Interpatch

Interpatch is a term used in patch-based data processing to describe operations, measures, or constraints that involve interactions between multiple patches rather than within a single patch. It is not a formal, standardized term with a single definition; rather, it describes a class of methods across fields such as image processing, computer graphics, signal processing, and geostatistics that seek to exploit cross-patch information.

In image processing and computer vision, interpatch methods assess similarity and relationships across patches located at different positions. This enables non-local denoising, texture synthesis, and super-resolution by leveraging information from distant parts of the data. Algorithms like PatchMatch rely on interpatch comparisons to find corresponding patches throughout the image, and cross-patch regularization enforces consistency between neighboring patches to reduce seams.

In texture synthesis, terrain and 3D modeling, interpatch blending and coherence steps help ensure seamless transitions and globally consistent appearances when stitching patches together. In geostatistics and related disciplines, interpatch interpolation may refer to estimating values by considering relationships between spatial patches rather than treating each patch in isolation.

Terminology and emphasis vary by field, but the core idea is to move information between patches rather than solely within a single patch. Intra-patch processing remains important, yet interpatch constraints can improve realism, consistency, and global coherence when combining patches.

Related concepts include intrapatch processing, patch-based methods, non-local means, and PatchMatch.