4D Facial Capture Pipeline Incorporating Progressive Retopology Approach

The pipeline for creating high-fidelity facial models often utilizes multi-view stereo techniques for reconstruction. However, the sub-sequent step of retopology often involves intricate manual work, limiting the extension of facial capture systems towards 4D acquisition. This paper proposes a facial 4D capture pipeline based on high-speed cameras. We employ standard multi-view stereo techniques for 3D reconstruction. Non-linear deformations of facial expressions are decoupled from rigid movements of the skull using QR code markers. Additionally, a progressive automated retopology approach is introduced for batch processing. Results demonstrate that our system can capture continuous facial motion sequences with detailed 3D models.