Depth of field rendering via adaptive recursive filtering

Abstract

We present a new post-processing method for rendering high-quality depth-of-field effects in real time. Our method is based on a recursive filtering process, which adaptively smooths the image frame with local depth and circle of confusion information. Unlike previous post-filtering approaches that rely on various convolution kernels, the behavior of our filter is controlled by a weighting function defined between two neighboring pixels. By properly designing this weighting function, our method produces spatially-varying smoothed results, correctly handles the boundaries between in-focus and out-of-focus objects, and avoids rendering artifacts such as intensity leakage and blurring discontinuity. Additionally, our method works on the full frame without resorting to image pyramids. Our algorithms runs efficiently on graphics hardware. We demonstrate the effectiveness of the proposed method with several complex scenes.