Robust Edge Aware Descriptor (READ)

The Robust Edge Aware Descriptor (READ) measures the similarity of the underlying structure to an edge

using the 1D Fourier transform on a set of points located on a circle around a pixel. The magnitude and the phase of READ can well represent the magnitude and orientation of the local gradients and present robustness to imaging effects and artifacts.

To define READ, we first define LFD, by considering a neighboring function defined on a circle of radius R at each pixel and applying the 1D Fourier transform to the neighboring function:

$F(n)=\sum_{k=0}^{P}f_{k}e^{-2\pi i(n-1)(k-1)/P},(n=1,...,P)$

We noticed the second component of LFD can be used for local gradient estimation. Formally, READ is defined by setting n = 2 in the above equation:

Below see compare the magnitude and phase of READ (columns 4 and 5) with the magnitude and orientation of local gradient computed by the central difference method (column 2 and 3):

Gradient information computed by the READ operator are accumulated in support regions around the region found by an affine region detector (e.g., Hessian Affine) using intensity order pooling method to construct the READ image descriptor.

The support regions are computed by anisotropic scaling of the elliptical region (along their eigenvectors) rotation as we proved that the regions found by this method are corresponding to each other (yellow: the original region, red the new support region):

See the following reference for details:

Maani, Rouzbeh, Sanjay Kalra, and Yee-Hong Yang, Robust Edge Aware Descriptor for Image Matching, Proc. 12th Asian Conference on Computer Vision (ACCV), Nov 1-5, 2014, Singapore.