Here's an interesting example of artifacts that can be generated when doing Photoshop work in standard colorspaces. This image, in sRGB, actually started out as a gray looking rectangle. The below attached images are in standard sRGB and a modified, gamma=1, sRGB. On a profiled, color managed setup they should look exactly the same. If the monitor profile is marginal, you may see some slight variation between the left and right sides.
Then I ran a wide, blurring brush over the right side. It turned into into a colored strip. How does this happen?
Well, Photoshop is rather stupid in many of the algorithms it uses. In particular those that upsize/downsize or blur images. The reason is that it just mathematically averages the RGB values without considering the encoding gamma. The average of RGB 254,254,254 and 0,0,0 is calculated to be 127,127,127. This is fine were the gamma=1, but for gamma=2.2, it's not. The average should be done in gamma=1 which means RGB 187,187,187. The dark gray area at the top of this image is caused by the blur tool averaging in sRGB gamma.
The original images have an alternating checkerboard of pixels on the right hand side. The top is alternating black/white followed below by differing alternating pixel colors that average out as a light gray (50% reflectance). The net effect is that the sRGB image (attached below), viewed in Photoshop at 100% zoom, just looks like a gray rectangle. But it won't print that way. If you upsize or downsize the sRGB one, but not the modified gamma=1 sRGB version, in Photoshop it will show horizontal, colored strips. If you run the blur tool over it you will get the same effect shown.
The cyan/greens/yellows that follow show the same problem.
I encourage people to download the attached files, one in gamma=1 the other in sRGB gamma and try out resizing and blurring in Photoshop to get a feel for what's going on. Zoom in and look at the pixels to see how the effect is done.
This is more of an interesting demonstration as the effect rapidly disappears if the image being printed is not both super sharp and has things in it that rapidly change pixel to pixel. You might see differences in subjects like referees with black/white striped shirts in sharp images when downsized to small prints such that the stripes blend.
So, whenever I upsize or downsize in PS for printing I typically convert to gamma=1 first. With the exception of the options designed to retain detail. These look for diagonal edges and expect a normal gamma. At least for some of their work. So you might want to test both approaches using these options.
ADDED:
The attached images that also show below, will show pretty badly colored right hand sides. Trust me. They look like very nice gray rectangles in photoshop at 100% zoom and nothing like what you probably see on the browser.
