Analysis results
Histogram, channel statistics, exposure diagnostics, thresholds, and comparison metrics.
Export results
Formula used
Bin frequency: hc(k) = Σ I(bk ≤ xc < bk+1)
Relative frequency: pc(k) = hc(k) ÷ N
Entropy: Hc = −Σ pc(k) log2 pc(k)
Mean: μc = (1 ÷ N) Σ xc,i
How to use
- Upload, paste, or load a sample image.
- Select a colour space and required channels.
- Choose bins, normalization, preprocessing, and threshold settings.
- Optionally draw a region, upload comparison data, or add a batch.
- Select Generate histogram to calculate and display results.
- Review statistics, warnings, similarity values, and export controls.
Example interpretation
| Observation | Possible meaning | Suggested action |
|---|---|---|
| Large spike near zero | Many shadow or black pixels | Check underexposure or intentional dark backgrounds |
| Large spike near maximum | Highlight clipping or bright background | Reduce exposure or inspect saturated regions |
| Narrow channel distributions | Low contrast | Consider contrast expansion or normalization |
| One channel shifted strongly | Possible colour cast | Review white balance and dataset consistency |
| Wide, balanced distributions | Broad tonal range | Usually suitable for feature extraction |
Machine learning applications
Colour histograms provide compact features for classification, retrieval, segmentation, and anomaly detection. They also expose exposure differences, clipping, colour casts, and domain shift.
Dataset-level averages can reveal class imbalance in visual appearance. Comparison metrics help measure preprocessing consistency across training, validation, and production images.
Frequently asked questions
It shows how frequently channel values occur within an image or selected region.
Use 256 bins for full 8-bit detail. Use fewer bins for smoother, more compact comparisons.
Different channels may contain similar frequencies at the same intensity values.
It divides bin counts by the analysed pixel count, enabling fair image-size comparisons.
The calculator measures the percentage of pixels near channel minimum and maximum values.
Higher entropy usually means greater tonal diversity and less predictable channel values.
Yes. Transparent pixels may be ignored, included, or composited over black or white.
It measures shared histogram mass. Larger values indicate more similar distributions.
No server upload is required. Analysis is performed locally with browser JavaScript.