Histogram Equalization

Updated 02 Oct 2026

•
What it is
- •
  uses an image's cumulative intensity distribution to construct a mapping that spreads its occupied intensity levels
•
Mechanism
- •
  count intensities, calculate the cumulative distribution, construct a lookpu table, and replace each pixel using that table
- •
  a common discrete mapping for a nonconstnatn image is $T(k)=\operatorname{round}[(L-1)(F(k)-F_{\min})/(1-F_{\min})]$ , where $F_{\min}$ is the first nonzero cumulative probability.
•
Limitations
- •
  discrete output histograms are not guaranteed to become perfectly uniform.
- •
  can amplify noise or produce undesirable changes in brightness.
- •
  constant images require special handling because the normalization denominator vanishes.
•
Project
- •
  the equalization route decodes to grayscale and applies cv2.equalizeHist.
- •
  its displayed "original" is therefore the grayscale input to equalization
•
Distinguish from
- •
  Histogram Matching aims at a supplid reference distribution
- •
  Unavailable reference uses an intensity-range mapping rather than redistributing values using the full cumulative histogram
•
Reference
- •
  https://docs.opencv.org/4.x/d5/daf/tutorial_py_histogram_equalization.html
•

Linked references 4