Portrait stylized rendering for 3D light-field display based on radiation field and example guide.

Area of Science:

• Computer Vision
• Computer Graphics
• Display Technology

Background:

• Traditional 2D portrait stylization methods fail to meet the depth and spatial information requirements of 3D light-field displays.
• Generating high-quality stylized content for 3D light-field displays necessitates innovative approaches beyond 2D image processing.
• Existing methods struggle to maintain view consistency and contour details crucial for immersive 3D experiences.

Purpose of the Study:

• To propose a new portrait stylization method specifically designed for 3D light-field displays.
• To ensure the generation of 3D stylized portraits with accurate depth, spatial information, and view consistency.
• To overcome the limitations of 2D stylization techniques in creating content for advanced 3D displays.

Main Methods:

• An example-based portrait stylization method is employed to transfer style while preserving contour information.
• A pre-trained deep learning model incorporates a Laplacian loss function to minimize color diversity and refine contour details.
• A mask-guided light-field coding method is used to reconstruct the 3D representation and generate the final stylized 3D light-field image.

Main Results:

• The proposed method successfully generates 3D stylized portraits that maintain dense view consistency required for light-field displays.
• Contour information loss is prevented, and color diversity is minimized, leading to enhanced portrait quality.
• Experimental results validate the effectiveness of the approach in producing high-quality 3D light-field portrait content from real photos.

Conclusions:

• The developed portrait stylization method effectively addresses the challenges of creating content for 3D light-field displays.
• The technique enables the generation of high-quality, consistent, and detailed 3D stylized portraits suitable for immersive viewing.
• This advancement contributes to the creation of richer and more engaging visual content for emerging 3D display technologies.