Abstract
We develop an algorithm to automatically color faces in a monochrome photograph. A face is one of the most important parts of human body and carries a large amount of expression. We use a 3D face model with reference colors and match the orientation of the model face to that of a face detected in a monochrome photograph to be colored. We make a 3D face Model with reference colors by mapping the texture to the model. This face model is painted with typical colors based on reference images from internet and image collection library. By transferring the model′ colors, the monochrome face is colored naturally. We computer similarities between the colored results from our system and the original color image before grayscale-converted. Human subjects are asked to evaluate the quality of colored images. This study will reduce a total time for coloring monochrome photographs, which is currently manually performed by experienced users.
