Lighting difference based wrinkle mapping for expression synthesis

Nowadays, facial expression synthesis is widely used in expression simulation, recognition and animation. Texture feature (wrinkle) after face deformation largely reflects the genuineness of the synthesized expression, whose mapping is a critical step to the expression synthesis application. However, current lighting ratio based wrinkle mapping is sensitive to small source lighting and large lighting difference. In this work, a novel wrinkle mapping based on lighting difference fitting is proposed to improve the genuineness and robustness of mapped wrinkles. Experimental results on expression synthesis with large lighting difference show that the proposed algorithm is competitive on synthesizing visually genuine expressions.