Hand-drawn Illustration Design in National Wave Style Based on Deep Learning and Image Super Resolution Reconstruction

Abstract

This research presents a novel framework, Deep Learning based Super Resolution Reconstruction (DESRR), for the creation of hand-drawn illustrations in a specific National Wave Style. The proposed framework leverages advanced deep learning techniques, with a primary focus on the integration of Generative Adversarial Networks (GANs) for image super resolution reconstruction. The objective is to enhance the resolution and fidelity of hand-drawn illustrations while preserving the distinctive characteristics of the chosen national wave style. The DESRR framework involves a two-step process: firstly, the utilization of GAN algorithms for generating illustrations that encapsulate the unique artistic nuances of the targeted national wave style; and secondly, the application of image super resolution techniques to refine and elevate the quality of the generated illustrations. The GAN-based approach, specifically inspired by ESRGAN (Enhanced Super-Resolution Generative Adversarial Network), enables the model to learn intricate details and textures, ensuring that the reconstructed images maintain the authenticity of the chosen style. To implement DESRR, a curated dataset of hand-drawn images in the specified national wave style is employed for training. The model is fine-tuned to strike a balance between increased resolution and the faithful representation of the targeted artistic style. The framework’s effectiveness is evaluated through a comprehensive analysis, considering both quantitative measures of image quality and qualitative assessments of style preservation. The proposed DESRR framework not only contributes to the field of artistic illustration design but also showcases the potential of combining deep learning and image super resolution techniques for creative applications.