(Digital) Noise after Scanning?

[Pridbor]

My Photo SW has a noise removal function which very effectively smoothens the image. It actually has 2 different functions for Noise removal, one-second one being Digital Noise removal, which I assume stems from the Digital Images.

My question is when I scan a negative into a digital format, whichever it might be, does it contain noise like an image from a digital camera??? Or does the noise removal actually remove the film grain, of which there's nothing left of after the application?

Rather new to the process of developing film and scanning them, so please advise, as I don't want to remove Film grain, but rather understand which film to use to get the grain I like.

Thanks

Preben

 

[Adrian Bacon]

My Photo SW has a noise removal function which very effectively smoothens the image. It actually has 2 different functions for Noise removal, one-second one being Digital Noise removal, which I assume stems from the Digital Images.

My question is when I scan a negative into a digital format, whichever it might be, does it contain noise like an image from a digital camera??? Or does the noise removal actually remove the film grain, of which there's nothing left of after the application?

Rather new to the process of developing film and scanning them, so please advise, as I don't want to remove Film grain, but rather understand which film to use to get the grain I like.

Thanks

Preben

All analog to digital converters will introduce some amount of noise as part of the digitization process. Some ADC converters are cleaner than others. The best way minimize noise is to have as large of a signal to noise ratio as possible.

As far as noise removal goes, most of it is software. There are a number of algorithms to do that sort of thing, however, I don’t recommend any of them when scanning film except for maybe light color noise removal for color film. Film grain by definition is noise, so you don’t really want to remove it as it will remove fine detail.

For color film, I would only ever apply chroma noise removal, and just barely enough cut down the visually objectionable chroma noise when viewing at 100%.

Again, all noise removal algorithms will smear and/or remove fine detail, and so should be used sparingly if at all.

 

[Kodachromeguy]

Adrian above is correct. I made the mistake of adding sharpening when scanning b&w negatives, and the software made fine details, like tree leaves, look grainy it jagged. Moral: NO sharpening during scanning. Do it later if needed.

 

[Adrian Bacon]

Adrian above is correct. I made the mistake of adding sharpening when scanning b&w negatives, and the software made fine details, like tree leaves, look grainy it jagged. Moral: NO sharpening during scanning. Do it later if needed.

For the new to digital scanning people, the optimum is:

1: digitize with as little gain and as short of an exposure time possible while still maintaining the highest signal to noise ratio possible. Basically, use as much light as you can. If you’re using a flatbed or dedicated film scanner, there’s not much you can do about this.

2. Do the initial digitization at the native sensor resolution, regardless of what the internet says the effective resolution of your scanner is. DO NOT apply sharpening, or noise removal at this stage.

3. Do any post processing at the native sensor resolution (I.e. invert, correct colors, spot dust out, etc). DO NOT apply sharpening or noise removal at this stage.

4. Scale the scan to your intended archival resolution if it is different than the native sensor resolution that you scanned at. Often times, this will be a smaller resolution than what you scanned at. Doing this very effectively reduces the amount noise borne out of digitizing because you effectively increase your signal to noise ratio when scaling down (a good reason why 2. above specifies using native sensor res., it’s not about actual scanner resolution, but having those additional samples to help reduce noise later down the pipe, specifically at this stage). Once at your archival resolution, if you need to, apply just enough chroma noise removal to dampen out any remaining chroma noise that is objectionable. DO NOT apply sharpening or luma noise removal.

For output:

1. From your archival resolution, scale the image to your intended output resolution, be it for print, or for digital display.

2. At the output resolution, do any additional post processing that is appropriate for the intended output. It is totally appropriate to apply sharpening at this point. Any other post processing done will vary depending on output target.

 

[PhilBurton]

For the new to digital scanning people, the optimum is:

1: digitize with as little gain and as short of an exposure time possible while still maintaining the highest signal to noise ratio possible. Basically, use as much light as you can. If you’re using a flatbed or dedicated film scanner, there’s not much you can do about this.

2. Do the initial digitization at the native sensor resolution, regardless of what the internet says the effective resolution of your scanner is. DO NOT apply sharpening, or noise removal at this stage.

3. Do any post processing at the native sensor resolution (I.e. invert, correct colors, spot dust out, etc). DO NOT apply sharpening or noise removal at this stage.

4. Scale the scan to your intended archival resolution if it is different than the native sensor resolution that you scanned at. Often times, this will be a smaller resolution than what you scanned at. Doing this very effectively reduces the amount noise borne out of digitizing because you effectively increase your signal to noise ratio when scaling down (a good reason why 2. above specifies using native sensor res., it’s not about actual scanner resolution, but having those additional samples to help reduce noise later down the pipe, specifically at this stage). Once at your archival resolution, if you need to, apply just enough chroma noise removal to dampen out any remaining chroma noise that is objectionable. DO NOT apply sharpening or luma noise removal.

For output:

1. From your archival resolution, scale the image to your intended output resolution, be it for print, or for digital display.

2. At the output resolution, do any additional post processing that is appropriate for the intended output. It is totally appropriate to apply sharpening at this point. Any other post processing done will vary depending on output target.

Adrian,

I'm not the OP, but I just want to say just how much I appreciate this post. I have long put off my "Big Scanning Project" but I have promised myself and my wife that I'm going to do it next year. I have 30+ years of Kodachrome slides, many thousands, plus some High Speed Ektachrome slides,with golf-ball grain. (I shot film with 2 Nikon bodies, so I could shoot Kodachrome and High Speed Ektachrome for low light subjects, at the same time.) I'm guessing that I have about 10 to 20 thousand slides total, before culling. I also have about 5 thousand negatives, a lot of Tri-X with some Plus-X and maybe some Pan-X. Years ago, I bought a Nikon 5000 ED scanner with SF-210 batch slide feeder. I use Lightroom as my main software application for my digital photos and I shoot only Nikon RAW NEF files.

My software choices for scanning are NikonScan, Vuescan, or Silverfast. Based on what I just wrote, do you have any observations, caveats, etc. With the volume of material to scan, I'm willing to swallow the price for Silverfast if it does the best job.

Phil

 

[MattKing]

Phil:
Sounds like a good question for its own thread!

 

[PhilBurton]

Phil:
Sounds like a good question for its own thread!

Matt,

Let's see how Adrian chooses to respond.

Adrian, do you want me to post this question as a new thread?

Phil

 

[Adrian Bacon]

Matt,

Let's see how Adrian chooses to respond.

Adrian, do you want me to post this question as a new thread?

Phil

For that much material to digitize... honestly, I’d pick the fastest route and get it all in at a reasonable quality/resolution so you can at least catalog it and figure out what material you want to have really good scans of, then go back and do just those as you have the time to do so. If you want it all at highest quality, you’re going to be at it for a while no matter what route you choose. That’s a lot of material. You can’t realistically expect to do highest quality for all of it in any reasonable time frame. Time is your most valuable asset. The cost of software and hardware is pretty secondary next to what your time costs. Paying for the fastest route that provides acceptable results is always worth it.

All that being said, I personally use Vuescan if doing scans with any of my dedicated film scanners or flatbed, however, the vast majority of what I scan nowadays is via dslr/copy stand and my own software. For me it’s about speed of capture while still providing acceptable results. Again, it’s all about time.

 

[PhilBurton]

For that much material to digitize... honestly, I’d pick the fastest route and get it all in at a reasonable quality/resolution so you can at least catalog it and figure out what material you want to have really good scans of, then go back and do just those as you have the time to do so. If you want it all at highest quality, you’re going to be at it for a while no matter what route you choose. That’s a lot of material. You can’t realistically expect to do highest quality for all of it in any reasonable time frame. Time is your most valuable asset. The cost of software and hardware is pretty secondary next to what your time costs. Paying for the fastest route that provides acceptable results is always worth it.

All that being said, I personally use Vuescan if doing scans with any of my dedicated film scanners or flatbed, however, the vast majority of what I scan nowadays is via dslr/copy stand and my own software. For me it’s about speed of capture while still providing acceptable results. Again, it’s all about time.

Adrian,

I bought the SF-210 batch slide feeder so I could do up to 50 slides at one time, unattended. I agree with your point about dong the fastest job for reasonable quality, but since I'm doing unattended scans I'm willing to set the scanner to do multi-scan (or whatever it's called) for best Dmax. I anticipate that each day, I will set up the scanner with 50 slides, and then let the scan operation run overnight. Then repeat the next day. Scanner resolution is fixed at 4000 x 6000 pixels.

Have you ever evaluated Silverfast as an alternative to Vuescan? Silverfast makes some pretty bold claims for better/unique features, and as a scanning newbie, I'm not sure that I can really properly evaluate those claims. I have read that Silverfast has an interface that is hard to use.

 

[Adrian Bacon]

Adrian,

I bought the SF-210 batch slide feeder so I could do up to 50 slides at one time, unattended. I agree with your point about dong the fastest job for reasonable quality, but since I'm doing unattended scans I'm willing to set the scanner to do multi-scan (or whatever it's called) for best Dmax. I anticipate that each day, I will set up the scanner with 50 slides, and then let the scan operation run overnight. Then repeat the next day. Scanner resolution is fixed at 4000 x 6000 pixels.

Have you ever evaluated Silverfast as an alternative to Vuescan? Silverfast makes some pretty bold claims for better/unique features, and as a scanning newbie, I'm not sure that I can really properly evaluate those claims. I have read that Silverfast has an interface that is hard to use.

I have not evaluated silverfast. My usage of vuescan is limited to getting the raw scanner sample data, which I feed into my software, so my usage isn’t really a fair comparison to anything else.

If you’re scanning mostly slides, and the scanner supports adjusting the exposure time, exposing so a clear (way over exposed) slide is just below sensor clip, then exposing everything at that exposure level and 16 bit depth will give you very good DMAX. If you want multi-exposure, you can do that, but in my experience it’s diminishing returns and significantly extends your scanning time.

6000x4000 is great. That’s plenty of resolution for archival and more than enough for display and most reasonable prints. It’s all too easy to get caught up in the “more resolution” train and kind of miss the forest for the trees so to speak. There’s “highest quality” and then there’s “good enough”. While it’s fun to investigate “highest quality”, practical reality dictates setting a reasonable “good enough” threshold so that we can actually get stuff done. From there, any time you can gain time while still maintaining that “good enough” threshold is always money well spent.

 

[Pridbor]

I appreciate your responses very much!

Thanks

Preben

 

[Kodachromeguy]

Hello Adrian, I have three questions if you have time.
1. I have a similar question as noted by Pridbor above. When I scan B&W or color negative film with a Plustek 7600i scanner (135 film) or a Minolta Scan Multi (120 film), I see grain but no obvious random noise. Am I just not seeing it?
2. The Silverfast software lets me do multi scans on the Plustek (not specified, possibly 2 times?) and 2, 4, 8, or 16 times on the Scan Multi. I usually do multi and 4x, but am not fully convinced that there is any benefit. What do you recommend?
3. What is the native resolution that you mentioned above? Is it the maximum resolution that the software lets you set? On the Minolta Scan Multi, I always use 2820 dpi, which is the max. On the Plustek, I use 3600 dpi. I have tried 7200 dpi (the max), but I am not sure if I am seeing any more actual data, just larger size grain.
Thanks!!

 

[Adrian Bacon]

Hello Adrian, I have three questions if you have time.
1. I have a similar question as noted by Pridbor above. When I scan B&W or color negative film with a Plustek 7600i scanner (135 film) or a Minolta Scan Multi (120 film), I see grain but no obvious random noise. Am I just not seeing it?
2. The Silverfast software lets me do multi scans on the Plustek (not specified, possibly 2 times?) and 2, 4, 8, or 16 times on the Scan Multi. I usually do multi and 4x, but am not fully convinced that there is any benefit. What do you recommend?
3. What is the native resolution that you mentioned above? Is it the maximum resolution that the software lets you set? On the Minolta Scan Multi, I always use 2820 dpi, which is the max. On the Plustek, I use 3600 dpi. I have tried 7200 dpi (the max), but I am not sure if I am seeing any more actual data, just larger size grain.
Thanks!!

1. Whether or not you see noise will depend how it was scanned and what you’re scanning. If scanning negatives, the noise won’t be where you think it is. It’ll be in the highlights/whites, not the shadows/blacks, but only on films that have a lot of density. If scanning black and white, it’ll be luminance noise, in which case don’t worry about it as it’s not very objectionable. If scanning slides, the noise will show up in the shadows/blacks as chroma/color noise. If you don’t see any, you either have a very good scanner, or something, somewhere along the way is removing it.

2. I wouldn’t bother. If you’re scanning at 16 bits and are already scanning where clear film base plus fog is just below sensor clip, you’ll already be getting excellent DMAX. Unless whatever you’re scanning has extraordinary density, the scanner should be able to handle it without much problem unless it’s crap. My 14 bit dslr handles slide film dynamic range with room to spare. A scanner with a 16 bit ADC should be able to stomp on it in the dynamic range department.

3. You’ll have to look it up. I think the plustek is natively 7200 dpi. Don’t know about the other one. Avoid interpolated resolution. Scanning at the native sensor resolution isn’t really about getting more resolution as they’re all optically limited to one degree or another, it’s about not prematurely limiting what you’re going to capture early in the chain, and keeping the software out of the way as much as possible early in the chain to avoid inadvertently doing not so great things to the image quality, like only sampling every other pixel off the sensor to get the lower resolution. That sort of stuff introduces all kinds other image quality problems. The easiest and simplest way to avoid all of that is to scan at the native sensor resolution and bit depth, then scale down later. That being said, just because you can’t see more detail doesn’t mean the scanner isn’t capturing more detail. If it’s below 50% contrast response, you’ll be hard pressed to see it without applying some processing to dig it out. At any rate, you tend to get better results by scanning at the native resolution, then scaling down to a “house format” for archival, it helps to give the lower resolution archival format a bit more fine detail and helps to reduce any noise that may be present. You’re basically oversampling when you do that, and I’ve yet to see anybody publish anything that shows that oversampling is to be avoided. Also, the different colors will capture different amounts of spatial resolution, the red channel the least, the blue channel the most.

 

[Les Sarile]

My Photo SW has a noise removal function which very effectively smoothens the image. It actually has 2 different functions for Noise removal, one-second one being Digital Noise removal, which I assume stems from the Digital Images.

My question is when I scan a negative into a digital format, whichever it might be, does it contain noise like an image from a digital camera??? Or does the noise removal actually remove the film grain, of which there's nothing left of after the application?

Rather new to the process of developing film and scanning them, so please advise, as I don't want to remove Film grain, but rather understand which film to use to get the grain I like.

Thanks

Preben

It sounds like you understand the difference between film grain and digital noise. The plain definition of digital noise after a scan is aberrations introduced that is not on the original film. Of course this means you have to be able to verify optically whether or not it is on the film. I have used something like a 400X loupe/microscope and a lightbox early on.

The details you can distinguish with this can be resolved by a 4000dpi scan from a Nikon Coolscan. Another use is to verify if the detail is on the film and not resolved by the scanner. So if it is not on the film, it has to be digital noise and there are many ways this can be manifested.