A Caricature-Style Generating Method for Portrait Photo

The existing methods can not automatically generate realistic cartoon style line drawings which can reflect the facial features. In order to solve the problem, a method based on the hybrid method of data-driven and filtering is proposed. The facial features and hair are divided into two parts. To deal with the facial features, the face in a portrait image and the facial features are recognized. Then, an algorithm is designed that uses the existing key points to match the appropriate facial features data from the database built and generated the facial features image. For hair, Canny edge detection, image binarization, adaptive threshold binarization and XDoG operator are used to generate hair line domain. In order to generate a picture closer to the painting style, different painting textures and hair line fields are overlayed. By integrating facial features and hair pictures, it can generate carica-ture style line paintings of portraits. The images are obtained from the Internet to construct data sets for experi-ments. Using different methods to generate different styles of line drawings, professional and non-professional people are invited to evaluate the image quality, the accuracy of characteristics and the overall aesthetic effect. Experiments show that the method can deal with side-face pictures in real time and handle complex-structured hair while generating comic-style exaggerated facial features. Compared with other methods, the method can generate caricature style pictures instead of cartoon style or sketch style.