Textile color has long been judged by a device that touches the fabric at a single point: the spectrophotometer. Accurate as it is, this instrument samples only a small aperture-sized patch of cloth, so it can miss the uneven dye distribution, mottling, and local shade shifts that determine how a finished garment actually looks. A new study published in Fashion and Textiles on 28 October 2026 presents a calibrated digital imaging framework that measures color across an entire fabric surface without contact, and validates it against conventional spectrophotometry on polyester fabrics dyed by two fundamentally different methods.
The research team, led by Yoonkyung Cho with colleagues including Ali Rayhan Sarkar, Shafira Fitri Baraja, and Ida Nuramdhani, built a custom stereo-photometric imaging chamber measuring 40 by 40 by 40 centimeters. Inside, a fabric sample rests on a white stage surrounded by matte black walls that suppress stray reflections. Four D65-equivalent LED light sources sit at azimuth angles of 0, 90, 180, and 270 degrees, and images are captured with only one lamp lit at a time, so each frame records the surface under a single, well-defined illumination direction. A Canon EOS 450D CCD camera mounted 22.5 centimeters above the stage captures images at 4272 by 2848 pixels, with illuminance held at 1844 plus or minus 14 lux throughout acquisition. A Datacolor SpyderCHECKR 24 color chart sits beside each sample for calibration.
Processing follows a pipeline the authors adapted from earlier stereo-photometric work. The four directional images are fused by pixel-wise averaging to remove directional illumination artifacts, and the device-dependent RGB values are transformed into the CIE XYZ color space using a least-squares linear matrix derived from the color chart. The calibrated data are then converted to CIE Lab* coordinates and on to the perceptual CIE LCh space of lightness, chroma, and hue angle. Four regions of interest per sample, selected manually at consistent size and excluding background and edges, yield mean values and standard deviations that describe both overall color and its spatial variation.
Three descriptors come out of this analysis. Color depth is expressed as a digital Kubelka-Munk K/S value, computed pixel by pixel from the Y luminance channel: the pixel’s relative reflectance is approximated as 0.01 times Y, and the standard Kubelka-Munk formula converts that into an absorption-to-scattering ratio, where higher values mean deeper, more light-absorbing color. Spatial uniformity is defined as 100 minus the coefficient of variation of the pixel-wise K/S values, so higher numbers indicate a more homogeneous dye distribution. Finally, the LCh coordinates capture lightness, chroma, and hue directly from the calibrated image.
To test whether these image-derived numbers mean anything, the team compared them with measurements from a Konica Minolta CM-600d spectrophotometer operated with a 2-degree standard observer, D65 illuminant, and 11-millimeter aperture, sampling four random spots per sample. The agreement for luminance was striking: image-derived and spectrophotometric Y values correlated at r = 0.99 with a coefficient of determination of 0.98. For K/S color strength the correlation remained strong at r = 0.95 and R2 = 0.90, slightly lower than for luminance. The authors attribute this partly to the Kubelka-Munk model itself, which behaves linearly only at low dye concentrations and deviates at higher color depths as scattering coefficients and absorption behavior change.
Agreement on absolute color coordinates was more modest. Comparing image-derived and spectrophotometric Lab* values across 36 samples, the mean CIEDE2000 color difference was 5.46, above the threshold commonly reported for perceptibility in direct color matching. The authors are careful about what this means. Camera-based and spectrophotometric systems differ in illumination geometry, detector spectral sensitivity, and spatial sampling, and here the two methods also measure different areas: the spectrophotometer reads a discrete aperture-defined spot while the image integrates pixel-level data over a larger fabric region. Correlations between the two methods’ coordinates were nonetheless high, at r = 0.98 for L*, 0.97 for a*, and 0.94 for b*, indicating that the imaging framework reliably preserves the relative color ordering of samples even when absolute values diverge.
The validation samples were deliberately diverse. The team dyed 100 percent polyester plain-weave fabrics with three disperse dyes, CI Disperse Red 1, CI Disperse Red 60, and CI Disperse Red 153, using both supercritical carbon dioxide and water as dyeing media, across temperatures of 80, 90, and 100 degrees Celsius and dyeing times of 30 to 60 minutes. In the supercritical CO2 experiments the pressure was fixed at 25 megapascals to keep fluid density stable, so temperature and time effects could be read cleanly. Supercritical CO2 dyeing is an emerging waterless coloration technology whose dye transport, governed by density-dependent solvent properties, enhanced diffusivity, and polymer swelling under heat and pressure, differs fundamentally from aqueous dyeing, making it a demanding test case for any color measurement method.
The image-based analysis distinguished these conditions quantitatively. In aqueous dyeing, color depth generally rose with temperature, most markedly for Disperse Red 1, while the supercritical system showed no consistent monotonic trend: each dye reached its maximum depth at a different temperature, at 90 degrees for R1, 80 to 90 degrees for R60, and 80 degrees for R153. Across time at 80 degrees, the supercritical medium generally produced deeper color than water, except for R153 at 50 minutes, and R1 achieved its deepest shade at the shortest time tested. On uniformity over a 40 by 40 square millimeter region, supercritical dyeing matched or exceeded aqueous dyeing in 13 of 18 conditions, and extending the analysis to areas up to 200 by 120 square millimeters revealed finer spatial variation that small-area spot measurements would miss.
The perceptual analysis added further detail. Aqueous-dyed fabrics generally showed higher lightness than their supercritical-dyed counterparts, while supercritical-dyed samples mostly showed higher chroma, consistently for R1 and R60 with one exception for R153. Hue angles clustered in the red region, roughly 350 to 20 degrees, for both media, but the supercritical samples clustered tightly while aqueous samples spread more broadly, with several around 80 degrees, a yellow-orange shift, appearing under the longest conditions of 80 degrees for 60 minutes for R60 and R153. Three-dimensional LCh visualizations made these contrasts directly comparable between dyeing media.
The authors position the framework as a complement to spectrophotometry rather than a replacement, noting that further work is needed to reduce the remaining absolute colorimetric discrepancies, to test the system under practical industrial imaging conditions, and to relate the measured parameters to human visual perception. Even so, the study demonstrates that a calibrated, multi-directional imaging setup can deliver non-contact, large-area, spatially resolved color data, spanning color depth, uniformity, lightness, chroma, and hue in a single measurement, and lays groundwork for computer vision and artificial intelligence driven quality inspection in textile manufacturing.
Subject of Research: A calibrated digital image-based method for quantitative color assessment of PET fabrics dyed in supercritical CO2 and aqueous media
Article Title: Development and validation of a digital image-based method for color assessment of PET fabrics dyed in supercritical CO2 and aqueous media
Article References: Cho, Y., Sarkar, A. R., Baraja, S. F., & Nuramdhani, I. (2026). Development and validation of a digital image-based method for color assessment of PET fabrics dyed in supercritical CO2 and aqueous media. Fashion and Textiles, 13(1), Article 45. https://doi.org/10.1186/s40691-026-00498-3
Image Credits: AI Generated
DOI: 10.1186/s40691-026-00498-3
Keywords: digital image-based color evaluation, PET fabrics, supercritical CO2 dyeing, aqueous dyeing, spectrophotometry, Kubelka-Munk color strength, CIE LCh color space, spatial uniformity, color calibration, stereo-photometric imaging, textile quality inspection, disperse dyes
News Source: Denise Maddox. (October 29, 2026). Research report: Development and validation of a digital image-based method for color assessment of PET fabrics dyed in supercritical CO2 and aqueous media. Scienmag.
