[Gimp-user] Re: smart resize
[EMAIL PROTECTED] (2004-10-15 at 0609.12 +): The question is tricky. And I'm very interested in a clear answer. My opinion is that (sometimes) bicubic for reducing smooths the image too much. I found this off comp.periphs.scanners: http://www.xs4all.nl/~bvdwolf/main/foto/down_sample/down_sample.htm Another similar pages talking about filtering (demo and/or code) just in case someone wants to get deeper: http://www.binbooks.com/books/photo/i/l/57186AF8DE http://www.antigrain.com/ http://www.path.unimelb.edu.au/~dersch/interpolator/interpolator.html http://www.xmission.com/~legalize/zoom.html Basically, there are better methods than current GIMP ones (or some other common apps, for that matter). Search yourself, not the first time the issue of best quality (transformations, noise, colour, etc) appears in GIMP lists. Though it doesn't tell about 'real-life' performance (PS CS's will likely produce 'sharper' looking image), it seems ImageMagick does the better thing. Well, the typical approach to test signal processing systems is to use some kind of simple input, mostly cos it is easier to compare with output. GSR ___ Gimp-user mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-user
Re: [Gimp-user] Re: smart resize
Hi, GSR - FR [EMAIL PROTECTED] writes: Another similar pages talking about filtering (demo and/or code) just in case someone wants to get deeper: http://www.binbooks.com/books/photo/i/l/57186AF8DE http://www.antigrain.com/ http://www.path.unimelb.edu.au/~dersch/interpolator/interpolator.html http://www.xmission.com/~legalize/zoom.html Perhaps you could add these links to our Wiki? Sven ___ Gimp-user mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-user
[Gimp-user] Re: smart resize
* Marco [EMAIL PROTECTED]: Barton Bosch wrote: Hey Marco, could you tell whether and why using bicubic for enlarging and bilinear for reducing is better than using bicubic for both? I was under the impression that bicubic was the highest quality algorithm for all resizing operations. No, I couldn't. I cannot find any reason based on a clear mathematical proof. However I'm quite sure that bicubic is not a panacea for all resizing operations. In http://www.msjc.edu/m2/gall/resampling-nearestbilinear/resampling.html is given an example of better results with Nearest Neighbour instead of bicubic (because of the image nature). In http://www.mav-magazine.com/Apr1999/resample/ is described this bilinear/bicubic option of PSP. The question is tricky. And I'm very interested in a clear answer. My opinion is that (sometimes) bicubic for reducing smooths the image too much. I found this off comp.periphs.scanners: http://www.xs4all.nl/~bvdwolf/main/foto/down_sample/down_sample.htm Though it doesn't tell about 'real-life' performance (PS CS's will likely produce 'sharper' looking image), it seems ImageMagick does the better thing. -- Psi -- http://www.iki.fi/pasi.savolainen ___ Gimp-user mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-user