Greetings everyone,

Something just recently hit me up side the head. Something that I should have thought of a LONG time ago.

In still photography people have been known to take images with the lens cap on. These images (various processes for various forms of noise) are then subtracted from the original accurately removing thermal and dark current noise. Could something similar work with a ground glass? That would be a _very_ helpful means as we could use just about and GG and since the ground glass is static (I'm trying for a static version. Can't stand having to power one of these things) the grain would be consistent in its placement.

EDIT:

Does anyone have a GG handy that they can experiment with? Maybe you could attach the adapter to your camera and then point it at a light source so the only thing visible is the glass and its imperfections. Then pull out a single frame (would work best with a progressive scan camera), invert it and subtract it from the main footage?

I think this COULD work with a non compressed format, the MiniDV has a sight compression that IMHO make useless to subtract the plain noise fo the GG.
I think using a digital camera with cap on whould lead to a complete dark video, without the grain image. Keep in mind that photography has a way higher resolution that a DV video...

That's of course IMHO...maybe someone else can add something :D

Why do you think the compression would cause that much of an effect (not saying you're wrong just curious)?

Quite right about the lens cap idea. I was just suggesting that we try an approach similar to that used to remove dark current noise (which is achieved by shooting a frame with the lens cap on). In our case you would have to point the video camera at a light source or relfected light source so you only pick up the ground glass and its imperfections.

This is a great idea -- something Rai and Markus talk about in their HD setup which, on a very cursory, theoretical level I understand, but implementing the reality of it in practice is over my head. It's like sampling the noise of a room and using an audio tool to extract those sounds from dialogue -- see, I'm great with a metaphor but terrible with the code/filters for doing so.

I'm right now climbing Sysiphus' mountain of Flash and ActionScript, so I can't volunteer much time to work on the job, but someone with a working knowledge of VDub or AviSynth should be able to cook something like this up.

Please keep us posted -- if someone's willing to take the job, shows results, and lives near NYC (or visits occasionally), I'll buy you a beer(s) or your beverage of choice :)

- jim

this i took from http://medfmt.8k.com/mf/vision.html

When using ground glass focusing screens, as on a 4x5" view camera, the ground glass is hard to focus critically due to the coarse grain size of the glass. One old timer's press camera trick is to use a drop of oil to smooth out the surface of the ground glass. The result is an oil spot on the glass that is smoother, rather brighter than the untreated glass, and useful for critical focusing efforts. I presume the brighter oiled screen is due to less light being scattered from the coarse ground glass surface, thanks to the oil.

The author suggests Linseed Oil. Could this be used to elminate the problem of grain, or will the oil show up too much or make too soft of an image?

Matt: I believe someone mentioned this idea somewhere else on this board as well. It would certainly work. I'm not sure how practical it would be for a mini35 type system though as oil won't stay on the GG for very long.

Jim: I was unaware that Rai and Marcus had discussed this in respect to their HD cam. I assume they were doing this to remove dark current noise, not grain as with a GG?

I'm going to give it a shot sometime this week. I'll post any results.

I'm not sure if it was from GG or what, but it was a "fixed pattern noise" that they worked around. From the Drake camera thread:

The software is capable of eliminating issues like fixed pattern noise and gain noise. It does that already almost perfect, what is left in terms of noise is changed to random noise, that looks like film grain, so not too bad.

I knew that everybody would jump on the fixed pattern noise thing. Oh, well. We know that it's there and we found a way to eliminate it. thing is that for every gamma curve you apply during shooting you have to define a huge set of paramters to get enough data so the software is able to filter the noise out.

- jim

I wonder is electronic noise and gain a pattern or is it random? Anyone have factual info on this?

Shaw-
Try a few lighting conditions and mix and match to see if this will work in the real world. As mentioned in the other thread the image of the GG will be unique under each lighting condition even at each lens stop. Never the less I'm interested in hearing what you find.

<<<-- Originally posted by Jim Lafferty : This is a great idea -- something Rai and Markus talk about in their HD setup which, on a very cursory, theoretical level I understand, but implementing the reality of it in practice is over my head. It's like sampling the noise of a room and using an audio tool to extract those sounds from dialogue -- see, I'm great with a metaphor but terrible with the code/filters for doing so.
- jim -->>>

Do you believe this is nearly the REAL method used to remove audio noise ??? :D
This method is used to remove echoes and reverb.

Actually, that's exactly how noise removal tools work - you sample the noise without any additional sounds and then you substract it. I use it often in my work, it's a great tool.

Echoes and reverb cannot be completely removed without affecting the dry signal because the frequencies of the reverb /echo are essentially the same as those of the dry signal.

<<<--Echoes and reverb cannot be completely removed without affecting the dry signal because the frequencies of the reverb /echo are essentially the same as those of the dry signal. -->>>

A friend of mine, who works as an engineer for yamaha shown me a system that by sampling the reverb in an empty theatre hall created with a known signal (I guess it was something like a sinus) was able to remove all the reverb from a music sample taken in the same hall.
They are studiyng a system that playing (adding) a certain sound in realtime can remove the reverb of a concert hall whatever the building materials are.

Dario,

I know what you mean, but I think that this kind of device works rather as a gate, cutting the unwanted signal when the level of the useful signal falls below a certain mark.

And by the way, I tried to remove static noise from an image using the same pattern and it works very well. Seamlessly, in fact. I believe this could be the saving grace for the static ground glass system.

<<<--
I know what you mean, but I think that this kind of device works rather as a gate, cutting the unwanted signal when the level of the useful signal falls below a certain mark.
-->>>

I'm getting A LOT off topic... :D

But I knew you won't understand that :D

The system was using a 3d model of the arena, then N emitters playing a known sample (le'ts say a 4 kind of sinus sound each one) where placed in some certain points of the arena.
Each one was sampled by the system, that knowing the phisical shape and material of the arena was able to calculate in realtime how the reverb in that place was generated.
in this way they where able to remove the reverb of a (let's say) guitar in realtime.
It was not only a filter... believe me :D

I saw that demonstration because I asked this friend how I could remove reverb from a video I shoot, he told me there was a way, but he neded a model of the set and some time to sample the base signal on stage. of course it was too tricky for my needs.

Anyway... really too much O.T.... :D

Chris Rubin-
So it worked? Did you do a full test using one image of the GG and seeing if it will work for ALL shots under different lighting conditions? If it doesnt work with just one image of the GG then the system isnt practical. You dont want to have to pull a new frame of the GG whenever your lighting changes...hell whenever you pan the camera. Post some links to the frame grabs...even better...5 seconds of full rez video. I want to be proven wrong on this.

Brett,

my adapter is disassembled at the moment so I cannot post a clip right now. But here's what I came up with using what I had:

http://www.liisikoikson.com/ftp/noise_redux_ex.jpg

As far as I understand, different lighting conditions should have no effect here. Just as the camera does not 'know' what's on the other side of the ground glass (it just captures the image on the gg), neither shoud this patter 'care' what the image itself looks like.

I hope someone chimes in and posts a more thorough test.

Looks great Chris. Do you have any full res images you could post?

Also, I'm not sure how you did this with your adapter apart. Could you please explain more about your setup used to create this?

I'm liking the looks of this so far!

Aaron,

I used old footage. Just happened to find an evenly gray frame of a wall for the grain pattern. Not a full frame, hence the cropping.

I like the looks of it too. Can't wait to put the adapter back together (I just got a new gg from opto sigma, waiting for the new condenser lens to arrive) so I could put this method to a good use.

cheers,
Chris

chris is that video noise or grain from the ground glass

I'd say GG from the look of the patterns.

It's a ground glass.

Wait, so how did you capture the noise again? Explain real slow for us real slow types :)

Thanks,

- jim

This is very interesting! You might have something here. Lighting conditions will effect the appearance of the grain in the GG but from this first picture it doesnt seem to be a major factor.

Did you try it in motion? Any weird artifacting showing up?

Post some more pics when you can and let us know how it looks in motion.

<<<-- Originally posted by Jim Lafferty : Wait, so how did you capture the noise again? Explain real slow for us real slow types :)
-->>>

I Quote, can U please explain how you got the grain image ?? Still trying to figure out the way :D

Anyway, very nice results, I was very skeptic about the results....sorry. :P

Alright, here's the process (very slowly):

1. Heat up a frozen pizza in your microwave. Eat it. You don't want to get hungry during your experiment.

2. Take your camera with your static mini35 and 35mm lens attached. Make sure your on-camera lens is perfectly focused on the gg.

3. Shoot a well lit piece of paper (a bright gray wall, fridge door, whatever). Make sure the target area is evenly exposed and that the white area fills the frame. If possible, use a monitor (rather than a viewfinder) to judge the lighting conditions. Do not overexpose. Shoot the target out of focus to minimize any detail that it might have. The more out of focus the better. To minimize the camera video noise, do not use gain.

4. Now grab your cam and shoot stuff. Slow pans, out-of-focus shots and hand-held stuff is the best, because that's where the ground glass texture is most evident.

5. Capture all of that into your computer.

6. Import the gg test into your NLE timeline, desaturate it and export a single frame into an uncompressed format like tiff or bitmap. That's your gg noise pattern.

7. Import your test footage (described in p.4) into a software like AfterEffects. Also import the gg test pattern frame, have the software treat it as a clip.

8. Place both items on the timeline, with the gg test pattern as the top layer.

9. Change the gg test pattern transparency mode to 'soft light'.

10. Invert it.

11. Change opacity to 50%.

12. Tweak the opacity level until you are satisfied with the result. It also helps to add a bit of contrast to the grain pattern at this point, but it may not always be necessary.

13. You're done! Export the cleaned up clip into a new file.

14. Go heat up another frozen pizza. You must be hungry by now.

Note: this process is subject to change as more tests surface. Some people prefer noodles to pizza.

Hahaha! Nice!

I'm still waiting to try my own tests. I want to try the test with medium format stuff but I don't have access to any at the moment so I'll stick with Nikon.

Chris, that's great! Thanks,

- jim

Chris, thanks for the tip!

I've been holding the adapter to the front of my camera. LOL! Now I'll have to figure out a good way to keep the adapter still (secured to the front of the camera). Otherwise, this trick won't work for me.

You can also use a similar technique to lessen the hotspot. The only difference is the blend mode.

chris this may alter the way we want to build this. I happened to see the ps tecknik in action and it had a much larger grain than the 1500 we are using it was maybe 500 or less ,maybe we want to use the same to allow more light

And just to add to the ground glass wonder:

I ran across _the_ perfect diffuse material for our needs. Well maybe not perfect... but I'll let you guys be the judge of that.


Holographic diffusers. These suckers do not create hotspots (no need for condensers etc!), they are quite fine grained, and most importantly, especially for those of us pondering the use of slower medium format lenses, they can allow OVER 90% light transmittance! That's at least twice that of a good GG!

EDIT: http://www.mdatechnology.net/techsearch.asp?articleid=33

Richard,

I'm not so sure about that. I just received a 1500 grit gg from Optosigma a few days back and its light transmission capabilities are the best i've seen so far. I have not tried it in front of the camera yet, but it seems that as far as the stop loss is concerned, I'm a happy camper.

I would also like to add a tidbit to my previous tutorial:

the gg noise pattern should be evenly gray (RGB 125), not brighter or darker. Inverting the noise pattern should not change the overall lightness of it, it should just 'reverse' the grain.

It is also important to play around with aperture settings on both lenses (cam & 35mm) to find the spot where the gg structure is most clearly visible.

cheers,
Chris

my first test shows promise need a better capture of the grain

Chris - great work! Let us know if it hold up to being put in motion and viewed on a TV screen. If so they question becomes does it help any to make your adapter oscillate? I know this method wont work on moving GG. What Im wondering is the quality difference between a shot with moving GG and one that you have the motor stopped and you try Chris' methods on that footage. My guess is the moving GG will be better but by how much? Perhaps it we wont have to move the GG anymore...I'm doubtful but then again we'll see.

So Chris is this process render intensive or will this be pretty eazy for us to deal with on full length projects?

Brett,

I'm sure that as far as gg grain is concerned, moving gg will always be better. But for me getting rid of the moving parts and the need to power them has been an important goal. Looking at stuff like 'Life without Memory' (enormousapparatus.com), I've come to a conclusion that artistic value of depth of field, if shot well, will overrun any grain problems, and if there's a way to minimize the gg structure to a point where it is almost invisible, that's good enough for me.

About the render time - it's pretty fast (less than 1s/f), so no problem there.

The motion holps up perfectly, because the only things that change are the non-moving elements in the image (static grain).

Another thing...
Dario Corno mentioned earlier the problems that could arise from the DV compression. He's got a point. I believe that in-camera processing (such as sharpening) could be more of a problem here although I did not see any of that come to play in the few tests that I did. I'm looking forward to trying this approach with Juan's Andromeda. As it has a resolution that exceeds HDV and 12bit output, the gg grain should be resolved much more accurately and thus should be more easily sampled and removed.

this sound cool. so the higher the rez the more you will see the grain to sample it out . plus the precision to make a ocillating adapter is beyond most of us here. I think the total parts list for a static adapter is under $150. with the chris rubin workflow this
may be enough for most members.

very interesting Chris. Do you have some footage you could post?

a thought just occured to me. if we were to attach the static adapter to a 35mm slr camera ,and get a super sharp picture of ground glass we are using. we could make this way better than high def. It sounds like the goal is to get the best possible sample of the grain pattern from our ground glass.

Hrm...

You would have a high-rez picture of the ground glass, with all the grain detail...but when you applied that back over the low-rez (720x480 SD) footage, it would possibly re-introduce grain into the footage. Some of the finer grain that wasn't picked up originally in the footage would be put back over it and it wouldn't really cancel out.

Or...maybe it would work. Just a thought.

I'd have to agree with Frank. I'm not sure we need anything higher res though. We can only record so much res so capturing - and removine - more detail than exists in the original would probably be a bad idea.

then maybe a non compressed raw digtal photo.

If the dv compression is a problem

higher rez capture of the grain will not help. The grain capture method has to be as similar to the actual footage capture as possible.

It is true, that higher rez 'noise print' would be more accurate, so this looks to be something for HDV. Although MPG compression creates artifacts during fast movement, which will not be there on the grain pattern. So it seems that for static noise removal HDV can be both good and bad.

The best results obviously come from uncompressed / mildly compressed footage, which we cannot get yet.

Chris

Chris.

With the night-vision, I have a process I use which limits but does not entirely eliminate the scintillation noise in the relayed image.

I call it layer stacking though I think frame blend in After-Effects amounts to the same thing.

Basically, scintillation noise endures for only one frame or so. If one makes multiple copies of the footage and offsets them successively by one frame on the timeline and assigns them an equal transparency which is a division by the number of tracks of 100%, then individual defects in the blended frame are less apparent. The downside is a smearing effect on movement. - However ----

To make up a valid mask relatively free of camcorder noise and compression artifacts unique to individual MiniDV frames, this method may be helpful. Movement is not an issue. The only granular "noise" of note will be the groundglass grain itself which is exactly as we want it, dead still. Let's say blend ten frames at 10% transparency then any other noise on individual frames will be effectively eliminated.

This process is also useful to reduce low-light gain noise in conventional video imaging.

There's one issue with your process which would be easily resolved with regular good housekeeping. Best practice will be to shoot a short gray-scale for every setup or less thoroughly, perhaps after each unpack, transport or dismantlement and re-assembly of the ALDU rig to the camcorder to allow for slight departure of the grain texture from the reference gray-scale between shots. If shot on each setup, that gray-scale might also be accurate for the color environment of that shot

Has anyone tried, or do you think you could do a simular process using multiple video tracks in vegas? You could at least see your results without having to wait for AE to render.

I'll be trying in Vegas once I get my wax adapter done, which reminds me...*goes and sets the oven to 450*...

- jim

Furthur to above, I have been doing this "layer-stacking" in Premiere 6.

There's one small wrinkle, you have to create new tracks and use those, not Track2, 1A or 1B, otherwise the transparency thing doesn't seem to work. I am probably mismanaging the software. There is likely a keystroke method of achieving the same outcome but I have not gone searching for it yet.

It was a bit of a hit and miss guess for me as I used the "rubber bands" and on the small display, 20% transparency for five tracks is a little difficult to get exact. In practical terms, it doesn't matter a lot.

Another means of doing this for a single frame would be to export several consecutive frames and layer them in Photoshop, or Ulead, or similar.

In Ulead, I used a slightly different method to create an individual demonstration frame. I used the "Edit" "Stitch" functions on five frames, merging each new frame back onto the previously merged and saved frames at 50% transparency until I got to the last frame and set that one alone at 20%. You have to save each successive stitch as a new file and close each previous donor file before moving onto the next.

Jesus. That sounds like a real digital workout :)

Vegas allows track opacity to be set several different ways -- you can type the numbers in manually, use rubber bands, pull levels, etc.

I'm going to have to do some research to properly understand Chris's method, but Vegas should get the job done. I'll have my tutorial up sometime first week of January (given the holidays and all...)

Speaking of holidays, I hope everyone here is in good health and I wish you all the best in the coming year -- and beyond :)

- jim

I find Aftereffects most convenient for jobs like this, because you have to create a project like Bob suggested only once. With every new clip, you just replace the incoming file and you're done. Aftereffects also supports realtime preview (given your system can handle it).

The 'layer stacking' Bob describes seems to be the same as native 'echo' function in aftereffects. I've used it to clean up nighttime footage in music videos I've done.

Chris

<<<-- Originally posted by Chris Rubin : I find Aftereffects most convenient for jobs like this, because you have to create a project like Bob suggested only once. With every new clip, you just replace the incoming file and you're done. Aftereffects also supports realtime preview (given your system can handle it).

The 'layer stacking' Bob describes seems to be the same as native 'echo' function in aftereffects. I've used it to clean up nighttime footage in music videos I've done.

Chris -->>>

You should be able to just replace media in Vegas, too. Ditto the realtime preview.

- jim

hi everyone I sent a email to cris hurd with a photo of the static adapter .I will be linking to it soon

I've been trying to replicate Chris's results using Vegas with mixed progress.

In Vegas, I'm running the inverted grain image along a track, and parenting it to a copy of the footage I'm degraining. Then, I set the grain track to luma mask, and on the copy of the footage (the composite child), I'm adjusting luma up, to bring the darker grain of the image within the range of the brighter grain of the original footage.

All of this is placed over the original footage, left untouched.

I will say it *supresses* the grain, but does not remove it entirely -- it looks good for stills, but in motion it doesn't add up to a 100% grainless image, and in reality there's little aesthetic difference between one kind of grain and the other.

I tried the AE 6.5 demo today, just to see if AE would treat the image differently than Vegas using Chris's technique -- the results aren't as good as I got with Vegas, but I'm left with the impression that with a little tweaking, I could get them close to one another. Still, it's not a completely grainless image.

I'd like to be proven wrong -- Chris, if you've got some footage to show at full DV res, I'd be happy to host it for a bit. I'd really like to see the results in motion...

- jim

just wandering - did anyone tryed to subtract grain with "digital fusion" (probably now "maya fusion")?
some time ago i saw brilliant results with that software, but i never use it.
maybe someone has expirience with it?

filip