Compositing with 2D Vector Fields by using Shape Maps that can represent Inconsistent, Impossible, and Incoherent Shapes
Ergun Akleman, Youyou Wang, Ozgur Gonen
TL;DR
Shape maps encode a 2D shape and thickness via a 2D vector field $(x(u,v), y(u,v))$ and thickness $d(u,v)$, enabling image-based reflection, refraction, and Fresnel effects without 3D rendering. The core contribution is a reformulated compositing equation $CI = \alpha FI + (1-\alpha) \big( f\,EI(R) + (1-f)\,BI(T) \big)$, where $R$ and $T$ are shape-map-driven mappings and $f$ is a Fresnel term, reducing the process to a single additional layer. Artists can paint shape maps directly in 2D (as $(r,g,b) = (\tfrac{1}{2}(x+1), \tfrac{1}{2}(y+1), d)$) and adjust parameters such as the index proxy $a$ and thickness $d$ to produce realistic or painterly effects with optional environment maps. The method supports impossible/incoherent shapes, enables glossiness via mipmapped environment maps, and can be integrated into standard 2D workflows for fast, dynamic editing of reflections, refractions, translucency, and Fresnel control. This yields a practical, painter-friendly alternative to 3D rendering for stylized 2D composites that still convey convincing optical cues.
Abstract
In this paper, we present a new compositing approach to obtain stylized reflections and refractions with a simple control. Our approach does not require any mask or separate 3D rendering. Moreover, only one additional image is sufficient to obtain a composited image with convincing qualitative reflection and refraction effects. We have also developed linearized methods that are easy to compute. Although these methods do not directly correspond to the underlying physical phenomena of reflection and refraction, they can provide results that are visually similar to realistic 3D rendering. The main advantage of this approach is the ability to treat images as ``mock-3D'' shapes that can be inserted into any digital paint system without any significant structural change. The core of our approach is the shape map, which encodes 2D shape and thickness information for all visible points of an image of a shape. This information does not have to be complete or consistent to obtain interesting composites. In particular, the shape maps allow us to represent impossible and incoherent shapes with 2D non-conservative vector fields.
