep,
But I have promises to keep,
And miles to go before I sleep,
And miles to go before I sleep"
Knowledge of where the shot is situated in the overall image is probably not enough to eliminate the need for the stitching software to correlate the images. Even if the X/Y stage is computer controlled, it won't be accurate to the pixel level.
However, dust will provide a natural target for correlation. Even very clean negatives have plenty of dust. I don't foresee any big problem with stitching. Some custom software may be required, but that's not the end of the world.
Ben based on my experience stitching, I don't think dust will be enough-maybe if there is a ton of it and you jack up the contrast-but I doubt it.
Thanks,
Kirk
at age 73:
"The woods are lovely, dark and deep,
But I have promises to keep,
And miles to go before I sleep,
And miles to go before I sleep"
Sheesh, Brian, that's easy to say but not so easy to live up to. What software are you talking about? Software that hasn't been written yet? I don't intend to write software for this activity. And you are assuming that the software knows, within 4 or 5 microns, the position of the film stage, or that the film stage can be moved with that degree of accuracy. Anything I build will not provide that, though it will keep the film in the correct plane. Just because it's easy to articulate the solution to a problem means the resources are at hand to achieve it. Do you know of a product that already does that?Originally Posted by Brian C. Miller
Others are happy to construct a vast complication, and that's fine, but I'd just as soon build mine from existing bits, preferably those I already own. I'm short of time and money, and requirements for much of either will make this project as inaccessible as a Scitex.
I have an old copy of Panavue but not on my newer machine. It allowed precision alignment using manually set alignment points, either based on landmarks in the image or on marks of our own making. With a distortionless flat-field lens (which is not that hard to achieve for bellows lenses that tend to be symmetrical), only two markings are needed and they could be placed on the support glass and removed later in Photoshop.
The software Kirk has suggested deserves careful study, which I will do at some point. But I wonder if a template would still require precision lateral positioning of each tile. I was thinking that approximate positioning, based on markings and manual positioning, would be the much easier machine to build.
Rick "keeping it simple" Denney
I'm hearing this loud and clear. The histogram for my test scan, which included quite a range of tones from deep shadows (base fog on the negative) to fairly dense highlights, only filled the center third-to-half or so of the full range recorded by the sensor. To restore contrast, I stretched the tone curve, but didn't see any hint of posterization either visually or by looking at the histogram.
I got better contrast at the pixel level than does the Epson, not that the Epson presents a difficult goal, but I can see how flare would blind the sensor to low-contrast fine detail. I made no attempt at any masking whatsover, and my light source was diffuse and much wider than the negative, let alone the field of view. So my test results above are perhaps as bad as it gets in this regard. Also, I converted from RAW to a high-gamut color space (ProPhoto), which in hindsight was a mistake--lots of empty gamut left at the extremes of the histogram.
I'm not that sure such detail survives conventional enlargement any better, but that would be sacrilegious to mention. Nor do I think any other scanner except for a laser/PMT drum scanner is likely to do that much better. I have a condenser head that I can probably find with some digging and use, if for no other reason than to provide a very even light source. It will be more collimated than a light box, but not nearly as much as a laser source. I suppose I'll find out how much it does or does not help.
Rick "who has already bettered the Epson with no special care" Denney
It appears to me that good portion of this gismo is worked out. I sometimes use PS8 for stitching and it works v. well. I believe the ptgui has better stitching software. The idea of the camera pointing down w/macro lens make sense to me. The camera could be sliding on solid rails (similar as the interior crane operates).....and set to better than 0.0001" accuracy. Indeed, this can be motorized by servo motors and controlled via CMC type software....or similar.
While the camera moves over the determined grid, it fires frames allowing 10-20% space to overlap...or even more if sweet spot of the lens is used. However, the bigger the file the more problematic will be the processing of it. I had a problem with 514MP pano file out of 7 shots and my computer has near 4GB of memory....so this is kind of heads up.
Once the sliding rails are firmly attached and calibrated, the camera/lens should be the same distance over the surface of the 4x5 or larger negative. Which also means that there will be a need to use live-view focus (13X on my D700).....and once that's accomplished (tether or manually)...the camera can then copy the 4x5 negative/positive. Naturally the WB should be set to manual....and on some cameras Kelvin degrees can be adjusted in increments. The camera has a direct hook up to USB so there would be no need to juggle the memory card....as well as the photos can be reviewed on large H. Def monitor. Using wi-fi camera controls from a laptop would allow all of this to work remotely. Also, the lens could have gearing and also can be rotated via wi-fi....in order to obtain precise focus.
True, the expense of the D700 is up there, but the files would definiately look clean. The only unresolved part would be the automation of running the camera on rails over the negative and triggering it at appropriate spot in order to cover the overlapping areas. However, that could also be done manually if necessary....and naturally it would take more time.
If i didn't convolute the issue, hopefully this was helpful.
Les
"a highly collimated light source would help to greatly increase contrast"
I know very little so please don't jump on me but might it be possible to cheaply set up a linear array of collimated light sources even if they were staggered a little? Fiber optics? Tiny tubes aligned that direct light? Even if the light source was not configured in a straight line a precision timed movement of the light source, film, or sensor in one direction should provide even illumination?
If only a small array of light can be set up, mow an area the size of the sensor with collimnated light and a small, highly accurate x-y controller, then move the film with a larger, less accurate x-y controller with some overlap, scan again with the highly accurate x-y controller and then stitch the smaller areas together? Could this possibly reduce the cost of trying to make a large, highly accurate x-y table?
If a collimated light source could be set up in a linear fashion the same length of the short axis of a sensor, the movement of the light, film, or sensor would only need to be linear and this should be much cheaper to build with precision.
So the idea is just make a small scanning area, highly accurate, and hopefully cost effective to build, and then stitch the small scanned areas together?
I'm talking out my butt and just throwing out ideas in hopes that better ideas might come of this so don't read too much into this.
http://www.robotshop.com/ for Arduino controllers etc. There is an online community that can help with programming the Arduinos but the x-y thing is supposed to be pretty common apparently.
This is a great project!
I like the ideas! Good work!
Regarding stitching, as I've said before I agree with Kirk that there might be some issues here, but we really don't know whether there's a problem or not until we stitch some samples at the magnifications we're talking about. Let's not get too involved in fixing a problem that might not even be there.
Regarding light sources, I'm a microscope neophyte, but don't some of them have collimated light sources? This might be a good option for experimentation.
Regarding moving the camera instead of the negative, this would certainly be feasible, and it would especially be good for people with 8x10" and larger negatives, since if you move the negative, the glass support plate needs to be twice as big as the negative with a moving x-y negative stage. My guess is this would be a little harder to implement with common materials, but I'd love to be wrong, and it'd be nice if someone investigated this. I really wish that I had a closet full of 80/20 components.
The best next steps appears to be:
1) Making an even light source. My first one is going to be diffusion.
2) Making a sliding negative stage. That way we can investigate the stitching issue.
I'm certainly up to doing that, although I'm going on location for a shoot this weekend, and so it'll probably be next week before this gets done.
Regarding the robotics, I'm sure I can do this too, but it's been a long time since the electro-magnetism semester of college physics. If someone with more electrical experience wanted to investigate positioners, stepper motors, stepper drivers (probably an Arduino), Arduino programming... That would be a big help. See, for instance: http://www.photomacrography.net/foru...ic.php?t=15673
“You often feel tired, not because you've done too much, but because you've done too little of what sparks a light in you.”
― Alexander Den Heijer, Nothing You Don't Already Know
Dave, as a fellow canuck I am hinting that you can dig this pit as deep as you like and there ain't no bone in it. The problems are not obvious until you have some montaging experience and that is why I suggest trying one of the commercial scanning systems before investing much time. On the other hand, you can learn lots and have fun if that is the goal.If we were wanting to build something anyway just to try and do better than an Epson V700 / 750 what would you recommend for a set up using a DSLR?
I was raised in Grimsby, Ontario so can you throw an old fellow Canuck a bone?
Note, I have no direct experience using DSLRs to scan film. Everything I did was with scientific cameras. Therefore, feel free to consider me as much a newbie as anyone. Perhaps there is a special sauce in the DSLR.
Funny, we made our mark using camera-based systems to replace scanning PMTs in demanding applications. Fuji Film's medical division used to sell one of our scanners in Japan as the "Beautiful Image System". And here I am being a wet blanket. Sorry.
Montage of about 25 tiles. She's kinda dirty and missing a leg but not bad for 150 years old.
Beings that I make my living from writing software, and I have experience in machine automation, I write from the authority of personal experience when I say, this isn't a problem.
The GIMP is driven by Script-Fu. The Script-Fu language is what is engaged when you click the buttons on the GUI. Here we have a situation where images, without edge landmarks, need to be tiled and overlapped together. This means that certain edges don't matter, so they can be ignored. Ignore the edges, and position according to the tile information. It's just that simple. There isn't that much adjustment needed to feather in 64 white or blue or whatever tiles.
The reason that the conventional tiling software has a problem is because it is written such that the edges must have landmarks. But here, in our situation, for this purpose, the camera's position is known. 2 follows 1, again and again, and with a reasonable degree of precision. Falling back to camera position is just that, falling back. If the primary stitch target doesn't exist, then camera position is all that matters. The tiles get feathered together in sequence.
This would be a minor modification to the existing stitching scripts.
"It's the way to educate your eyes. Stare. Pry, listen, eavesdrop. Die knowing something. You are not here long." - Walker Evans
Posting Permissions
Reply With Quote
Bookmarks