< Previous | Contents | Next >
The following nodes are used to re-combine channels in different ways:
— Channel Boolean: This is a 3D node used to remap and modify channels of 3D materials using a variety of simple pre-defined math operations.
— Channel Booleans: Used to shuffle or rearrange RGBA and auxiliary channels within a single input image, or among two input images, to create a single output image. If you only connect a single image to this node, it must be connected to the background input to make sure everything works.
— Copy Aux: The Copy Aux node is used to remap channels between RGBA channels and auxiliary data channels in a single 2D image. The Copy Aux node is mostly a convenience node, as the copying can also be accomplished with more effort (and flexibility) using a Channel Booleans node.
— Matte Control: Designed to do any combination of the following: (a) re-combining mattes, masks, and alpha channels in various ways, (b) modifying alpha channels using dedicated matte controls, and (c) copying alpha channels into the RGB stream of the image connected to the background input in preparation for compositing. You can copy specific channels from the foreground input to the background input to use as an alpha channel, or you can attach masks to the garbage matte input to use as alpha channels as well.
Understanding Premultiplication
Now that you understand how to direct and recombine RGB images and alpha channels in Fusion, it’s time to go more deeply into alpha channels to make sure you always combine RGB and alpha channels correctly for each operation you perform in your composite. This might seem simple, but small mistakes are easy to make and can result in unsightly artifacts. This is arguably one of the most confusing areas of visual effects compositing, so don’t skip this section.
When alpha channel and RGB pixels are both contained within a media file, such as a 3D rendered animation that contains RGB and transparency, or a motion graphics movie file with transparency baked in, there are two different ways they might be combined, and it’s important to know which is in use.
— Unpremultipled (Straight): An RGB image unaltered by the semi-transparency information in a fourth channel (alpha channel)
— Premultiplied: An RGB image that has each channel multiplied by its alpha channel before compositing.
The term Premultiplied alpha is a term that has historically been used by editors, visual effects artists, and motion graphics designers, but it’s imprecise. The alpha channel itself is not multiplied. The R,
G, and B channels are multiplied by the alpha. In the end, the alpha channel stays the same, but the values contained in the R, G, and B channels are modified.
A RGB image (left) and its alpha channel (right)
Non-premultiplied images, sometimes called “straight” alpha channels, have RGB channels that are unaltered (not multiplied) by the alpha channel. The result is that the RGB image has no anti-aliased edges and no semi-transparency. It’s usually obvious where the RGB image ends and the alpha matte begins. The image below is an example of the ragged edges seen in the RGB channels when using a non-premultiplied alpha channel. But notice the smooth semi-transparent edges found in the alpha.
A detailed view of a non-premultiplied RGB image (left) and its alpha channel (right)
A premultiplied alpha channel means the RGB pixels are multiplied by the alpha channel. This method guarantees that the RGB image pixels include semi-transparency where needed, like anti-aliased edges. Most computer-generated images are premultiplied for convenience, because they’re easier to review smoothly without actually being placed inside of a composite.
A detailed view of a premultiplied image (left) and its alpha channel (right)