Research on digital reconstruction and image processing technology of watercolor style driven by big data
Data and Metadata · 13.05.2026 · 5 · 783
DOI: 10.56294/dm2026783 ↗Kurz gefasst
Fachartikel in der Zeitschrift Data and Metadata, online erschienen am 13.05.2026. Verfasst von Mei Li. Die Arbeit nennt 17 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Introduction
The increasing growth of digital art technologies and large-scale visual datasets has enabled new directions in image synthesis, particularly in watercolor style reconstruction.
Objective
This research focuses on the development of a big data-driven approach for digital reconstruction and image processing of watercolor-style images, aiming to preserve the artistic characteristics of transparency, fluidity, and brushstroke diffusion.
Method
A curated dataset of 11,100 high-resolution paintings is used, followed by preprocessing with noise removal to separate structural and stylistic features, followed by preprocessing step including noise removal to separate structural and stylistic components. To enhance feature representation, Histogram of Oriented Gradients (HOG)-based feature extraction is employed to capture edge structures, gradient orientations, and texture patterns that contribute to the overall composition and stroke direction in watercolor images. The proposed method integrates a Quokka swarm optimized Intelligent Residual Network 50 layers (QSO-IntResNet-50) architecture to capture both global composition and fine-grained artistic textures. IntResNet-50 learns coarse-to-fine watercolor features and captures global and fine details, while QSO optimizes parameters for better convergence. Perceptual and localized style loss improve realism, and a multi-receptive field discriminator enhances robustness across scales.
Results
Experimental results implemented in Python demonstrate that the proposed model achieves FID = 14,75, Diversity Score = 0,875, and IS = 8,70[user2.1][user2.2]. The results show that the proposed method successfully generates high-quality watercolor-style images while preserving both structure and artistic detail.
Conclusion
It demonstrates the effectiveness of integrating big data, Deep Learning (DL), and image processing for style reconstruction and digital art preservation.
Zitieren
APA
Li, M. (2026). Research on digital reconstruction and image processing technology of watercolor style driven by big data. Data and Metadata, 5, 783. https://doi.org/10.56294/dm2026783
Vancouver
Li M. Research on digital reconstruction and image processing technology of watercolor style driven by big data. Data and Metadata. 2026;5:783. doi:10.56294/dm2026783
BibTeX
@article{Li2026,
author = {Li, Mei},
title = {Research on digital reconstruction and image processing technology of watercolor style driven by big data},
journal = {Data and Metadata},
year = {2026},
volume = {5},
pages = {783},
doi = {10.56294/dm2026783}
}
EUVIMED