Hardware Accelerated Motion Blur Generation

Clement Shimizu,Amit Shesh,Baoquan Chen

2003-01-01

Abstract:Motion blur occurs in photography by the motion of objects during the finite exposure time that the cam- era shutter remains open for to record the image on film. The traditional method of rendering a motion blur with a computer is to render the scene at many discrete time instances in every frame. In this paper, we present an efficient motion blur generation method that leverages modern commodity graphics hard- ware. Our method avoids rendering the entire complex scene many times per frame. It first renders the scene into a texture, next renders the optic flow, created based on object transformation, to a vector field texture. The scene texture is finally efficiently blurred according to the vector field using texture- mapping hardware to do a piecewise iterative line integral convolution. Though our method uses vertex velocities to calculate image pixel velocities, the line integral convolution is performed on an image, mak- ing our method largely independent of scene complexity.