FabSoften: Face Beautification via Dynamic Skin Smoothing, Guided Feathering, and Texture Restoration

Sudha Velusamy, Rishubh Parihar, Raviprasad Kini, Aniket Rege; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2020, pp. 530-531

Face retouching is a widespread application in modern smartphone cameras with its high business value evidenced by its broad user base. We propose a real-time face softening approach that smooths blemishes in the facial skin region, followed by a wavelet band manipulation to restore the underlying skin texture, which produces a highly appealing `beautified' face that retains its natural appearance. Softening is carried out by an attribute-aware dynamic smoothing filter that is guided by facial attributes, including the number of blemishes and coarseness of facial skin texture. The proposed solution is robust to wide variations in lighting conditions, skin nonuniformities, blemishes, the presence of facial accessories, and delicate hair-like regions. The method includes an explicit facial hair preservation module to preserve their delicate texture while smoothing blemishes. We perform a qualitative comparison of our proposed face softening approach with numerous state-of-the-art techniques and commercial products. We demonstrate the power of our method in producing beautified faces at a minimal performance cost, which enables smooth execution on low-power devices like smartphones.