Trying to understand the math of scanning

Kavika

My goal - to create large B/W prints from 35 mm color slides.
First task - How to scan and convert the slides?
History - Ten years ago, I had posters made from a 35mm slide that were excellent, sharp! This was done by a large commercial printer, with 4 color separations, etc. I think the scan was done by a Scitex ??? The poster was 24 x 36 and the image was 18 x 24.
Questions - How can I do this on my own? What scanner will produce a file good enough for that size enlargement? How big a file do you need to create a sharp 18 x 24 image.

I recently tried a flatbed scanner and was not happy with the results; just not sharp enough, like my earlier poster. And the file, saved as a tiff, was close to 100MB ! If a 10 or 20 MP camera will do the job, why does a scan have to be 100 MB? Please help this dumb blonde with the math!:dunno

Angelo

I say embrace your inner blonde!

I can't help you with the math but whenever I've needed to scan 35mm negatives or slides I've had them done on a drum scanner.

A drum scanner is a transparent vertical tube housed inside a light array housing that scans by transmitting rays through the negative / slide and captured on an imaging plate for the best possible resolution. Really the only way to go, in my book.

If you seek out a high-end printing house or film developing company in your area they should be able to help.

Expect to pay about $50 per scan with a resulting file about 30mb. I've always asked for a TIFF file on disk and recommend you do the same.

ziggy53

Kavika,

The first thing to understand is that there is no direct correlation between digital and film.

Digital (CCD/CMOS and variants) is a regular array of fixed pitch and sized image elements, converted from discreet individual colors into a mapped sequence of calculated pixels (Bayer technology).

Film images are discreet image elements of randomly arranged, unevenly sized, granular, semi-transparent crystals.

Since you are going from color transparency to gray scale, you may be thinking that there is a direct translation of tonality from color to gray. That may be true but probably is not. You may be better off emphasizing particular colors to enhance contrast between color differentials.

There are many methods for converting color images to gray scale, but a good thread is:

http://www.dgrin.com/showthread.php?t=2070

I suggest the best starting strategy is to overscan with the highest resolution possible, and the highest color bit depth possible, trying to capture the greatest detail and color quality intrinsic in the image. It takes 4 times the resolution to accurately scan colors, because a Bayer chip has 3 color filtered pixels, RGB, but twice as many Green pixels, so an RGBG pattern or some such. You also want to preserve as much of the original grain structure as possible, which means each grain in the image has an actual shape, which should be preserved for greatest image detail. Obviously, to capture the shape and size of each grain, there have to be multiple pixels, since a single pixel doesn't measure or represent size. (Grain aliasing is a common manifestation of too little resolution.)

Drum scanning is a completely different beast, and rather than describe the advantages, I will defer to another link:

http://www.luminous-landscape.com/reviews/scanners/drum_scans.shtml

So the answer to your question is dependant upon the film type and grain structure, and also the condition of the image on the film, dust, scratches and mold/mildew contamination, etc.

Kodachrome is treated somewhat differently in that it has a smoother, more continous tone when projected, but a very unique and evident grain when scanned. Wet mounting with a drum scanner is highly recommended for Kodachromes.

There is also a fairly new method for wet mounting in a tabletop scanner:

http://www.wetmounting.com

After scanning, but before the transformation to gray scale, (before any other process, for that matter), I also recommend a pass through a grain reduction software. Noise Ninja and Neat Image are great for this process.

ziggy53

colourbox

Kavika wrote:

And the file, saved as a tiff, was close to 100MB ! If a 10 or 20 MP camera will do the job, why does a scan have to be 100 MB?

In addition to there being no direct correlation between digital and film, there is no direct correlation between quality and file size.

When someone says "You need a 20MB file" that really does not tell you anything.

You can scan an image and have it come out 10 different file sizes depending on which buttons you push. One would guess that typically, when pros talk about a file of X MB, they are probably assuming it is 8 bits, RGB, TIFF, probably LZW compressed, no alpha channels, no layers, unknown pixel height and width. You change any one of those variables and the file size would change dramatically with no visual difference. The same file could be 100K (compressed JPEG) or 100MB (big burger with everything on it).

On top of that, the reason drum scanners are brought up is that the quality of the scan depends to a critical extent on the quality of the analog-to-digital conversion. (OK, I guess that part is a lot like a digital camera.) Make a scan with a $149 scanner and it could be noisy with bad color. Make a scan with the exact same specs on a drum scanner and it could come out with no noise and spectacular color, and would enlarge much better.

Angelo

Wow, great input guys!



I just came across this...

http://pogue.blogs.nytimes.com/2006/11/21/21pogues-posts-2/

ziggy53

Angelo wrote:

Wow, great input guys!



I just came across this...

http://pogue.blogs.nytimes.com/2006/11/21/21pogues-posts-2/

Thanks Angelo.

DavidTO already pointed out that article:

http://www.dgrin.com/showthread.php?t=47871

My main observation there is that we, the readers, know nothing about the way the images were displayed nor about the print method used to make the prints.

The system may have been insensitive to any more pixels than the minimum used, but that doesn't mean that every image and every system are so limited, in fact I could easily prove the opposite, choosing a system that easily shows the difference.

ziggy53