Text-region detection
Find candidate text regions before character recognition.
Characters are pixels before a recognizer assigns text to them.
Construct channels or features suited to the chosen detector.
Score region candidates with the detector’s model.
Detection returns geometry. Recognition is a separate operation.
Group compatible candidates into text regions.
Illustrative example. The stages explain the method; they are not a live OpenCV execution.
Try it on an image
Experiment at pixel level
Detect text-like extremal regions with the bundled OpenCV NM classifiers and draw grouped text boxes. These are regions, not recognized words.
The engine loads on your first run. Your images stay in this browser.
Scroll over either image to zoom at the pointer. Use the scrollbars to pan both views over the same relative area. Zoom is relative to the input; pixel coordinates belong to each image. Warps can change scene correspondence.
Pixel inspector RGBA · native values · matched scale · 9 × 9 output pixels
Select a pixel
Select a pixel
When to use it
Locate scene text so recognition can work on focused, rectified crops.
How it works
- 01Construct channels or features suited to the chosen detector.
- 02Score region candidates with the detector’s model.
- 03Group compatible candidates into text regions.
candidate regions → text likelihood → grouped text boxes
What to tune
Load the detector’s required classifier assets and use its documented channel, scale and grouping settings.
Where it breaks down
Detection locates likely text; it does not decode the words. Classical ER filters and CNN detectors require different models and preprocessing.
TypeScript API
Open an entry for its exact overloads, parameter descriptions, result ownership and pinned upstream source.
All of these calls execute on the CPU. Native objects need explicit disposal. See matrices and ownership and build compatibility.