unwarping

Unwarping is the process of removing geometric distortions, or warps, from an image, video, or other signal to recover a representation that more accurately reflects the original scene or a standard projection. Distortions may arise from imperfect lenses, camera geometry, curved document surfaces, or motion during capture. The term is often used interchangeably with undistortion, dewarping, or image rectification.

In digital imaging, unwarping typically involves estimating a transformation that maps distorted coordinates to undistorted ones. This can be a global transform such as a homography for planar scenes, or a non-linear model for lens distortion, such as radial and tangential coefficients, polynomial warps, or thin-plate splines. Calibration data or feature-based estimates guide the correction, and the result is produced by resampling the image through the inverse warp.

Applications include correcting barrel or pincushion distortion in photographs, rectifying satellite or aerial imagery to geographic projections, flattening curved pages in scanned documents, and reconstructing undistorted views in microscopy, astronomy, or medical imaging. In non-static scenes, motion or depth changes may require optical-flow techniques or 3D reconstruction to inform a non-rigid unwarping.

Challenges include ensuring an accurate warp model, avoiding interpolation artifacts, handling occlusions, and preserving sharpness and photometric consistency. Real-time unwarping demands efficient algorithms and careful implementation. The concept is closely related to image rectification, perspective correction, and dewarping.

See also: undistortion, image rectification, perspective correction, dewarping.