I have some medium format trans’ that are dupes of the originals. The grain was only multiplied when copied to the new film. Now I have the task of taking the 2×2’s and making 48×48 out of them and the grain is so bad it looks like chunks of wood floating around in the emulsion 😉
The grain is course, the shots are mostly outdoors, some with heavy texture and others not. Other than hand mousing over the entire image, what would you guys suggest to remove course grain without making the image too soft? I was thinking is there a way to make a couple of different masks and apply gaussian only to those masks without losing sharp??? Luminance mask for the highlights grain?? I’m not at the studio to try it and won’t be for a few days.
Try: http://www.neatimage.com. This sucker does an incredible job and doesn’t reduce any of the detail in the image. It is a little slow, but does a fantastic job.
Thanks for all the replies gentlemen. I did dload the demo of Neat Image. Didn’t have much time to use it, but couldn’t load a TIFF since it doesn’t work with compressed images. Didn’t have the time to decompress & resave. I do have grain surgery 2, and have tried to correct with it but still have the unwanted side effect of softening. On certain images, I can simply mask out the sky and soften out the grain but this does not help for the higher detail areas. I think most of these "degrainers" are mostly using blurs of one sort or another to accomplish their goal, maybe with a unsharp mask on the end, but they all soften the detail areas.
No one has an opinion on the use of masks to eliminate grain away from the detail?
When I get to the studio tonite I’ll sample an area of sky & detail together to show the grain, hope the compression will still show the grain correctly.
Thanks again for tagging the thread.
and how does one attach a jpeg to their message here?
But they wont do well with film grain, as they are designed for digital noise. However I may have other tools that would help you, but I need to ‘see for myself’ first.
You cant attach JPEGs here; You must upload to a webserver (PBASE for instance), and then put a link here.
Mathias, I just checked out the link for the actions, the tool looks good for digital noise. The grain I’m talking about is very old film grain. These emulsions are probably 30-40 yrs old, and the trans quality back then was terrible. When blown up, the grain appears as dents and sticks floating in the emulsion. The grain is not linear and preset filters have a tough time with non-linear distortions in the image. Problem there is, if I could get rid of the problem grain, I’d be eliminating important detail data in the photo as well.
I’m gonna do a temp account on pbase and set up the images tonite, or rather about 5:00 in the morning tomorrow.
Yes, film grain is tricky, very tricky, to remove. Especially with pre-sets, as there´s no way of knowing how big the scan (thus the grain) really is.. 90% of the time, living with the grain is probably your best option..
I think most of these "degrainers" are mostly using blurs of one sort or another to accomplish their goal, maybe with a unsharp mask on the end, but they all soften the detail areas.
I don’t think that is how they work. Not the good ones, anyway. Such a simplistic approach wouldn’t be worthy of spending money for. Grain Surgery is an example of a program that does not work that way. It works by sampling the noise so it will know what pattern to suppress. I suggest you persist in learning how to use Grain Surgery 2.
I did dload the demo of Neat Image. Didn’t have much time to use it, but couldn’t load a TIFF since it doesn’t work with compressed images. Didn’t have the time to decompress & resave.
Thanks for the reply Burton. My problem with these images, and I will up a few tonite, is that the grain is too large and variant that it doesn’t get recognized as noise. Actually noise, or finer semi-random patterns would be a blessing compared to this. Picture a nice white house, then up to 100% and imagine seeng little dark sticks all over the siding. I’m sure some of the users here have seen old med format emulsions before, but this is exactly the kind of thing the degrainers don’t get rid of because it’s too large. Fine noise and artifacts actually have somewhat of a pattern that can be read compared to this stuff. I’ll post on the site suggested above and let you kind folks judge. In the meantime, I’m masking and blurring to get this done. These are freebees for someone else who isn’t worried about inconsistency from panel to panel. 42 4×8 panels to be done, 22 left to go!
My problem with these images … is that the grain is too large and variant that it doesn’t get recognized as noise.
Neat Image lets you select noise samples in your image. You could deliberately sample areas containing those "little dark sticks". Neat Image gives you a lot of ways to select and define the sort of noise you want eliminated. For some information, see:
Early this morning I downloaded and installed the new version 3.0 of Neat Image Pro+ and I am delighted with the new improvements. I am still learning about them. If you want, you could post some of your problem images and give some of us a shot at them. After struggling with them we may have a greater respect for the challenges posed by the grain in your medium format trannies. (grin)
In the meantime, I’m masking and blurring to get this done.
As an alternative to blurring, have you tried Photoshop’s Filter> Noise> Median? It removes small details selectively. I use Median quite a bit. Don’t overdo it — experiment with its pixel setting. Sometimes I invert the image before applying Median and then invert again after applying it. That makes a difference.
I’ve used Neat Image several times on slides that I’ve scanned in to restore and it’s worked almost every time. another trick is to duplicate the green channel and blur it a little, that way the filter doesn’t cause as much blurring over all. This may or may not be something you’ve already tried, but you can duplicate your image and paste it over itself as a new layer (I’m assuming you’re using high resolutions here) then lower the transparency of the new layer a bit and nudge it over by one or two pixels. I’ve done this with one and two extra layers and it can help alot. Play with the opacity and see how much effect you can get out of it. You can follow that up with either the Neat image option, or the single channel blurr, or both. I hope it helps.
trying to get these files ul’d and it’s a real pain. I’m not that familiar with the aging effects of film… is reticulation what happens with age to the emulsion? dry lake bed would be more appropriate than "sticks" or grain.
Thanks for that input r_harvey. Was playing around earlier with neatimage again. I got away with it on one file, close up though it has that "painted" look as some details were lost, but it’ll pass for this project.
Usually from processing errors, occasionally from extreme storage conditions, reticulation leaves patterns, sometimes random, hit-or-miss, where there are tiny cracks, clumps, or bubbles.
I just did a search for reticulation on Google. Some of the results yielded sites with samples of what it could look like and I didn’t see anything quite like what I have although it certainly sounds like the problem. The problem also could be characterized as someone taking a picturem the crumpling it up in a wad and then unfolding it and laying it flat. You still see all the little folds & creases in the image even though it’s laid out flat. Does older film grain usually look like noise or specs or can it look like the "crumpled paper"?
Just to remind anyone, I am taking 2×2" and blowing it up to 48×48".
Does older film grain usually look like noise or specs or can it look like the "crumpled paper"?
Nope. If properly processed and carefully stored, grain a hundred years old will not look any coarser than the day it was processed. Many newer films have finer grain structure, but grain is grain. "crumpled paper" sounds like handling problems… too hot, too cold, extreme changes, cold and wet then hot and dry–reticulation.
Regardless of the cause you have abberrations in your image that you wish to be rid of correct? Have you accomplished it yet? If so, what worked for you? If not, have you tried the options that have been offerred up so far? If not, go back a few pages and try some of the ideas we’ve given you.I’m interested to see how it comes out for you.
BTW, have you got the image anywhere online where we could see it?
having probs getting pbase to take my pics…. all small in size that show the detail of the problem. The link doesn’t link to my pics, sometimes get nothing sometimes other folks pics.
I’ve used Neatimage Pro 2.61 (version I think is correct) to reduce some noise on one, softened the details but still useable. No luck with the most common problem in most of the shots.
All these trannies are from the Italian National Bureau of Tourism, and are shots of everywhere in Italy. These are copies of the origs, and are kept by them in books laying on shelves. 30-40 yrs old and people have been thumbing through them since they were made. Lot of good shots, but with the "problem" in most, I can only fix the ski without losing detail in the buildings or higher detail areas.
Still trying to get them onto Pbase…. when the link works I’ll post.
Thanks for tagging along and trying to help everyone. Much appreciated.
I upped a raw scan that shows what most of these trannies have in it. I’ve tried median, blurrs, neatimage. With these I usually get an unacceptable amount of softening of the needed image to get rid of the "crap".
I admit I don’t know how to use NeatImage, so using the fine tune method is what I’ve been doing. Probabaly not the best use of it, but if anyone knows a good setting for taking care of this I’d be all ears!
From what I see, there is some obvious dust and dirt but no grain. The overall texture of the entire image appears to be an accurate capture of the original object, probably a mass produced tourist print.
Film grain appears as a uniform grittiness. I think what you are seeing is the texture in a man-made panel that carries a printed image. You could have a dup of a mass produced picture.
For a small picture of film grain, see figure 15 at Kodak
Is that a stereo pair? It looks a bit like one. The "reticulation" is not film degredation, it is simply the paper texture on which the original photos existed. This is apparently a picture of a print and the picture faithfully reproduced the decorative pebble texture of the print. I’ll see if Neat Image can handle this. Admittedly, it will be a challenge because this is not noise or film grain.
That’s not grain or reticulation. This is a copy. It would save a lot of work if you could get the original film.
The texture doesn’t repeat, so it was not done in software–if it does repeat, it’s over a larger area, which would mean that it is part of the paper. The texture is sharper than the image, so it was added after the fact. The black junk is sharper than the texture, so it was added in the scan. The lighting looks even, but directional, so it was probably not made on a multi-light copy stand. The texture is more contrasty than the image–a bad thing.
You can get rid of most black spots by using an anti-static brush on each piece of film before scanning.
Since the image is not as sharp as the texture, it will take an awful lot of work to eliminate the texture, and by the time the sharp texture is smoothed-away, the image will lose some. If your example is typical, detail in the image is lower than the detail in the texture, so the loss will probably be acceptable.
Scan at very high resolution, smooth it, sharpen it a bit, then see what you have left.
they’re the same image… I actually uploaded it more than once, since I couldn’t get it to take the file the first time. From what I’ve been told, these are copies of the original transparencies. The original ones are supposed to be first generation and not photos of prints. Although it does look like paper grain, I was told they’re just copies of original landscape chromes.
When I tried NeatImage, I got super soft results when the "prob" was minimized. Almost giving the image a watercolor painting effect. I couldn’t use the image before or after. Median doesn’t work either.
I’m not sure you will get the same results as I did seeing as how there isn’t much detail information in the jpeg. I can’t up the original scans as there’s a 10MB limit to pbase. So I don’t know if anyone can successfully try to work on one of these upped images. But at least you can now see the problem I have. I don’t think I’m gonna make the full 42 by the end of the month 😉
Are you guys fairly sure these are copies of prints? I can always go back to the source and as if I can get originals, but I doubt I can.
Is the grain sharper than the image? I see it has more definition as it’s small. The dust is from MY scan I’m pretty sure. I did take note of what r_harvey said about the lighting direction to the grain. This trans was scanned from top to bottom, if it was reticulation, would the scanning light produce a shadow on it as appears? Or if this was grain, would the scanning light cause shadow?
I don’t know now what I have, except that my partner is trying to do a favor for some other people and it’s really a huge PITA. He has no other sources for these images other than the one we have now, and this is all I have to work with.
One thing I don’t get, is that these are Med format chromes. I don’t know what purpose they serve as they aren’t made for a slide projector. So that leaves what as the purpose for a 2×2 chrome? Reproduction is the only thing I can think of, and even that can’t be done decently.
Ok, it’s a whole other deal if these are scans of prints. If you can rescan it yourself, then (now this may sound odd) flip the image and scan it again. That will give opposing highlights and shadows to the texture in the surface of the print while duplicating the actual image data. Paste the second scan over the first and it will help cancel out the texture. It’s not perfect but it will give you a great reduction in the visibility of the texture without loss of image detail since the only thing different is the texture lighting, not the smooth image data. Adjust the opacity of your second layer up and down and you’ll see the effect. After that you should be able to clean it up in a more standard fashion.
Going down to the source and finding out once ‘n fer all what these are trans’ of. All I do know is I scanned this in as 2×2 positive transparency, top to bottom.
TempusMn, this sounds interesting, I’m trying to picture this in my mind here…flip it which way to cancel out the garbage? All I can see is scanning the other side than the emulsion but that puts everything backwards.
As I said in a previous message, the "reticulation" is actually just a good representation of the pebbled surface of the original "cardboard" prints.
I applied Neat Image 3.0 Pro+ to your image by selecting several samples of the paper texture. Nothing beyond the Neat Image processing was done. As you can see from the image in the link below, Neat Image completely removed the pebble texture of the paper and left the rest of the image untouched.
Unfortunately some parts of the rest of the image that were untouched were flaws from some bits of dirt and fibers and possible nicks in the print surface or trannie surfaces. Apparently the original pebbled print may have had some dirt on it. I suppose it is possible that some additional fibers and dust appeared on the scanned transparencies. Apparently both the original pebble-surfaced prints and the transparencies of the photographs of those prints provided surfaces that were vulnerable to environmental dust, fibers, and other contamination.
Before doing a scan you should blow off the print or trannie with canned air and/or use a special antistatic brush to remove any foreign material from the surface. You probably have no control over cleaning the original pebbled prints because they were done long ago, but you can clean the trannies with canned air and/or an antistatic brush before scanning them.
Ideally the original photos and subsequent scanning of their transparencies would be done in clean room conditions, but that isn’t always possible, so large foreign particles and fibers have to be manually removed in Photoshop. The Rubberstamp (Clone Stamp) tool has long been used to deal with these individual image flaws. The new Healing Brush tool might be a faster way to fix these individual flaws.
If you think any of this dirt was on the surfaces of the transparencies that you scanned, in the future you can save yourself some work in Photoshop by cleaning them thoroughly before scanning them. In my experience, something always gets through and you still need to do a little manual retouching in Photoshop. But hopefully you can keep that to a minimum.
In my own mind, Neat Image 3.0 Pro+ successfully dealt with the pebbled texture of the original photo prints while leaving the rest of the image intact, and I definitely would use it on this project. An alternate approach will be more work and/or reduce image quality. Neat Image can’t magically solve all of the problems. But without Neat Image, those pebbled print surfaces are a major problem for you. It is ironic but impressive that Neat Image removed the pebbled texture while leaving little fibers of lint in sharp detail.
Thanks for taking time to work on that pic Burton. Neatimage did do some work on that image, although for the intended output of 48×48" what was accomplished would still be very visible to the eye from 6-8 feet distancce. I must say though that it is a valuable tool. I know it can’t get rid of everything, but it appears even with a JPEG it did more than just blur as I originally thought it would do. Would running it over again and resampling the new "grain" help at all?
As far as the trans goes, this was just one that I grabbed from the box and scanned in without prep to post on the site. In usual practice, all trannies are first inspected under magnification to check for scratches and damage, then cleaned with film cleaner and oil mounted if necessary. Most of the time the fibers and dust are not part of the equation but for down & dirty grabbing to share this problem with you good folks I just blasted through and went home.
Burton, do you feel neat image 3.0 pro is much better than 2.61 or do they accomplish the same results?
Thanks to everyone for their advice & time, hopefully this thread will come to a close and I will have learned something more from everyone’s time.
Neat Image is great, but if overdone can look plasticy (not a real word). If one desires significant smoothing but doesn’t want this look, one can duplicate the background layer and apply Neat Image on the top layer and then mask the top layer. Painting shades of grey of the top layer mask will often give a more realistic look.
[quote]TempusMn, this sounds interesting, I’m trying to picture this in my mind here…flip it which way to cancel out the garbage? All I can see is scanning the other side than the emulsion but that puts everything backwards.[quote]
Flip the image vertically so that the top right hand corner becomes the bottom left hand corner and so on. Not over to the back side. In other words, rotate it 90 degrees on the scanner bed.After you scan the image you’ll have to flip it vertically to match your bottom layer. you’re effectively just using your scanner light to illuminate the image from the opposite direction so that the shadows and highlights fall the other way on the textured surface. I’m assuming here that your image is not a transparency.
These are medium format positive transparencies. But now I get the idea. Flip it so the scanning light hits it from the opposite angle then flip the image right side up and the different angles should cancel out. Is that a standard solution to 3d images or for correcting actual grain probs?
Older images often have a surface texture on the actual paper it’s printed on and sometimes 35mm slides will as well. I had gotten the impression that these images had been scanned from prints of some sort by the look of them and by some of the comments further back. flipping and rescanning is only going to help out on reflective images, paper, tin, cardboard, etc, that have a surface texture that’s being picked up in the scan. I tried it out on a 35mm slide to some minimal effect, but on film where the damage is "in" the image and has light passing through it, I don’t think flipping the image is going to work. You’re back to layers and masks and maybe some grain removal software if that’s the case.
Sorry this didn’t help you with your current project. Maybe another time it’ll be useful. Good luck!
2 1/4" slide projectors are manufactured by several major companies.: Hasselbad, Bosch & Lomb, Rollei, etc.
You trans may have been done for and educational purpose; art course, architecture, etc…
HTH
Mr3
wrote in message
Morning….
I have some medium format trans’ that are dupes of the originals. The
grain was only multiplied when copied to the new film. Now I have the task of taking the 2×2’s and making 48×48 out of them and the grain is so bad it looks like chunks of wood floating around in the emulsion 😉
The grain is course, the shots are mostly outdoors, some with heavy
texture and others not. Other than hand mousing over the entire image, what would you guys suggest to remove course grain without making the image too soft? I was thinking is there a way to make a couple of different masks and apply gaussian only to those masks without losing sharp??? Luminance mask for the highlights grain?? I’m not at the studio to try it and won’t be for a few days.
I know it can’t get rid of everything, but it appears even with a JPEG it did more than just blur as I originally thought it would do. Would running it over again and resampling the new "grain" help at all?
Yes, that is a good technique. As a matter of fact I used three passes on the sample I uploaded.
Do you feel Neat Image 3.0 Pro is much better than 2.61 or do they accomplish the same results?
I think 3.0 is significantly better than 2.61. Neat Image 3.0 Pro has the following improvements and new features:
The noise reduction engine has been improved and has increased processing speed. Neat Image now works 2.5 times faster on a standard PC and even faster on HyperThreading or multiprocessor systems. A Profile Matcher was added to automatically select a device noise profile most suitable to process the input image based on an automated inspection of the image. A Variant Selector has been added to make it easier to optimize the filter settings. Photoshop CS compatibility has been added. There is an improved graphic user interface for Windows XP. JPEG comments are now preserved by the Neat Image standalone. For details, see the Neat Image "History of Changes" page:
In short, Neat Image 3.0 is faster and better than 2.61 and well worth the upgrade. Since I was a recent purchaser, the upgrade was free, which is even better still. Incidentally, there is a new 50-page Neat Image 3.0 User Guide, which you will want to download.
Neat Image is great, but if overdone can look plasticy (not a real word). If one desires significant smoothing but doesn’t want this look, one can duplicate the background layer and apply Neat Image on the top layer and then mask the top layer. Painting shades of grey of the top layer mask will often give a more realistic look.
Good points and good tips. No doubt that Neat Image can be overdone, and those Photoshop techniques would help remedy the situation. Neat Image also has percentage sliders to let you control the amount of each of its effects.
Another technique I use when a Neat Image result looks too smooth is to bring it into Photoshop, if it isn’t already in Photoshop, and do a Filter> Noise> Add Noise. That lets you add a uniform, controlled amount of "good noise" to replace the "bad noise" that Neat Image removed. A little noise makes an image look more "real". At least, it does for me, although I guess that is a subjective thing.
Incidentally, I think that "plasticy" should be a word. Those dictionaries are missing out on a good thing.
Anyone still subscribing to this thread? WELL, in all I found out what the problem was with that crappy grain in the scan. Are you ready?
ANTI-NEWTON bed glass!!!!!!! The scitex Jazz + we have has the anti-newton glass on the bed, which has a slightly hazey appearance when reflecting light, caused the apparrent grain we saw. Why I didn’t just scan an empty glass and look at the results I don’t know, but a blank scan shows the exact same crumpled paper look as the trans scans did. We went to a fellow shop and used their Jazz + which incidently has NO anti-newton glass on it, and the scans show up fine. We just sent out for a piece of standard glass to swap with the bed glass and we’ll see how much better the scan is.
