Generation of Chinese ink portraits by blending face photographs with Chinese ink paintings

Abstract

In this study, an interactive Chinese portrait rendering system was developed. This portrait rendering system can generate a user-lookalike ink portrait by blending the user’s face with a selected Chinese ink painting. It first automatically analyzes the user’s facial features and then integrates them into a selected Chinese painting. This system comprises two processes: an offline process and an online process. During the offline process, a collection of Chinese portrait paintings is configured (e.g., the face masks and facial coordinates of the paintings are determined). Subsequently, blending-ready templates (faces without facial features) are prepared for the online process. During the online process, the user integrates their photograph into our rendering system. The system automatically analyzes the face orientation, color, and facial features and adjusts the attributes of the photograph to match the template’s configuration. The produced facial image is blended into a selected template, which preserves the textures of the original Chinese painting. The results reveal that our system preserved both the user characteristics and original painting styles. In this study, user-portrait matching was experimentally evaluated, and a questionnaire survey on satisfaction with painting style was conducted.