Technical „Did you know?“
This is going to be my first article in English, since some people told me they use translators to read the stuff I write here. I’m guessing most of my German readers don’t have problems with English, so this might be a good idea for the future, depending on how much extra effort it is to write this in English as opposed to German. This time I want to just throw in some small bits and pieces of information that you might find useful/interesting. Sort of a technical „Did you know?“.
Did you know that…
1. The Gaussian Blur filter is stronger than it should be? This is probably due to some code optimization. Therefore if you compare Photoshop’s Gaussian blur with e.g. GIMP’s Gaussian blur you only need a part of the radius for the same effect:
2. If your channels are of different brightness in the channels view than in the Channel Mixer, you should probably check your „Edit“-„Color Settings“ and set your working space Gray profile to something with the same gamma as your RGB color space. Create a new profile if needed by using the dropdown list and choosing „Custom gamma…“.
3. But wait. What does Gamma actually mean? Gamma is basically an exponential function in the form of , so this doesn’t really describe the tonal values themselves, only how they are calculated. It would be technically accurate to say the image has a tone response curve (TRC) of gamma 2.2, saying the image has a gamma of 2.2 is wrong – however it’s used quite often in this context, so I’ve adapted to ignoring it and try to think of it whenever I write it.
4. One more on this one and we’re done for today ;). The TRC for sRGB is based on, but not equal to Gamma 2.2. The lower part is a little refined for extra room in the shadows.
5. Every time you’re using the Photo Filter adj. layer your image gets temporarily converted to the XYZ color model to ensure the most realistic results.
6. The „Filter“-„Noise“-„Median“ and „Filter“-„Noise“-„Dust & Scratches“ do the exact same thing? They both create the median for the radius, but Dust & Scratches also offers a threshold so that values that are below it aren’t filtered.
7. Speaking of filtering. Did you ever wonder why blurring your image doesn’t overblur everything as the blur builds up and stacks with each pixel that is already blurred used to blur the next one? Easy solution: A temporary copy of the unblurred image is created in memory. The blur is calculated on this image and applied to the original.
8. Adjustment layers aren’t intercalculated in the layer stack. When you create a curves adj. layer that changes the output of the point 128 to 192 you can’t just invert the process by putting another curves adj. layer above it that changes the output of 192 back to 128. Instead the layers are calculated with the underlying layers until they find actual pixels to work with. This is actually the main reason why I created the „Allzweckkurven“ because this allows you to potentially reduce posterization by smoothing the curves of combined adj. layers.
9. Motion blur with an angle that is not dividable by 90° the complete image gets rotated (of course it’s quite possible that for 90° angles the image gets rotated as well, but this is a lossless rotation, so it would only have a slight performance impact). You can easily test this yourself. Just go to „Edit“-„Preferences“-„General“ and change the „Image Interpolation“ before applying the „Filter“-„Blur“-„Motion Blur“ filter with an odd angle. Compare the results of two interpolation methods and compare the results with the „Difference“ blend mode.
10. Perceptual rendering doesn’t really work for most RGB color space conversions? This is because most of Photoshop’s default RGB profiles are created as matrix profiles which only support „Relative Colorimetric“ rendering intent, so no matter what you select, this is what is used. However there is a solution for this: intermediate color profiles: http://www.luminous-landscape.com/forum/index.php?topic=56978.msg464486#msg464486
11. Did you ever wonder why the „Invert“ command doesn’t work in 32 bpc mode? Easy. Because it’s a simple 1 – value calculation; but that doesn’t work in 32 bpc where there are values way above 1 and below 0. What you can do instead though is use levels. Just set the output levels to 255 and 0 and it will invert your image and all the values above 1 and below 0 with it.
12. This one is probably rather well known, but I still wanted to mention it. DPI is not PPI. The DPI is the number of dots per inch that a printer will put on paper, the PPI is the number of pixels that are shown per inch on your monitor. It doesn’t matter how you set your PPI in your image settings, since the PPI is dependent on the physical size of your screen and your resolution. You can calculate/lookup your current PPI though to have a nice preview of the actual print size by using „View“-„Print Size“: http://www.scantips.com/no72dpi.html
So I hope you enjoyed these quick infos, if you have any questions or need more sources, I’ll try to dig them up again ;).