A Deep Learning-Based Non-Photorealistic Rendering (NPR) Generation Method
Wei Zhou, Minnan Cang · 2024
With the advancement of deep learning technologies, non-photorealistic rendering (NPR) techniques have made significant breakthroughs in the field of computer graphics. This paper proposes a deep learning-based framework for NPR generation, combining convolutional neural networks (CNNs) and generative adversarial networks (GANs), aimed at automatically generating graphics in various artistic styles, such as sketches, ink paintings, and oil paintings. Through image style transfer and feature learning, the framework optimizes the preservation of details and enhances the artistic effects of the generated images. Experimental results demonstrate that the proposed method significantly outperforms traditional algorithms in terms of style transfer quality, generation efficiency, and adaptability. It shows strong potential for wide applications in animation, game design, and virtual reality, providing robust technical support for digital art creation.