An Affine Invariant Patch Similarity

Abstract

Image and video comparison is often approached by comparing patches of visual information. In this work we present a detailed description and implementation of an affine invariant patch similarity measure that performs an appropriate patch comparison by automatically and intrinsically adapting the size and shape of the patches. We also describe the complete implementation of the proposed iterative algorithm for computation of those shape-adaptive patches around each point in the image domain. Source Code The implementation, written in C++, is available at the IPOL web page of this article 1. Compilation and usage instructions are included in the README.md file of the archive.

Exploring Patch Similarity in an Image

“…Note that one can go far beyond normalizing patches by translation, rotation, and affine maps on the intensity as we do here. In [29] for example, one finds the implementation of a method that extracts image patches normalized with respect to a geometric affine transformation. As such, it needs to define a sampling grid and interpolate the image to sample the patches, much like Algorithm 5.…”

Non-iterative Covariant Feature Extraction Based on the Shapes of Local Support Regions

“…And they achieved the affine shape adaptation by iteratively fitting an anisotropic Gaussian function to the blob features by means of a nonlinear least squares approach. Paper [18], [19] proposed an affine invariant similarity comparison between image patches from the point of view of Riemannian Manifolds. According to their derivation, the tensor product of the gradient vector can be used as the affine covariant structure tensors, which is the second moment matrix essentially.…”