Portrait Relief Modeling from a Single Image.

Area of Science:

• Computer Vision
• Computer Graphics
• Digital Imaging

Background:

• Existing image-based portrait relief methods are limited to 2.5D and struggle with 3D features like undercuts.
• Accurate geometry reconstruction, facial detail recovery, and depth structure preservation are key challenges in high-relief modeling.

Purpose of the Study:

• To propose a novel template-based framework for generating 3D portrait reliefs from single images.
• To address limitations of previous methods in creating high-relief models with undercuts.

Main Methods:

• A template-based framework utilizing bi-Laplacian mesh deformation for relief modeling.
• Integration of Shape-from-Shading (SFS) with user interaction for depth optimization.
• Fitting a template face to the input portrait and aligning facial features.

Main Results:

• Successful generation of 3D portrait reliefs with preserved depth structures and geometric details.
• The proposed method effectively handles complex features, including undercuts.
• Experimental results demonstrate superior performance compared to existing techniques.

Conclusions:

• The novel framework enables high-quality 3D portrait relief modeling from single images.
• The combination of template fitting, mesh deformation, and SFS offers a robust solution for relief generation.
• The method advances the state-of-the-art in 3D face reconstruction and relief modeling.