Template matching with noisy patches: A contrast-invariant GLR test

TL;DR

This paper introduces a new contrast-invariant and noise-robust criterion for patch matching in noisy images, improving the reliability of dictionary-based image processing methods across various noise models.

Contribution

It proposes a novel GLR-based test for patch matching that is invariant to contrast changes and robust to different noise types, with theoretical analysis and empirical validation.

Findings

Effective under Gaussian, gamma, and Poisson noise

Ensures contrast invariance in noisy patch matching

Improves robustness and accuracy of image patch identification

Abstract

Matching patches from a noisy image to atoms in a dictionary of patches is a key ingredient to many techniques in image processing and computer vision. By representing with a single atom all patches that are identical up to a radiometric transformation, dictionary size can be kept small, thereby retaining good computational efficiency. Identification of the atom in best match with a given noisy patch then requires a contrast-invariant criterion. In the light of detection theory, we propose a new criterion that ensures contrast invariance and robustness to noise. We discuss its theoretical grounding and assess its performance under Gaussian, gamma and Poisson noises.