Glossary · Machine vision
Image preprocessing
Also known as: Image enhancement, Image filtering
German: Bildvorverarbeitung
In image processing, image preprocessing is the set of operations applied to a raw image before analysis to improve it for the task, such as noise filtering, contrast adjustment, distortion correction, color conversion or morphological operations.
- Machine vision
In one sentence
Image preprocessing filters, corrects and converts raw images before analysis to make features easier and more reliable to evaluate.
Example
Preprocessing removes lens distortion and applies a median filter so the edge measurement on the stamped part becomes stable.
How it applies
- Engineering: Preprocessing compensates limitations of optics and lighting, but it also costs processing time and can remove real features. Better lighting often beats heavier filtering. Each additional step should be justified by a measurable improvement in the inspection result.
- Machine learning: For AI-based inspection, preprocessing at runtime must match the preprocessing used for the Training data; differences degrade the model silently. Typical steps include flat-field correction, noise filtering, contrast stretching and morphological opening or closing.
- Documentation: The documentation team should document the preprocessing chain in order with parameters, because each step influences later results and must be reproduced after software updates.
Image preprocessing vs. image analysis
Preprocessing transforms the image; analysis (for example Edge detection or Blob analysis) extracts measurements and decisions from it.