Sharpness on scanned 35mm

[SergioGuerra]

Greetings everyone 🙂

I have been doing about one roll/month since last year I rediscovered how I loved to photograph the world.. Although none of my photos are really good, I like them and like to tickle with old rangefinder cameras (Canonet gIII ql17, Fed3/Zorki4 + j8,j12,etc) .
The problem is that most my photos lack sharpness... I develop BW at home and colour on labs, but I only ask to develop the film cause I have a film scanner at home.
My scanner is a Agfa Duoscan T1200, and it can scan 35mm and larger formats...

If you look at my Flickr you can notice most my shots are always very grainy, even those with fp4 and colour (100asa). Am I doing anything wrong? I also notice that if i scan at more than 800dpi when i look at 100% crop i start to see pixels... (scanner soft goes till 2400).

I think i have some pretty good sharp lenses and looking at the negatives they seem pretty good, what might I be doing wrong? I want to have nice contrasty BW like I see on RFF gallery 🙂

Please take a look: http://www.flickr.com/photos/visuallab/

Thanks!

Sergio

 

[jlw]

What you're seeing is grain aliasing, a phenomenon that's caused by scanning rather than what film you use or how you develop it.

It causes images made on conventional b&w films to look much grainier when scanned than they do when printed conventionally. (Chromogenic b&w and color films don't suffer from it as much because their grains consist of soft-edged dye clouds rather than hard-edged clumps.)

Here's a link with a very comprehensive explanation of it.

"Aliasing" is just a term for any image feature that appears because of the digitizing process, rather than being present in the original. To oversimplify a bit, what's happening in your case is that the film's grain structure consists of hard edges with a random pattern, while the scanner's CCD array consists of hard edges with a regular pattern.

You can see that if a grain edge and a CCD edge happen to line up with each other, you'll get a more pronounced texture than if a grain edge happens to land in the middle of a CCD pixel or vice-versa.

Since the grain pattern and the CCD pattern are almost but not quite exact multiples of the same size, they interact as the scanner sweeps across the film to produce a larger random pattern of these enhanced edges. Your eye interprets this as exaggerated grain.

Since this is a fundamental characteristic of the way film and scanners work, there's no easy way to eliminate the effect. You can use a variety of techniques that subdue it by blurring the film's grain structure, but these also soften fine details in the image. Or you can try software filters such as Noise Ninja, which try to analyze the pattern and cancel it out -- but I've found these also affect subject textures such as fabric and hair.

One technique I've used with some success is to scan the negative twice (usually I do one scan that favors the highlights and another that favors the shadows, to make sure I'm capturing the negative's full density range.) I combine the two scans as layers in a Photoshop file and use the transparency controls to blend them together.

Since the scanner head never returns to exactly the same position for the second scan, the grain patterns of the two images are very slightly offset, and this somewhat cancels out the aliasing effect. If I need more cancelling, I'll sometimes select one layer and use the arrow keys to nudge it by one pixel in various directions, until I find a direction that suppresses the grainy appearance.

PS -- If you're mostly concerned about getting nice-looking images to post in an online gallery, you can smooth out the grain to some extent by scanning at your scanner's highest resolution, then using image editing software to downsample the file to an online-friendly size (about 600 pixels wide, maximum.) Downsampling averages several input pixels together to yield one output pixel, and this averaging process helps smooth out defects. You may need to use your image editor's sharpening functions to restore some of the crispness that will be lost through downsampling.


(Sorry for those who read my previous post from which this was copied, but this is a subject that just seems to keep coming up as more and more people try scanning their treasured b&w negatives and being disappointed with the results.)

 

[hth]

For B&W film try to increase scan resolution to double of the desired result, then after adjusting it (levels, curves, contrast etc), reduce size to the desired.

I do not seem to need to do this with color print film, but might do it anyways.

/Håkan

 

[SergioGuerra]

Oh... Thanks for the fast and very explanatory answers 🙂 I will try to use max resolution for the images I really care, and may try doubling scans and blend them in photoshop for the real important ones ..

I guess If I had a dedicated 35mm film scanner the problem would still exist, right? (I questioned myself this, although I currently dont have oportunity to buy myself one)

Thanks once again and greetings from Portugal 🙂

Sergio

 

[AusDLK]

>a dedicated 35mm film scanner the problem would still exist, right?

Yes, theoretically. My Minolta 5400 has a grain reduction option that scans using more diffused light that would be helpful.

Frankly, I always use this mode and grain aliasing has never been a concern for me.

 

[bafonso]

Sergio, always scan at a higher resolution than you plan on using at the end, be it for printing or putting up online somewhere. That said, use your good judgement. When working with crappy scanners, there's not much difference between the top two resolutions it can use normally...
It's always better to do processing at much higher "definitions" than the final. Same applies to digital audio for example.

The effects you are seeing have been explained in a previous post. There are some algorithms to get rid of the effect but as always, there's a tradeoff...

abraco
b

 

[ffttklackdedeng]

I'd recommend scanning at the physical resolution of your scanner (even though a flatbed won't really have the stated resolution) and downsampling later.

And yes, same problems with the dedicated film scanners.. But there are ways to improve / get around, the 16x oversampling of Vuescan (where the scan module is moved around for every scan) producing the best results for me on the Coolscan V.

 

[charjohncarter]

Thank you, JLW, I tried both your methods of reducing scan grain. They both worked pretty well for me. I even found that a scanned photo (not two) is improved with the transparency and move tool method. I'm not sure how to personally thank from RFF, so I hope my thanks somehow gets to you. If somebody reads this and knows how to contact JLW let me know.

 

[willie_901]

I hate to say it, but third-party scanner drivers such as SilverFast or VueScan get the most out of many scanners. At least I noticed a huge difference with my Canon 9950F using SilverFast.

For prints I scan 35mm negatives/transparencies at 4800 ppi using 16 bits per channel (48bit scans). For computer-screen viewing 2400 ppi works well. I've fought many battles over color casts, but I never had a sharpness problem that wasn't caused by the photographer (me).

willie

 

[wintoid]

Has anyone tried ALE? http://auricle.dyndns.org/ALE/