The W3C Scalable Vector Graphics Working Group has posted the first drafts of three specifications about SVG 1.2 filters:
According to the primer:
Filter effects are defined by 'filter' elements. To apply a filter effect to a graphics element or a container element, you set the value of the 'filter' property on the given element such that it references the filter effect.
Each 'filter' element contains a set of filter primitives as its children. Each filter primitive performs a single fundamental graphical operation (e.g., a blur or a lighting effect) on one or more inputs, producing a graphical result. Because most of the filter primitives represent some form of image processing, in most cases the output from a filter primitive is a single RGBA image.
The original source graphic or the result from a filter primitive can be used as input into one or more other filter primitives. A common application is to use the source graphic multiple times. For example, a simple filter could replace one graphic by two by adding a black copy of original source graphic offset to create a drop shadow. In effect, there are now two layers of graphics, both with the same original source graphics.
When applied to container elements such as 'g', the 'filter' property applies to the contents of the group as a whole. The group's children do not render to the screen directly; instead, the graphics commands necessary to render the children are stored temporarily. Typically, the graphics commands are executed as part of the processing of the referenced 'filter' element via use of the keywords SourceGraphic or SourceAlpha. Filter effects can be applied to container elements with no content (e.g., an empty 'g' element), in which case the SourceGraphic or SourceAlpha consist of a transparent black rectangle that is the size of the filter effects region.
