Histogram of Radon transform with angle correlation matrix for distortion invariant shape descriptor
Résumé
A shape matching descriptor based on the histogram of Radon transform (HRT) with an angle correlation matrix is proposed. Our descriptor based on HRT is robust to shape rotation, scaling, and translation. Shape distortions provide sparse and dense distortions relative to the angle coordinate on our descriptor. Therefore, we compute an angle correlation matrix and apply the dynamic time warping for a non-linear angle matching to be robust to these transformations. Based on the beam search algorithm, we speed-up the time complexity of our method. Robustness to affine distortions for our approach is shown experimentally on different kinds of datasets and compared with the literature.