Hybridizing Expressive Rendering: Stroke-Based Rendering With Classic and Neural Methods.

Area of Science:

• Computer Graphics
• Artificial Intelligence

Background:

• Nonphotorealistic rendering (NPR) traditionally uses techniques like edge detection and toon shading for artistic effects.
• Stroke-based rendering is a key area within classical NPR, focusing on simulating artistic strokes.
• Deep learning has emerged as a significant advancement, offering new paradigms for NPR.

Purpose of the Study:

• To analyze and compare classical and neural network-based NPR techniques.
• To highlight the strengths and limitations of both approaches, especially in stroke-based rendering.
• To propose a framework for integrating classical and deep learning methods for novel expressive rendering.

Main Methods:

• Comparative analysis of classical NPR methods (edge detection, toon shading, geometric abstraction) and deep learning-based NPR.
• Focus on stroke-based rendering techniques within both paradigms.
• Exploration of quality and artistic control trade-offs.

Main Results:

• Classical NPR excels in established artistic styles and control.
• Deep learning-based NPR offers novel aesthetic possibilities but may require more data and computational resources.
• Both approaches have unique strengths and limitations in achieving expressive rendering.

Conclusions:

• A hybrid framework combining classical and deep learning NPR can unlock new creative potentials.
• Integrating these methods allows for a balance of artistic control and novel stylistic generation.
• Future research can explore synergistic combinations for advanced NPR applications.