Digital Makeup Using Poisson Vector Graphics
In the digital age, portrait and self-portrait photographs are extremely popular and have spread to every corner of the world. There are many programs and apps available for photo editing, spanning from the general purpose software, such as Adobe Photoshop, to the makeover tools to beautify one’s face. These programs can edit eyes, brighten skin, remove wrinkles, reshape face and even perfect the body shape.
However, the majority of the tools are designed for only editing raster images. It is well known that vector graphics provides several practical benefits over raster graphics, including sparse representation, compact storage, geometric editablity, information reuse and resolution-independence. Early vector graphics supports only linear or radial color gradients, diminishing their applications for photo-realistic images.
Diffusion curves allow users to only manipulate the boundary colors (i.e., the Dirichlet boundary condition of Laplace’s equation) and do not support control of color gradients; therefore it is difficult to generate photo-realistic images.
Recently, the PI and his team have developed a completely new vector graphics, called Poisson vector graphics (PVG) . PVG is the solution of Poisson’s equation, whose solution space is much larger than that of Laplace’s equation, hereby providing users more control of images. Armed with two new types of primitives, called Poisson curves (PC) and Poisson regions (PR), PVG can easily produce photorealistic effects such as specular highlights, core shadows, translucency and halos
This project aims at developing a new facial image editing tool using Poisson vector graphics. This tool can automatically extract salient facial features, such as silhouettes, eyes, mouth, etc, and convert them into PVG primitives. It allows users to edit high-level features via simple slider control (e.g., enlarging eyes), and edit low-level features by fine-tuning the control points of the feature curves. Users can also edit tones (skin colors) and hues (brightness) via simple slider control of PVG primitives. The vectorized facial images are reusable, so that users can copy and paste part or whole facial features to other raster image or vector graphics.
Related Research Topics
Orzan et al. pioneered diffusion curve images (DCIs), which are curves with colors defined on either side. By diffusing these colors over the image, the final result includes sharp boundaries along the curves with smoothly shaded regions between them.
Thanks to its compact nature and the ability of producing smoothly shaded images, diffusion curves quickly gain popularity in the graphics field and inspire many follow-up works, such as improving runtime performance and numerical stability and generalization to 3D and non-Euclidean domains.
However, diffusion curves allow users to only manipulate the boundary colors (i.e., the Dirichlet boundary condition of Laplace’s equation) and do not support control of color gradients; therefore it is difficult to generate photo-realistic images.