A layered depth-of-field method for solving partial occlusion

Abstract

Depth of field (DoF) represents a distance range around a focal plane, where objects on an image are crisp. DoF is one of the effects which significantly contributes to the photorealism of images and therefore is often simulated in rendered images. Various methods for simulating DoF have been proposed so far, but little tackle the issue of partial occlusion: Blurry objects near the camera are semi-transparent and result in partially visible background objects. This effect is strongly apparent in miniature and macro photography. In this work a DoF method is presented which simulates partial occlusion. The contribution of this work is a layered method where the scene is rendered into layers. Blurring is done efficiently with recursive Gaussian filters. Due to the usage of Gaussian filters big artifact-free blurring radii can be simulated at reasonable costs.