Image preprocessing is the set of operations applied to raw image data before it is fed into a computer vision or machine learning model. It standardises and conditions images through resizing, normalisation, colour-space conversion, denoising and contrast adjustment to improve downstream model accuracy and robustness. Preprocessing also includes augmentation transforms that synthetically expand training data and feature-oriented steps that emphasise salient structures.
Overview
- Preprocessing transforms heterogeneous raw inputs into a consistent representation suitable for modelling.
- Normalisation and resizing ensure numerical stability and architectural compatibility.
- Augmentation improves generalisation by exposing models to plausible input variations.
Key aspects
- Resizing, cropping and aspect-ratio handling
- Pixel normalisation and standardisation
- Colour-space conversion (RGB, grayscale, HSV)
- Denoising and contrast enhancement
- Data augmentation (rotation, flip, jitter)
Applications
- Training pipelines for Image Classification
- Preparing inputs for Facial Recognition
- Medical and satellite image analysis
- Optical character recognition pipelines