Face photo-sketch portraits transformation via generation pipeline

Portrait sketching is widely used in digital art, forensic security and other fields with its unique value. However, existing portrait sketch style transfer techniques often focus on overall style transformation, neglecting the hierarchical drawing steps and the integration of details in the sketching process. The generated images still fall short in terms of layering effects and detail representation. To address this issue, this paper proposes a generative adversarial network-based portrait sketch temporal generation pipeline (PSTG). The pipeline simulates the artist’s layer-by-layer drawing process, sequentially executing from outline to facial features and then to hair, to generate high-quality sketch images with detailed expressiveness. Additionally, we designed a composite network structure that includes both global and local generators. The global generator is responsible for capturing overall contours and proportions, while the local generators focus on the detailed depiction of facial features and hair. This structure excels in capturing both overall proportions and local details. Experimental results demonstrate that the PSTG method not only restores the detail hierarchy of sketches but also achieves significant success in retaining sketch details and brushstroke effects. It also effectively imitates the brushstroke style of artists, generating portrait sketches that visually resemble the works of professional artists, outperforming existing holistic stylization methods overall.