User Rating 0.0 โ˜…โ˜…โ˜…โ˜…โ˜…
Total Usage 0 times

Drop an image here or click to browse

Supports PNG, JPG, WebP, BMP, GIF • Max 4096×4096

Is this tool helpful?

Your feedback helps us improve.

โ˜… โ˜… โ˜… โ˜… โ˜…

About

Color desaturation reduces chromatic intensity in an image by interpolating each pixel between its original RGB value and a computed gray equivalent. The gray value depends on the method: the Luminosity method applies ITU-R BT.709 perceptual weights (L = 0.2126R + 0.7152G + 0.0722B), producing results that match human brightness perception. The Average method treats all channels equally. The Lightness method uses the midrange of channel extremes. Choosing the wrong method produces flat, muddy tones or crushed highlights. This tool processes pixels natively in the browser using the Canvas API with Web Worker acceleration.

Channel-selective desaturation isolates hue ranges (Reds, Yellows, Greens, Cyans, Blues, Magentas) and removes saturation only from selected families. This technique is standard in portrait retouching and product photography where skin tones or brand colors must remain vivid while the background loses chroma. The feathering algorithm applies a cosine falloff at hue boundaries to prevent hard color transitions. Note: results approximate professional software behavior but do not replicate ICC color-managed workflows.

desaturation grayscale image editing color removal photo editor saturation image converter

Formulas

The core desaturation operation is a linear interpolation between the original color and its computed gray value, controlled by an amount parameter a ranging from 0 (original) to 1 (fully desaturated).

Cout = Cin โ‹… (1 โˆ’ a) + gray โ‹… a

where Cout is the output channel value, Cin is the original channel value, a is the desaturation amount (0 - 1), and gray is the grayscale equivalent computed by the selected method.

graylum = 0.2126 โ‹… R + 0.7152 โ‹… G + 0.0722 โ‹… B
grayavg = R + G + B3
graylight = max(R,G,B) + min(R,G,B)2

For channel-selective desaturation, the hue angle H is computed from RGB. A mask weight w is derived using cosine feathering across the target hue range with a configurable feather width f (default 15ยฐ). The per-pixel desaturation amount becomes a โ‹… w.

Reference Data

MethodFormulaPerceptual AccuracySpeedBest Use Case
Luminosity (BT.709)0.2126R + 0.7152G + 0.0722BHighFastPhotography, print
Luminosity (BT.601)0.299R + 0.587G + 0.114BMedium-HighFastVideo, legacy systems
Average(R + G + B) รท 3LowFastestArtistic, flat look
Lightness (HSL)(max + min) รท 2MediumFastIllustration, design
Desaturate (HSL)S โ†’ 0MediumMediumPrecise hue control
Channel-selectivePer-hue S reductionHighSlowerPortrait retouching
Single channel (R)gray = RVery LowFastestForensic, infrared look
Single channel (G)gray = GMediumFastestLandscape, B&W film sim
Single channel (B)gray = BVery LowFastestDramatic sky effects
Decompose (Value)max(R,G,B)LowFastestBright, washed look
Luma (sRGB linear)Linearize โ†’ BT.709 โ†’ gammaHighestSlowColor-accurate archival
RMSโˆš(R2 + G2 + B2) รท 3MediumMediumHDR tone mapping

Frequently Asked Questions

The Luminosity method (BT.709) weights green at 0.7152, reflecting the human eye's peak sensitivity to green wavelengths. The Average method weights all three channels equally at 0.333. In practice, Average produces noticeably darker greens and brighter blues compared to perceived brightness. For photographic work, Luminosity preserves tonal relationships more accurately. Average is acceptable for stylistic or artistic effects where perceptual accuracy is not a priority.
The tool applies a cosine falloff function at hue range edges. For example, if you desaturate Reds (centered at 0ยฐ), pixels at exactly 30ยฐ (orange) receive partial desaturation rather than an abrupt cutoff. The feather width defaults to 15ยฐ. This prevents visible banding artifacts at the boundary between desaturated and saturated regions. The mask weight follows w = 0.5 โ‹… (1 + cos(ฯ€ โ‹… d รท f)) where d is the angular distance from the target hue center.
The preview renders at display resolution, which may differ from the source image resolution. The export always processes the full-resolution original. Additionally, browser color management may apply monitor ICC profiles to the canvas display but not to the exported PNG blob. If color accuracy is critical, calibrate your monitor and compare exports in a color-managed viewer.
Yes. Use channel-selective mode and enable desaturation on all hue families except the one you want to keep. For example, to isolate red objects: enable desaturation on Yellows, Greens, Cyans, Blues, and Magentas at 100%, leave Reds at 0%. The feathering ensures smooth transitions. For very specific hues (e.g., only fire-truck red, not skin red), the feather width may need manual reduction.
The tool processes images up to 4096 ร— 4096 pixels. Beyond that, the image is downscaled proportionally. For images within this limit, a Web Worker handles pixel processing off the main thread. Typical processing time for a 3000 ร— 2000 image is under 200ms on modern hardware. Older devices may take up to 1s.
No. The alpha channel is preserved unchanged. Only the R, G, and B channels are modified. Transparent PNG regions remain transparent after processing and export.