Article: Personalised colour harmony generation for clothing using a perception-guided diffusion model Journal: International Journal of Reasoning-based Intelligent Systems (IJRIS) 2026 Vol.18 No.20 pp.54 - 75 Abstract: Generating personalised colour combinations for apparel often suffers from colour perception mismatches, a lack of inter-regional coordination, and coarse luminance modelling, leading to significant deviations between generated results and user expectations. To address these challenges, this paper proposes a perception-guided diffusion model for personalised apparel colour generation. Specifically, the proposed method achieves colour perception alignment and explicitly models inter-regional coordination through perceptual space mapping and a graph convolutional network. Additionally, it incorporates a luminance-aware diffusion decoder to enhance the fidelity of highlight and shadow details. Furthermore, a controllable creative diffusion mechanism is introduced to enable flexible and highly personalised colour synthesis. Experimental results demonstrate that the proposed method achieves substantial improvements, yielding at least a 4.06 dB increase in peak signal-to-noise ratio and a 0.043 improvement in structural similarity index measure compared to state-of-the-art baselines, thereby providing a highly effective solution for intelligent fashion. Inderscience Publishers - linking academia, business and industry through research

Title: Personalised colour harmony generation for clothing using a perception-guided diffusion model

Authors: Xiaoli Lin; Shunhua Luo

Addresses: School of Textile and Fashion, Xinjiang Vocational University of Technology, Kashgar, 844004, China ' School of Fashion, Shandong University of Art & Design, Jinan, 250300, China

Abstract: Generating personalised colour combinations for apparel often suffers from colour perception mismatches, a lack of inter-regional coordination, and coarse luminance modelling, leading to significant deviations between generated results and user expectations. To address these challenges, this paper proposes a perception-guided diffusion model for personalised apparel colour generation. Specifically, the proposed method achieves colour perception alignment and explicitly models inter-regional coordination through perceptual space mapping and a graph convolutional network. Additionally, it incorporates a luminance-aware diffusion decoder to enhance the fidelity of highlight and shadow details. Furthermore, a controllable creative diffusion mechanism is introduced to enable flexible and highly personalised colour synthesis. Experimental results demonstrate that the proposed method achieves substantial improvements, yielding at least a 4.06 dB increase in peak signal-to-noise ratio and a 0.043 improvement in structural similarity index measure compared to state-of-the-art baselines, thereby providing a highly effective solution for intelligent fashion.

Keywords: perception-guided; diffusion model; colour combination generation; colour perception alignment; fashion design.

DOI: 10.1504/IJRIS.2026.155783

International Journal of Reasoning-based Intelligent Systems, 2026 Vol.18 No.20, pp.54 - 75

Received: 19 May 2026
Accepted: 25 Jun 2026
Published online: 13 Aug 2026 *

Title: Personalised colour harmony generation for clothing using a perception-guided diffusion model

Keep up-to-date