[Gimp-user] Test images and test suite (was Re: [Gimp-developer] GIMP should fork babl and GEGL)
On 11/06/2014 11:24 AM, Gary Aitken wrote: Even if the result is not obvious visually, I need a heads-up so I can pay close attention to what's happening and undo the operation if appropriate. On 11/07/2014 04:25 AM, Simos Xenitellis wrote: Is there a test suite available that could show the expected behavior? If not, let's try to build one, both the test-suite code plus the images (before/after). As Gary hinted, when doing X to an image, whatever X might be, sometimes results are visually obvious and sometimes not. Many times when editing images, I've done X and thought X was just fine, only to realize later that X had changed the file in an unwanted way that only became obvious several editing steps later. And sometimes X can make a change in an image that the monitor's limited color gamut can't display, that might only pop up visually when the image is printed or displayed on a wider gamut display. There are a lot of nice copyright-encumbered test images available on the internet, but not many copyleft test images, and not many images that reflect the full range of colors as captured by digital cameras and then interpreted using typical matrix input profiles. When I was testing unbounded sRGB image editing, the first task was to put together some appropriate test images and then figure out which tests needed to be made, as per Simos's test suite. A good test suite of copyleft images would be a nice thing to have, whether for testing new and existing editing algorithms, or for whatever testing that individual GIMP users might want to do on their own (printer-related, for example). Best regards, Elle Stone ___ gimp-user-list mailing list List address:gimp-user-list@gnome.org List membership: https://mail.gnome.org/mailman/listinfo/gimp-user-list List archives: https://mail.gnome.org/archives/gimp-user-list
Re: [Gimp-user] Test images and test suite (was Re: [Gimp-developer] GIMP should fork babl and GEGL)
Hi, On Fri, Nov 7, 2014 at 1:07 PM, Elle Stone ellest...@ninedegreesbelow.com wrote: On 11/06/2014 11:24 AM, Gary Aitken wrote: Even if the result is not obvious visually, I need a heads-up so I can pay close attention to what's happening and undo the operation if appropriate. On 11/07/2014 04:25 AM, Simos Xenitellis wrote: Is there a test suite available that could show the expected behavior? If not, let's try to build one, both the test-suite code plus the images (before/after). As Gary hinted, when doing X to an image, whatever X might be, sometimes results are visually obvious and sometimes not. Many times when editing images, I've done X and thought X was just fine, only to realize later that X had changed the file in an unwanted way that only became obvious several editing steps later. And sometimes X can make a change in an image that the monitor's limited color gamut can't display, that might only pop up visually when the image is printed or displayed on a wider gamut display. There are a lot of nice copyright-encumbered test images available on the internet, but not many copyleft test images, and not many images that reflect the full range of colors as captured by digital cameras and then interpreted using typical matrix input profiles. When I was testing unbounded sRGB image editing, the first task was to put together some appropriate test images and then figure out which tests needed to be made, as per Simos's test suite. A good test suite of copyleft images would be a nice thing to have, whether for testing new and existing editing algorithms, or for whatever testing that individual GIMP users might want to do on their own (printer-related, for example). That's a good idea. What kind of images would be the most interesting? Basically should that be images, taken with a good digital camera, of a lot of objects of various colors? It could also be images with color gradients, I guess (sunset/rise and such)? Or do you already have such copyleft images at your disposal that you could provide? Indeed if we could gather these for access to anyone as reference (then various software could use them for their own tests), it would be great. Jehan Best regards, Elle Stone ___ gimp-user-list mailing list List address:gimp-user-list@gnome.org List membership: https://mail.gnome.org/mailman/listinfo/gimp-user-list List archives: https://mail.gnome.org/archives/gimp-user-list ___ gimp-user-list mailing list List address:gimp-user-list@gnome.org List membership: https://mail.gnome.org/mailman/listinfo/gimp-user-list List archives: https://mail.gnome.org/archives/gimp-user-list
Re: [Gimp-user] Test images and test suite (was Re: [Gimp-developer] GIMP should fork babl and GEGL)
On 11/07/2014 10:25 AM, Jehan Pagès wrote: A good test suite of copyleft images would be a nice thing to have, whether for testing new and existing editing algorithms, or for whatever testing that individual GIMP users might want to do on their own (printer-related, for example). That's a good idea. What kind of images would be the most interesting? Basically should that be images, taken with a good digital camera, of a lot of objects of various colors? It could also be images with color gradients, I guess (sunset/rise and such)? Or do you already have such copyleft images at your disposal that you could provide? Indeed if we could gather these for access to anyone as reference (then various software could use them for their own tests), it would be great. Jehan Nicolas Robidous's test image collection is very nice, in particular the baby's face and the brightly colored buildings make great test images. His images are already converted to sRGB, which means they can't really fully exercise the color gamuts of reasonably decent printers and wider gamut displays. I can make available straight from the camera interpolated raw file, no enhancements added images of very saturated (outside the sRGB color gamut) natural objects, mostly flowers. I've also put together various artificial color ramps, granger rainbows, stepped gray scales, and such. And I have IT8 target shots from several cameras, which I think the photographers would release under an appropriate license. I wish that I had a nice set of pictures of people's faces, young to old, male and female, of diverse skin colors. Skin tones are something that everyone wants to get just right, so faces make great test images. Such photographs ideally would be shot raw under natural daylight, more or less full frame, and properly white balanced, preferably with a white balancing object discretely placed somewhere in the image frame (styrofoam cups, PVC plastic, white coffee filters all work really well, often as well as commercially available white balancing aids). High quality images with good gradients would be a nice addition to a collection of test images. Interpolated raw files that have been output in a wider gamut color space would be more versatile than images that have already been converted to sRGB. Here are links to some sample collections of copyrighted test images: http://www.pcstats.com/articleimages/200601/samsungspp2040_300dpi1.jpg I would love to have enough copyleft images to put together a copyleft composite similar to the one in the above link. http://www.northlight-images.co.uk/article_pages/test_images.html http://www.outbackphoto.com/printinginsights/pi048/essay.html Thinking more about what kind of images, it depends on who's testing what. Here are some possible reasons for wanting test images: * Testing scaling algorithms. * Testing ICC profile conversions from wider gamut color spaces to printer profiles and/or to display screen profiles. * Testing the quality of prints made by a commercial or personally owned printers. So the first question is: What kind of test images, for what kinds of testing, do you all, as a diverse group of GIMP users and developers, wish you had access to? Elle ___ gimp-user-list mailing list List address:gimp-user-list@gnome.org List membership: https://mail.gnome.org/mailman/listinfo/gimp-user-list List archives: https://mail.gnome.org/archives/gimp-user-list