Film versus digital, a noisy issue

[rjlittlefield]

This posting was inspired by Tony T's picture of the Flannel Moth caterpillar over in the closeup/macro gallery ("Cat or Mouse?").

That image is scanned from Kodachrome 64 slide film. When I look at the OOF background and leaves, it seems quite "noisy" compared to modern digital images. Yet Kodachrome 64 has a reputation for having fine grain and high resolution -- not as good as Kodachrome 25 by any means, but quite good as films go.

I have scanned only a few hundred frames of film, but what I see in Tony's image is consistent with my own limited experience. The film grain that's visible in scans of 35mm film is much worse than the per-pixel random noise in modern digital captures at equivalent exposure.*

Of course noise is only one of the issues. Another is maximum resolution, where typically film has much better numbers. But the usual numbers don't tell the whole story. There is a fascinating and non-obvious relationship between noise and resolution. It turns out that what wins for high contrast subjects can lose for low.

I took a crack at investigating and documenting this issue last summer. You can read about it in this writeup: "35mm Film vs DSLR: Gradation, Resolution, and Dynamic Range". As phrased, the comparison sounds like a moot point, given the impending demise of film. However, the results most likely carry over into comparisons of various size digital sensors, since they have intrinsically different noise levels due to the limit on how many photons they can count per square-mm of the sensor.

--Rik

* "Equivalent exposure" = same lens position, same aperture diameter, same exposure time, hence same perspective, same DOF & diffraction blur, and same motion blur (if any).

[Tony T]

"it seems quite "noisy" "
Perhaps the scanner is too "good", or perhaps I scanned it at too high a resolution. Scanned at 4000 ppi Nikon Supercoolscan 5000ED. Reduced to 72 ppi for posting.

[rjlittlefield]

The scanning process you describe does not sound like a problem.

My scanner is a CanoScan FS 4000 US. I've never noticed a significant difference between scanning at 4000 dpi and then downsampling in Photoshop, versus having the CanoScan run at a some lower resolution. After initial testing, I standardized on scanning at the highest resolution needed to capture the finest image detail. That way I only had to scan once, while keeping all my options open.

Other people have noted that different scanners produce different noise on the same films. There's definitely an issue if the scanner looks at only small spots of film and ignores the spaces between them. And it seems there are some subtletles of the illumination method, like the old tradeoffs between condensing and diffusing heads on enlargers. But even the best scans I have seen look noisy compared to modern digital.

Early digital cameras had ghastly big noise. It was mid-2001 when I got my first compact digital. I was very disappointed by the noise levels and dynamic range. In fact I wrote to a friend not too much later that digital was clearly on the way, but I thought it would be at least 5 years before I switched over. That turned out to be pessimistic. It was March 2004 that I bought the Canon Digital Rebel (300D) that I am still using. I have not shot a dozen rolls of film since.

--Rik

[Edit] BTW, "72 ppi" is not a meaningful unit for web display. That's a common spec, but at best it's a crude approximation to reality. What you have is 795 pixels wide, period. My monitor displays the image as 8.2" wide, hence about 96 pixels per inch on my monitor. In terms of the original film, it's 795 pixels in 36 mm, hence equivalent to about 560 pixels per inch scanning resolution.

[Tony T]

.
[Edit] BTW, "72 ppi" is not a meaningful unit for web display. That's a common spec, but at best it's a crude approximation to reality. What you have is 795 pixels wide, period. My monitor displays the image as 8.2" wide, hence about 96 pixels per inch on my monitor. In terms of the original film, it's 795 pixels in 36 mm, hence equivalent to about 560 pixels per inch scanning resolution.

What
The moth wing I am about to post was photographed at 300 ppi (that's what Photoshop tells me), has a size of 4288x2848 pixels and is an horrendous 69.9 megapixels. I then change the resolution to 72ppi. I then resample to 800 pixels wide. I then "Save for Web" at quite a high quality but keeping it below 200K.
So my (Photoshop's) 72 ppi is meaningless?

[rjlittlefield]

So my (Photoshop's) 72 ppi is meaningless?

Pretty much. I broke out the answer into a separate topic. See Pixel Dimensions, Document Size, and Resolution for more information.

--Rik

[Epidic]

As phrased, the comparison sounds like a moot point, given the impending demise of film.

Last time I checked film was still available. Is there anything in particular that gives you this impression? Is Kodak or Fuji going out of business?

[rjlittlefield]

I have no special insights, Will. I just read the newspaper.

[ref] The rate of decline is apparent from film sales — since only people who buy film need to have it developed. Over the last four years, the sale of film has been dropping at a rate of 25 to 30 percent each year. In 2006, 204 million rolls were sold, a quarter of the 800 million sold at the peak in 1999. “It’s pretty alarming,” said Bing Liem, senior vice president of sales for the imaging division of Fujifilm USA.

Sales of single-use cameras are declining, too, but more slowly than film, said Mr. Pageau [publisher of PMA Magazine, a photo industry trade publication]. “You’ll still get people using one-time-use cameras in places like amusement parks where people don’t want to damage their digital camera,” he said. Mr. Pageau said that about 6 percent of households in the United States used single-use cameras exclusively.

Still, Mr. Pageau said, film developing is becoming a niche service, thanks to digital cameras. “It’s all about the replacement cycles of the technology,” he said. “Most reloadable films are 35 millimeter, and there hasn’t been a significant new 35-millimeter camera introduced in two or more years.”
...
The Eastman Kodak Company’s film business has plummeted, and the company has spent the last few years making a transition to digital technology.

There is no dearth of images. In the heyday of film, said Mr. Liem, some 25 billion images were not just captured but printed as well. By 2009, as the use of digital cameras continues to grow, some 135 billion images will be captured, but far fewer printed. Instead, those images tend to stay on people’s computers in electronic shoeboxes. The challenge, say companies like Kodak and Fujifilm, is getting people to print those images out.

Kodak is aiming at third world economies where home computers are less common, to extend the life of film. It is selling low-cost film cameras throughout Asia, said Christopher Veronda, a spokesman for Kodak. “We’re seeding places with large populations and limited incomes, trying to take picture taking to those markets,” he said. In India last month, Kodak had a record sales month.

--Rik

[Epidic]

This is a very complex issue. Gain and noise are not the same thing. You are also making a comparison of a camera digital image of the subject and a digital image of a piece of film. The digitizing process affects the results. What would be useful is one of a digital print to a chemical print.

A few things to note. Chemical systems have no aliasing. At my old company we always used a seimens star as resolution changes with orientation of sensors. Granularity (grain is not the term as you are not seeing grain) changes with density. Test bars are not a good granularity test target. A step wedge is better.

[Epidic]

I have no special insights, Will. I just read the newspaper.

[ref] The rate of decline is apparent from film sales — since only people who buy film need to have it developed. Over the last four years, the sale of film has been dropping at a rate of 25 to 30 percent each year. In 2006, 204 million rolls were sold, a quarter of the 800 million sold at the peak in 1999. “It’s pretty alarming,” said Bing Liem, senior vice president of sales for the imaging division of Fujifilm USA.

Sales of single-use cameras are declining, too, but more slowly than film, said Mr. Pageau [publisher of PMA Magazine, a photo industry trade publication]. “You’ll still get people using one-time-use cameras in places like amusement parks where people don’t want to damage their digital camera,” he said. Mr. Pageau said that about 6 percent of households in the United States used single-use cameras exclusively.

Still, Mr. Pageau said, film developing is becoming a niche service, thanks to digital cameras. “It’s all about the replacement cycles of the technology,” he said. “Most reloadable films are 35 millimeter, and there hasn’t been a significant new 35-millimeter camera introduced in two or more years.”
...
The Eastman Kodak Company’s film business has plummeted, and the company has spent the last few years making a transition to digital technology.

There is no dearth of images. In the heyday of film, said Mr. Liem, some 25 billion images were not just captured but printed as well. By 2009, as the use of digital cameras continues to grow, some 135 billion images will be captured, but far fewer printed. Instead, those images tend to stay on people’s computers in electronic shoeboxes. The challenge, say companies like Kodak and Fujifilm, is getting people to print those images out.

Kodak is aiming at third world economies where home computers are less common, to extend the life of film. It is selling low-cost film cameras throughout Asia, said Christopher Veronda, a spokesman for Kodak. “We’re seeding places with large populations and limited incomes, trying to take picture taking to those markets,” he said. In India last month, Kodak had a record sales month.

--Rik

I still do not see anything than film has a shrinking market. This does not mean that it will disappear. Oil printing, theater, and sail boats are still around even though they were replace by different technologies.

[Epidic]

BTW, your 15-stop exposure latitude is a good estimate. The 7-stop claim by Clark is way off. After looking at his site it appears he is confused between scanner dynamic range and film dynamic range. I have printed lots of color negative film and I can tell you that c-41 film does not have a 7-stop range. It is far greater as you have found.

Clark is right about one thing. There can be no claim of superiority for film or digital.

[rjlittlefield]

There can be no claim of superiority for film or digital.

I'll go along with that. Different characteristics for different tasks. It happens that for most of the tasks I do, digital fits better. But there definitely are exceptions.

BTW, you might be amused to know that the reason my web page exists at all is because I wanted to defend film against defamation by a fellow on the Yahoo Microscope group who was asserting the absolute superiority of digital (ref). The illustrations of granularity and resolution for low-contrast detail sort of came along for free, and served to document earlier informal observations.

This is a very complex issue. Grain and noise are not the same thing. You are also making a comparison of a camera digital image of the subject and a digital image of a piece of film. The digitizing process affects the results. What would be useful is one of a digital print to a chemical print.

Gosh, if that were the worst criticism that could be directed against that investigation, I'd be delighted! I would guess offhand that far bigger differences would be introduced by choice of film and processing. That one roll just happened to be the slowest film that I could buy at the local drugstore, and I ran it through their 1-hour service, which from personal experience I can assure you is not the most tightly controlled photo lab you could ever want. I posted the same article to the PanoTools group, and a typical response there went along the lines of "You should test Velvia. It has much finer grain than Kodak."

But it was a same-subject, same-time, same-place comparison, and I thought the results provided a helpful explanation of the differences both between film and digital, and between the treatments of Clark (7 stops) and Kreunen (15 stops).

I didn't see any evidence that Clark is confused about scanner dynamic range versus film dynamic range. Actually, his treatment seems OK to me for what he does -- fine art and wedding pictures where there are lots of white in the scene and he wants the best possible gradation for those large important light areas. It just means he exposes for the highlights, and sacrifices the shadows.

What's not so OK is that he then turns that common but restricted case into a completely general statement, and wraps an analysis and presentation around it that at first blush seems so airtight that the fellow on Yahoo could legitimately write that "Roger N. Clark posts to Usenet and is often called upon in the photo newsgroups to defend his data and methods. Hence that web page has been sorted out by many, and has stood up to the test." [ref]. Oh well, I can't fight all the battles, and many folks seem to have strong feelings about their favorite medium.

I agree that the issues are very complex. That's why I think it's helpful to spend time teasing out which ones are core, then explaining and illustrating what's actually going on and why it might matter to a particular audience. Positive contributions toward achieving that goal are most appreciated.

Oil printing, theater, and sail boats are still around even though they were replace by different technologies.

Good point. I plead guilty to a bad choice of words, and rejoice in the thought that my little investigation may actually apply more than I suggested.

--Rik

[Epidic]

FYI, this is from a book called Materials and Processes of Photography. This is the first edtion printed in 1986. The prints are made from a series of negatives with a stop difference in exposure. H is normal exposure. The interesting thing is the last in the series, Q, is not showing it is reaching a limit. This is very much like the test you conducted. True, black and white emultions have a greater latitude than color negs, but not that much greater.

[Epidic]

Gosh, if that were the worst criticism that could be directed against that investigation, I'd be delighted! Smile I would guess offhand that far bigger differences would be introduced by choice of film and processing. That one roll just happened to be the slowest film that I could buy at the local drugstore, and I ran it through their 1-hour service, which from personal experience I can assure you is not the most tightly controlled photo lab you could ever want. I posted the same article to the PanoTools group, and a typical response there went along the lines of "You should test Velvia. It has much finer grain than Kodak."

I would not really fault you on the processing. Nor the film choice. You may find this strange, but I don't think that is as critical as some would have you believe.

I guess I find most of the stuff out there (and actually I like a lot of what you did and the image of the aerial image is a very interesting and good inclusion) is what is the relavance of all this? The examinations are at the limits of what the system can do - not the usual viewing environment. I would suggest that most of this has little to do with photography and photographers and more to do with imaging scientists and engineers. When I was getting into this stuff (that page above is from one of my first year college textbooks), that the folks actually taking pictures (the applied photo students) didn't give a rat's donkey about this stuff as they felt it had very little to do with the final image. Now in this world of photoshop where most folks can enlarge an image to silly sizes (you needed special equipment before), all this is coming up - usually on internet sites dedicated to the "truth" about one media or the other.

Now, film has grain. It is not an error or a fault. It is what film is made of. Digital cameras have pixels. There again it is what it is made of - a regular pattern of pixels is not as good as a random pattern of grain in resolution as aliasing is a factor (unless displayed on a regular pattern of pixels like the monitor you are looking at). I have noticed one thing thing with the resolution question is many cannot separate granularity with resolving power. I think the problem is that a digital image looks very smooth when enlarged implying it can be made larger. The film image is grainy and gives the impression of breaking up. I saw a review on Luminous Landscapes that concluded that a certain digital camera was as good as or better than medium-format film. One example (now removed) showed some glass windows on a modern building. The digtial image was very nice tonally. The film image was grainy. The conclusion was the digital image was resolving the target better. However, the film image was showing one pain was actually two split with a seam that was not shown in the digital image.

The other problem is how does this appear in the final image. For example, the Clark tests would make a horrible image if printed that expanded. Manufacturers, whether digital or chemical, aim for pleasing images. The contrast and dynamic range of film and sensors are not fixed (nor are they infinite), however, for an image to be appealing, there are limits placed on the response (more so in film than digital). To make a media to have a large dynamic range would not be useful. In fact, it would make pretty lousy images. All imaging systems are developed in relation to the human visual system which includes the idea of "what looks good." Yes, it would be nice to have an infinite dynamic range (and sensor so small that we would use 1mm focal length lenses) so photographers don't need to worry about exposure (and focus), but that is just blaming technology for human limitations.

And I think this is what the technology debate is about (present company excluded), placing all the responsibliity of photography in the technology. That would make this disipline a lot easier if you can just buy a new widget. If I can't expose, I just get something that can. And I think this is the bane of the craft. I doubt you or many here would say you can buy your way into good photography. It all comes back to the photographer and his/her control.

I will get off my soap box as I have strayed far beyond your intent. I enjoyed your essay and hope to see more.

[Leif]

That is an excellent test, and oddly enough I was thinking about conducting the same kind of test.

Most such tests have used a consumer grade scanner, such as a Coolscan, and that usually leads to criticisms that the scanner is the cause of the poor results from film. Clearly your tests prove that have to be incorrect.

FWIW the results agree with my own experiences, using KR64 and Provia 100F scanned with a Minolta 5400, and Nikon D70 and Nikon D200 cameras.

[Tony T]

When I was getting into this stuff (that page above is from one of my first year college textbooks), that the folks actually taking pictures (the applied photo students) didn't give a rat's donkey about this stuff as they felt it had very little to do with the final image.

OK, so why should I care, or to use your words "give a rat's donkey", about what the photo students thought (or think). Just curious.