Deep Sketch-Guided Cartoon Video Inbetweening.

Area of Science:

• Computer Vision
• Computer Graphics
• Artificial Intelligence

Background:

• Generating cartoon videos often requires complex animation processes.
• Existing frame interpolation methods struggle with large motion and user control.

Purpose of the Study:

• To develop a novel framework for cartoon video generation.
• To enable user control over video sequences through sketch guidance.
• To improve the quality and flexibility of cartoon video synthesis.

Main Methods:

• Estimating dense cross-domain correspondence between user sketches and cartoon frames.
• Employing a blending module with occlusion estimation for middle frame synthesis.
• Utilizing an arbitrary-time frame interpolation pipeline for in-between frame generation.
• Implementing a temporal consistency module for refined video output.

Main Results:

• The proposed framework successfully generates cartoon videos by synthesizing frames guided by user sketches.
• The system demonstrates superior performance compared to existing image synthesis and frame interpolation methods.
• Achieved high-quality results with enhanced temporal consistency and user controllability.

Conclusions:

• The novel framework offers a flexible and effective approach to cartoon video production.
• User-guided sketch-based synthesis significantly enhances video generation quality and control.
• The method generalizes well across different movie frames, outperforming current solutions.