• Page
• Discussion
• View source
• History

From Odwiki

MotionBlur : Deformation Motion Blur

There are three types of motion blur supported by Houdini, TransformationBlur, VelocityBlur and DeformationBlur.

Deformation blur assumes that the SOPs being output for Rendering is changing over time. The SOP will be cooked at the beginning of the shutter and at the end of the shutter. The two sets of geometry will then be output to the target renderer.

The difficulty with deformation blur is that some SOPs (notably Simulation SOPs like particle systems) are dependent on the previous state of the geometry. This can cause problems when outputting geometry to a renderer. Consider a shutter time of 2 frames. When outputting the geometry, the SOP will be cooked at the beginning of the frame and then at a frame offset of 2 from the beginning. When the renderer attempts to generate the geometry for the next frame, the SOP will already be advanced beyond the time of the next frame, causing erroneous output.

The above case of deformation problems is an over-simplification. There can be many more complex cases where problems might occur. An alternative to DeformationBlur is VelocityBlur.

The more serious problem with deformation blur is that the topology of the geometry cannot change frame-to-frame. That is, deformation blur doesn't work very well with a particle system which is birthing points every frame. There's no way for either Houdini or the renderer to determine which point at the beginning of the frame matches which point at the end of the shutter time.

Note: DeformationBlur is a superset of TransformationBlur. That is, any transformational blur will also be taken into account with deformation blur turned on.