Why do colored pixels appear around fine fabric threads in my RAW product photo?

Understand where the colored pixels originate

Many camera sensors record color through a color-filter array. Individual sensor positions do not directly provide complete red, green and blue information for every output pixel.

Demosaicing estimates the missing color components using neighboring samples. Very fine repeating structures, sharp edges and high-frequency detail can make that reconstruction difficult.

The darktable manual documents that demosaicing methods differ in how they handle fine details, colored edges, moiré and false maze patterns. Some methods preserve more detail but can introduce color overshoot in difficult regions.

This mechanism is distinct from a genuinely colored thread. It is also distinct from JPEG compression, which acts later on rendered image data.

Do not classify every colored edge as demosaicing failure. Chromatic aberration, sensor noise, lighting reflections and actual colored fibers can produce superficially similar observations.

Inspect the original RAW before resizing

Locate one small fabric region containing the suspicious colored pixels. Preserve the untouched RAW and record the current converter and processing settings.

Inspect the rendered image at an appropriate pixel scale. Avoid judging only a heavily enlarged screenshot or a browser thumbnail, since display scaling can introduce additional artifacts.

Compare the suspicious pixels with a reliable physical fabric reference or a separate close-up. If the textile genuinely contains colored threads, removing them would falsify the product.

If the colored pixels appear only in a resized storefront derivative, investigate the display-preparation stage instead. If they already appear in the RAW converter's full-resolution rendering, a RAW-stage reconstruction issue becomes more plausible.

Compare demosaicing methods without changing the source

Use a RAW editor that supports alternative demosaicing methods for the camera's sensor type. The current darktable manual documents several methods and their different artifact characteristics.

1. Keep one original RAW file and duplicate the processing treatment where supported.
2. Render the same fabric crop using method A.
3. Render it using a second supported method without changing exposure, white balance or sharpening.
4. Compare false color, thread definition and any new maze patterns.
5. Only then evaluate optional color smoothing or other artifact reduction.

If the colored speckles change substantially with the demosaicing method, reconstruction is implicated. If the colors remain attached to the same real threads across methods and independent references, investigate whether they are genuine.

Do not assume that the method with the smoothest appearance is most accurate. It may suppress real fine detail along with false color.

Illustrative diagnosis: a fine striped shirt

Hypothetical editorial scenario; no RAW conversion or garment test was performed. A shirt has narrow black and white woven stripes. Its physical reference contains no red decorative thread.

In illustrative conversion A, a RAW processor produces scattered red and cyan pixels along the narrow stripe boundaries. The pixels appear in the full-resolution rendering before storefront resizing.

Conversion B uses another supported demosaicing method with the same RAW source, white balance and exposure. The red and cyan speckles become less prominent while the black and white stripe spacing remains recognizable.

A stronger smoothing treatment removes additional speckles but also blurs the weave. That treatment is rejected.

The completed hypothetical diagnosis is demosaicing-related false color, supported by its sensitivity to the reconstruction method. Conversion B is selected as the better source treatment, subject to comparison with the actual shirt.

Keep RAW reconstruction separate from pattern accuracy

Record the converter version, sensor format, demosaicing method and any color-smoothing adjustment. These controls are application- and sensor-dependent; a Bayer-specific method should not be assumed available for every sensor arrangement.

Check the real pattern after correction. A conversion can remove false colors yet also reduce fine-thread definition. If the capture cannot resolve the fabric accurately, a different source photograph may be more useful than increasingly aggressive processing.

CartVisual does not provide the external RAW demosaicing controls described here. Use a reviewed source image and compare any generated garment against the real weave, stripe direction and seam relationships. A generated image cannot prove which RAW reconstruction was correct.

Record your own observations

The filled rows are illustrative examples. Use the blank row for your own files and observations.

Download the editable CSV worksheet

What the references establish

• darktable — Demosaic module: Primary software documentation describing color-filter-array reconstruction, alternative demosaicing methods, fine-detail behavior, false color and color smoothing. Controls and available algorithms depend on darktable version and sensor type.
• darktable 4.2 — Demosaic module: Version-specific historical documentation useful for distinguishing algorithm behavior across releases. It should not be used to assert current interface availability.