Contrast Limited Adaptive Histogram Equalization (CLAHE): How It Works and When to Use It

The short answer

Contrast limited adaptive histogram equalization, usually shortened to CLAHE, is an image-processing method that increases local contrast. It splits the image into small regions and spreads out the brightness values within each region. It also caps how much any region can be stretched, so noise and flat areas are not exaggerated. The result brings out detail in shadows and highlights without lifting or darkening the whole frame evenly, as a single contrast slider would.

CLAHE runs as a processing step in software. It is not a standard webcam control. When your camera settings offer contrast, gamma or backlight compensation, those are different adjustments. The material supplied for this guide does not confirm that any particular webcam, driver or calling app offers CLAHE, so check your own software's documentation before assuming it is available.

Terms you need first

• Histogram: a count of how many pixels fall at each brightness level. A dim, flat image has its pixels bunched in a narrow range.
• Histogram equalization: remapping brightness so the pixels spread across the full range, which raises overall contrast.
• Adaptive: the remapping is calculated separately for small regions, called tiles, instead of once for the whole image.
• Contrast limited: a cap, called the clip limit, restricts how steeply any tile's brightness can be stretched.

How CLAHE works, step by step

• 1. Divide the image into a grid of tiles.
• 2. Build a brightness histogram for each tile.
• 3. Clip each histogram at the clip limit. Any count above the limit is cut off and redistributed evenly across all brightness levels. This stops a tile with a large uniform area, such as a plain wall, from being stretched so hard that sensor noise becomes visible texture.
• 4. Turn each clipped histogram into a mapping from old brightness to new brightness. This mapping comes from the cumulative distribution of the histogram.
• 5. Blend the mappings of neighboring tiles for each pixel, typically with bilinear interpolation. Without this blending, tile edges would show as visible blocks.
• 6. Write out the remapped brightness values.

How it differs from contrast and gamma settings

Adjustment | What it changes | Typical side effect

Contrast slider | Stretches or compresses all brightness values around a midpoint, the same way everywhere | Can clip bright highlights or crush dark shadows

Gamma | Bends the brightness curve to lighten or darken midtones globally | Can make the picture look washed out or muddy if pushed too far

Global histogram equalization | Spreads one histogram for the whole image | Can over-brighten large dark areas and look harsh

CLAHE | Equalizes small regions separately, with a cap on stretching | Can produce halos, a gritty or over-processed look, or amplified noise if set too strongly

Applying CLAHE to color images

Equalizing the red, green and blue channels separately tends to shift colors, which shows up especially in skin tones. The usual approach is to convert the image to a color space that separates brightness from color, such as LAB or YCrCb. You then apply CLAHE only to the brightness channel and convert back. This changes contrast while leaving hue largely intact.

The example below uses the OpenCV library in Python on a still image. The clip limit and tile grid size are starting values to adjust, not tested recommendations.

import cv2

img = cv2.imread("frame.png")
lab = cv2.cvtColor(img, cv2.COLOR_BGR2LAB)
l, a, b = cv2.split(lab)

clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))
l_eq = clahe.apply(l)

result = cv2.cvtColor(cv2.merge((l_eq, a, b)), cv2.COLOR_LAB2BGR)
cv2.imwrite("frame_clahe.png", result)

Choosing the two main settings

• Clip limit: a higher value allows stronger local contrast and also more noise and halos. A lower value gives a gentler effect that moves toward plain, unenhanced output.
• Tile grid size: more, smaller tiles respond to finer local detail but can look busy or patchy. Fewer, larger tiles behave more like global equalization.

How to check the result

• Compare the original and processed frames side by side at full size, not just as thumbnails.
• Inspect dark areas for noise that now looks like grain or texture.
• Look along high-contrast edges, such as a face against a bright window, for light or dark halos.
• Check plain surfaces like walls or backdrops for blotches or a tiled pattern.
• Check that skin tones have not changed color. If they have, make sure only the brightness channel was processed.
• If any of these problems appear, lower the clip limit first, then try a different tile size.

Limits and next steps for video calls

CLAHE improves how existing detail is displayed. It cannot recover detail the camera never captured. Fully black or fully white areas stay that way, and very noisy low-light footage usually gets noisier. Improving the light on your face, or reducing strong light behind you, is generally more effective than any contrast processing.

Applying CLAHE to a live call requires software that processes each frame before your calling app receives it. Whether that is possible depends on the tools you have. The supplied material does not confirm any specific app that does this. Keep in mind that a camera which is detected and streaming is separate from the image looking good on the other end. The calling app may apply its own compression and enhancement afterward, so judge the final appearance from what remote participants actually see.