Article: Digital art image design method based on fractal geometry and lightweight convolutional networks Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.40 pp.49 - 74 Abstract: This study proposes a digital art image design method based on fractal geometry and lightweight convolutional networks to overcome inefficiency and instability in existing artistic image generation. The approach integrates fern fractal and random tree mechanisms with pattern fusion, affine transformation, noise suppression, and hyperparameter optimisation to achieve diverse modelling of complex natural structures. Results show edge continuity of 97.9% and 96.3% at 1,024 px and 2,048 px, respectively - significantly higher than controls. With iteration depth 12, the artefact rate is 2.3%, and response linearity reaches 0.99. The latency is 205.4 ms, five times faster than the generative adversarial network, with an energy efficiency ratio of 3.2 samples/joule. These results demonstrate that the proposed method efficiently generates high-quality, structurally complex, and aesthetically natural images, offering reliable support for intelligent digital art design. Inderscience Publishers - linking academia, business and industry through research

Title: Digital art image design method based on fractal geometry and lightweight convolutional networks

Authors: Liang Yang; Yujun Liu

Addresses: Department of Literature and Art, Jinshan College of Fujian Agriculture and Forestry University, Fuzhou, 350001, China ' Computer and Information Engineering College, Guangdong Industry Polytechnic and Trade University, Guangzhou, 510510, China

Abstract: This study proposes a digital art image design method based on fractal geometry and lightweight convolutional networks to overcome inefficiency and instability in existing artistic image generation. The approach integrates fern fractal and random tree mechanisms with pattern fusion, affine transformation, noise suppression, and hyperparameter optimisation to achieve diverse modelling of complex natural structures. Results show edge continuity of 97.9% and 96.3% at 1,024 px and 2,048 px, respectively - significantly higher than controls. With iteration depth 12, the artefact rate is 2.3%, and response linearity reaches 0.99. The latency is 205.4 ms, five times faster than the generative adversarial network, with an energy efficiency ratio of 3.2 samples/joule. These results demonstrate that the proposed method efficiently generates high-quality, structurally complex, and aesthetically natural images, offering reliable support for intelligent digital art design.

Keywords: image art design; Fern fractal; random tree; hyperparameter gradient optimisation; lightweight convolutional neural network.

DOI: 10.1504/IJICT.2026.153316

International Journal of Information and Communication Technology, 2026 Vol.27 No.40, pp.49 - 74

Received: 30 Oct 2025
Accepted: 16 Jan 2026
Published online: 01 May 2026 *

Title: Digital art image design method based on fractal geometry and lightweight convolutional networks

Keep up-to-date