Simulating depth of field using per-pixel linked list buffer

Yang Liu, Purdue University

Abstract

In this thesis, I present a method for simulating three characteristics of depth of field image: partial occlusion, bokeh and blur. Retrieving color from occluded surfaces is achieved by constructing a per-pixel linked list buffer, which only requires two render passes. Additionally, per-pixel linked list buffer eliminates the memory overhead of empty pixels in depth layers. Bokeh and blur effect are accomplished by image-space point splatting (Lee 2008). I demonstrate how point splatting can be used to account for the effect of aperture shape and intensity distribution on bokeh. Spherical aberration and chromatic aberration can be approximated using a custom pre-built sprite. Together as a package, this method is capable matching the realism of multi-perspective methods and layered methods.

Degree

M.S.

Advisors

McGraw, Purdue University.

Subject Area

Computer science

COinS