CompuServe Thread

Replying to my own message . . . By further studying this problem I have discovered that it occurs while creating selections from alpha channels as well as masks. And I was wrong about white areas of the alpha channel, they act as expected, producing a fully opaque mask or a full selection. The problem only occurs in black areas of the mask (and possibly intermediate shades of gray, I haven't determined this yet). The mask or selection is altered by the brightness in the main image. White areas in the image (over a black area in the alpha channel) produce a fully opaque mask or selection and darker levels of brightness produce less opaque masks and selections. Black areas in the image act as expected, producing a fully transparent mask or no selection. This is probably why this problem was missed. My guess is the developers and testers used a typical rendering with an alpha channel from 3D Studio to run tests with. The default is to render on a black background. These types of images produce a mask and selection as you would expect them to (except for possibly in the single row of antialiasing pixels where it would be impossible to see). To see the full effect of this problem try rendering an object in 3D Studio with an alpha channel and with a gradient from black to white as the background. I hope you can fix this soon. Thanks Don Hayward