 |
| ff0268 |
Notes
March 09, 2008
Scanning
Once again it needs to be emphasized that a print is not a photo. A print is screened and has a narrower gamut compared to a photo. It is mostly recommended to optimize a scan already at the stage of scanning, requiring to adjust some settings of your scanning software. These comprise scan resolution, colour depth, number of sampling passes, scan histogram, and descreening and are for scanning a print differ from scanning a photo.
In my default workflow I use a resolution of 300 dpi for a full A4 or letter page and increase resolution for smaller originals up to a max of 600dpi. (Contrasting to photos for prints it is definitely not meaningful to go to higher resolutions).
If RAM and hard disk space is not an issue I recommend to scan at 16 bit per channel. File format should be TIFF with an appropriate colour profile embedded.
To reduce the noise generated by the scanner itself, which impacts the quality of shadows, I usually do multisampling. (Signal noise ratio is proportional to the square root of the number of samples. I routinely do 4 passes, which means the scanner noise is reduced 2 fold compared to a single pass).
Regarding the scan histogram I usually only set black and white point. Normally I set the output level of the scan's darkest tone to black and the brightes level to white in order to use the maximum number of tones the scanner hardware can provide (256 or 65536). However, sometimes doing so can provide wrong results, eg for a greyscale print which is low contrast (flat) on purpose, or for an original by purpose having a colour cast. So, although the automatic adjustments by the scanner mostly give quite good results, you should check if the scan meets your esthetic expectations. (Remember - a print is not a photo).
Also be careful with noise reduction and descreening, which in some cases can result in pictures that are blurred or distorted by secondary interference patterns. Of course this is resolution dependent, but also dependent on your hardware and on your equipment's built in descreening algorithm. To find your optimum you will need to experiment. (To be honest, even after 15 years of editing pics I still am experimenting).
Regarding scanning software I prefer 3rd party software over the software provided by the hardware manufacturers. My favourite programs are Vuescan and Silverfast, although neither of them really meets all of my requirements. Vuescan currently can be used very platform independently - this applies to both OS and scanning hardware, does not interfere with the OS, has its own universal scanner drivers, multisampling, IT 8 calibration and full 16 bit/channel support. Main disadvantage of Vusescan IMHO is that adjusting the histogram often requires fiddling quite a lot with the black and white point controls.
Silverfast also allows multisampling, 16 bit/channel operation, a very user friendly adjusting of the histogram and has a fully automated IT8 calibration. The built in descreener works well, but is way too slow to be useful for routine scanning (you would not like to wait for 10 minutes for a single descerening run to complete). Another disadvantage of Silverfast is its insufficient support for operating systems other than the 32 bit versions of Vista, Windows XP and MAC OS's prior to Version X.5. (In my case I am still desperately waiting for a Vista x64 version).
Descreening
Descreening filters are part of most better image editors. However they do not always give good results. Still the best workflow for me (in Photoshop) is to transform the pic into the LAB colour space and to manipulate the channels separately using the fiilter "Gaussian blur". Sometines the new filter “Lens Blur" is an alternative, because it preserves more edges than the Gaussion blur. I apply Lens Blur in the L channel as the first filter to reduce the moiré pattern and then remove the residual pattern via Gaussian blur. Filters working reasonably well - at least for colour prints - are eg the built in descreeners of Corel Photo Paint and Silverfast HDR, although the former only works in 8 bit/channel mode and the latter is really hungry for RAM and slow.
After descreening I transform the pic back from LAB mode in to RGB. BTW according to my experience descreening of greyscale pics seems to be much tougher than descreening of colour pics.
Well, that's it for today. The next chapter will focus on noise reduction.
You can find recent Notes archived here: