Virtual Makeup is a real-time, fully automatic application for women enabling them to try multiple makeup options virtually using just a computer and a webcam.
The architecture can be divided into two sequential parts: automatic facial feature tracking and coloring. For facial feature tracking, multiple well established algorithms were reviewed and the comparison led to Active Shape Models (ASMs) for being fast (covering the real time demand) and accurate. An ASM comprised of 115 feature-points was selected. The model was trained exhaustively using a manually annotated database of 12300 faces and a test database of 135 female faces. The trained model (derived from 135 faces) provided sufficient fitting in a wide range of faces, facial expressions and poses.
The facial features delivered by the ASM were used for coloring. In order to produce a realistic result and support a full range of colors the Hue Saturation Value (HSV) cylindrical representation was selected. Furthermore, Gaussian filters and Spline functions were used to provide smoother edges. A result of the Virtual Makeup application can be seen in the following image
An advanced framework for immersive media capturing, representation, encoding and semi‐automated collaborative content production.