PDB: Point-Based Deformation Blending for Facial Animation Retargeting

2026-10-06

AI Summary

This study addresses the challenge of cross-mesh facial animation retargeting, where preserving expression fidelity while suppressing surface artifacts remains difficult. To this end, this work proposes a direct mesh reconstruction framework based on sparse control point prediction and weight reuse. By incorporating ReLU non-negativity constraints, row normalization, and per-vertex deformation blending, the method achieves efficient retargeting without predefined cages or global solvers. Furthermore, it supports cross-identity transfer with unpaired data using only self-retargeting supervision. The proposed approach significantly improves both retargeting accuracy and inference speed while effectively eliminating surface artifacts. Perceptual evaluations further validate its high fidelity and superior visual quality.

Abstract

Mesh-agnostic facial animation retargeting transfers expressions across meshes with different structures, but preserving facial motion without surface artifacts remains challenging. To address this, we present PDB, Point-Based Deformation Blending for facial animation retargeting. PDB predicts a compact set of deformed control points from a source neutral-expression pair and blending weights from the target neutral mesh. The weights are computed once per target and reused across frames, while the control points vary with each source expression. ReLU enforces non-negative weights and permits exact zeros, followed by row-wise normalization. The target mesh is reconstructed directly by multiplying the weights and control points, without a predefined cage, precomputed coordinates, a learned per-element deformation decoder, or a global reconstruction solve. Trained only with self-retargeting reconstruction supervision, PDB supports cross-identity transfer without paired cross-identity training expressions. Experiments demonstrate accurate retargeting, fast inference, and localized support in the learned weights. Joint evaluation of expression accuracy and local surface preservation shows reduced surface artifacts relative to the evaluated dense displacement method while retaining the intended motion. Perceptual evaluations further support expression fidelity and visual quality in both self- and cross-retargeting.

Problem

• facial animation retargeting
• mesh-agnostic
• surface artifacts
• expression transfer
• cross-identity

Innovation

• Facial Animation Retargeting
• Mesh-Agnostic
• Deformation Blending
• Control Points
• Cross-Identity Transfer