Hey everyone,
We are seeing a crazy bug in Photoshop CS3 and I wanted to know if anyone else is running into this. It appears that the alpha channel in Photoshop CS3 for an 8-bits per channel image is not actually 8-bit. Here is a simple test:
Create an alpha channel in Photoshop.
Select a brush and choose the color value (28,28,28)
Paint in the alpha channel
Watch it paint (29,29,29) in the info
Mouse over the color. It reads (29,29,29)
Additionally, 31 paints 32, etc. Has anyone else run into this? It appears that the alpha channel is actually 7-bit. If we make the image 16-bits per channel, it works as expected, however our pipeline does not handle 16-bit per channel Photoshop images. Any suggestions? Does this same case repro in CS4 or CS5?
Turns out it is a setting in the color settings that is causing it. Make sure to turn off “Use Black Point Compensation” and “Use Dither (8-bit/channel images)” in the “More Options” section of the Color Settings dialog.
Why those would be on by default makes no sense to me.
I originally ran into it when figuring out how some our DXT converter made a mess of things which were supposed to be in perfectly 4x4 pixel squares with prequantized steps of grayscale in them. As DXT works the way it works (decides on two colors and interpolates the rest based on a few magic numbers), I was baffled at the results until I started inspecting the source image alpha channel…