Real-time depth-of-field simulation on GPU

This paper proposed an efficient and real-time algorithm for the simulation of camera imaging in the three-dimensional scene based on both ideal and real camera models.The algorithm took extensive use of the multiple render targets(MRT) technology and graphics processing unit(GPU) for programmable performance.It outputed the image of the three-dimensional scene and stored as a texture,and also outputed each pixel's depth and blurriness information and also stored as texture.At last,filtered the image,calculated the size of confusion(CoC) for each pixel by the blurriness,and used CoC to blend between the original(clear) image and blurred image.Our experiments on simulation of depth-of-field in a geome-try scene demonstrate various effects with different parameters.