Digital Makeup Face Generation Wut Yee Oo Department of Mechanical Engineering, Stanford University
Introduction
Work flow of Make up Transfer Model
This project applies the make up from a reference image to a target by performing four main steps (1)Facial landmarks recognition (2)Facial alignment (3) Layer Decomposition (4)Transfer of face details, highlights and colors through each layer.
Detail Detail
Make up applications offer photoshop tools to get users’ inputs in generating a make up face. While this has been a well established area of development, automatic transfer of make-up from a reference to a target image presents more challenges. Lightness
Large Scale
Detail
Weighted Transfer a
Target
Facial extraction Color a
Color b b
Methodology Alpha Blending
Active Shape Model
Facial Alignment
Thin Plate Spline Warping
Layer Decomposition
RGB to CIE Lab space
Color a
Reference
Weighted least square
Make up Transfer
Large
Detail
Color a,b
Poisson Laplace editing
Weighted Transfer
Alpha blending
Large
Facial Landmark Recognition
Color b
Target with Make up
Alpha blend
Facial extraction Warping Color correction
Lightness
Large Scale Scale Large
Detail
Gradient Editing
Results and Limitations Reference
Target
Reference
Target
Reference
Target
References [1] S. Milborrow and F. Nicolls, "Locating Facial features with an Extended Active Shape Model," in EECV, 2008. [2] Z. e. a. Farbman, "Edge-preserving decompositions for multi-scale tone and detail manipulation," ACM Transactions on Graphics (TOG), vol. 27, no. 3, p. 67, 2008.
• Face contours, not due to make up, were transferred • Distortion of the reference image when warping
• Needs frontal views with no facial expression
Ideal reference image