High Relief from Brush Painting.

Area of Science:

• Computer Vision
• Digital Art
• Computational Geometry

Background:

• Relief art blends 2D painting and 3D sculpture.
• Generating 3D relief from 2D images is challenging.
• Existing methods struggle with accurate depth map recovery.

Purpose of the Study:

• To develop a novel approach for generating texture-mapped high-relief models from single brush paintings.
• To extract brushstrokes and generate corresponding relief proxies, not exact depth maps.
• To preserve the artistic style and features of the original painting in the 2.5D model.

Main Methods:

• Layer decomposition and stroke segmentation using boundary constraints to extract brushstrokes.
• Applying inflation and displacement maps to individual brushstroke proxies.
• Combining relief proxies to form a 2.5D high-relief model.

Main Results:

• Successfully generated convincing high-relief models from diverse paintings (humans, animals, flowers).
• Preserved the unique style and features of brushstrokes in the resulting relief models.
• Demonstrated the algorithm's effectiveness on purposeful brushstroke paintings like Chinese and Rosemailing art.

Conclusions:

• The proposed method effectively converts 2D brush paintings into 2.5D high-relief models.
• Brushstroke extraction and relief proxy generation maintain artistic integrity.
• The algorithm has potential secondary applications in image editing, particularly for stylized paintings.