|Page (1) of 1 - 01/02/02||email article||print page|
Post-Multiplication in Digital Fusion
When dealing with semi-transparent layers in Digital Fusion, there are two ways to deal with them. In this tip I explain how, and more importantly, why you need the additive/subtractive slider in the merge tool.
Mathematically, there are two ways to make a merge. The first is to use the alpha channel from the foreground to cut, or subtract, a hole from the background image. Then you simply fit the foreground image into that hole. That's known as a 'subtractive' merge. The second way is to add the pixel values from the foreground to the background, using the alpha channel as a guide.
Above you can see a non-pre-multiplied image, which doesnt show any anti-aliasing as well as a raw color image. Next to it is a pre-multiplied image, which is anti-aliased, and requires an additive merge.
When you are placing semi-transparent objects in front of other elements, the color values of the background plates should have an impact on the color values of the foreground object. Glass is the best example of this, but all transparent material is affected the same way, including the edges of anti-aliased pixels like text or 3D renderings. The advantage to this 'additive' method is that none of the color values from the background are lost when you perform the merge, so in semi-transparent areas, you can actually see the background element through the foreground.
Try merging large white text over a bright background. (I used R0.7, G0.8, B0.9.) Set the merge to subtractive, and then set the size to 2.0. See the black halo around the text? Now set the Merge to additive. The halo disappears, and the quality of the text improves. (Set your Hi-Q checkbox on for this example)
When you post multiply an image, you multiply the value of each color channel by the value of the alpha channel. For example, if I have a pixel which is opaque, then the value of the alpha channel will be 1.0. So if my pixel is blue (R0, G0, B1), then performing a post multiplication gives me:
R(0x1) =0 G(0x1) =0 B(1x1) =1
Which is no change at all. But if that pixel was semi transparent alpha (0.5), then I would get:
R(0x0.5) =0 G(0x0.5) =0 B(1x0.5) =0.5
This basically halves the value of the blue channel. If the alpha channel was transparent A(0.0), then blue becomes B(1x0) which evaluates to zero.
Keep on Fusioning
Related Keywords:compositing, special effects
Source:Digital Media Online. All Rights Reserved