User talk:Leonard Vertighel/Requests

Technical Suggestion

I did some experiments trying something similar with image conversion from colour --> greyscale --> black & white images. (see 6 Symphonies, Op.9 (Stamitz, Carl Philipp) ) I found a result using the following....

  • convert colour to greyscale (possibly adjust colours first if paper colour is off-white)
  • increase pixels by 300% - this gives a nice number for curved interpolation - namely B-spline.
  • program I used is freeware irfanview
  • command in irfanview is Image --> Resize/Resample --> Set new size: pixels and Size Method: Resample and select B-Spline filter
  • then use appropriate conversion for change of grey --> black and white as you have described.

(posted by Homerdundas)

Thanks for the suggestion. Unfortunately, irfanview seems to be available only for Windows (I'm on Linux). I'm using GraphicsMagick, which can be easily called from a script.
I'm not sure what's the rationale for the intermediate greyscale step. In my tests, this didn't seem to make much of a difference (except for the slower processing...). Did you get different results? As for the resize factor, I tried 200%, 300% and 400% (all with Lanczos resampling), and found that 200% is a good trade-off between file size and quality (tested by printing on a 600dpi laser). I'm not sure if GraphicsMagick does B-splines; I didn't find it in the manual, but maybe it goes by some other name. If I find it, I can give it a try.
I've uploaded the first files to the FTP server (folder "Munich Digitization Centre"; ask Feldmahler if you don't have access yet). If you like, you can take a look and tell me if you can get better results with different settings.
--Leonard Vertighel 02:39, 20 January 2009 (EST)

Hi Leonard. The source scans were in colour - with the full gamut of colour levels and at a lower resolution that one would want for a printable copy. My rational for the grayscale step was to reduce processing time from the colour while preserving information contained in the grey levels. I used 300% because the originals were 200 pixels per inch - and 600 is nice. I judged the quality of the result by some very slender white lines between 16th note beams. The spline seems to have made the lighter grey pixels between the beams lighter - enough that when converted to monochrome, the slender white lines were visible. Which method is best? I guess it depends on original scanning conditions etc. --Homerdundas 09:49, 26 January 2009 (EST)