Abstract
This study evaluates six computational models for fusing texture and color to create realistic digital simulations of dyed fabrics. By comparing statistical mapping techniques with physics-based dichromatic reflection models, the researchers demonstrate that combining CIELCH color space transformations with texture strength adjustments yields the highest color fidelity, essential for the textile industry's digital design workflows.
Introduction
In the textile and apparel industries, the ability to visualize how a specific fabric texture, such as a twill weave, will appear when dyed a target color is crucial for pre-manufacturing design. The core challenge lies in "fusion"—the process of integrating color into a texture while maintaining realism. Simple overlays often fail to capture the complex physical interactions between light, fabric geometry, and dye, resulting in "flat" simulations. This paper investigates which mathematical models best preserve texture integrity while achieving high color fidelity.
Methodology
The researchers compared three primary families of models, encompassing six specific methods:
- Gray-to-Color Mapping (GCM): Maps 1D grayscale texture data to 3D RGB space using statistical correlations.
- Color-to-Color Mapping (CCM): A 3D-to-3D mapping that shifts an existing colored texture to a new target color. This was evaluated across three color spaces: CCM-RGB, CCM-LCH (Lightness, Chroma, Hue), and CCM-$l\alpha\beta$ (a decorrelated space).
- Dichromatic-based (DICH) Models: Physics-based models separating reflected light into surface reflection (highlights) and body reflection (base color). These include DICH-GC (grayscale source) and DICH-CC (colored source).
A significant innovation introduced is Texture Strength Adjustment, which scales the standard deviation of simulated color channels to match the expected texture intensity of the target color.
Key Insights
Channel Correlation
A fundamental finding is the extremely high correlation (often exceeding 0.98) between the Red, Green, and Blue channels in textile textures. This correlation serves as the mathematical foundation for mapping 1D grayscale information into a 3D color space effectively.
The Hue Stability Principle
The study observed that despite significant variations in texture and shadowing, the hue of a single-colored fabric remains remarkably constant across the image. This stability justifies the use of models that preserve hue while adjusting lightness and chroma.
Experimental Results
The models were evaluated using an Image Similarity Ratio based on CIELAB color differences ($\Delta E^*_{ab}$).
- Baseline Performance: Without manual adjustments, the physics-based DICH-CC method performed best, achieving a similarity ratio of approximately 0.87.
- Optimized Performance: With the application of Texture Strength Adjustment, the CCM-LCH model achieved the highest overall fidelity with a similarity of 0.95, followed closely by DICH-CC at 0.93.
- Color Space Observations: Surprisingly, the $l\alpha\beta$ space, typically effective for natural images, performed poorly. This suggests that the statistical properties of structured textile patterns differ significantly from those of natural scenes.
Conclusion
The research highlights that for high-fidelity fabric simulation, starting with a colored source and utilizing the CIELCH color space with texture strength adjustments is the most effective approach. While grayscale-to-color mapping provides acceptable results, it cannot match the realism of color-to-color techniques. Future work will aim to extend these models to multicolored prints and complex materials like silk or velvet, where surface reflections are non-neutral.
Summary updated