Demons registration for 2D empirical wavelet transforms

TL;DR

This paper introduces a numerically robust method for 2D empirical wavelet transforms using demons registration, improving computational stability and applying it to texture segmentation in microscopy images.

Contribution

It proposes a novel numerical scheme based on demons registration for efficient 2D empirical wavelet computation, addressing previous numerical challenges.

Findings

The method enhances numerical stability of 2D empirical wavelet transforms.

Application to microscopy images demonstrates effective texture segmentation.

The approach outperforms traditional methods in robustness and accuracy.

Abstract

The empirical wavelet transform is a fully adaptive time-scale representation that has been widely used in the last decade. Inspired by the empirical mode decomposition, it consists of filter banks based on harmonic mode supports. Recently, it has been generalized to build the filter banks from any generating function using mappings. In practice, the harmonic mode supports can have low constrained shape in 2D, leading to numerical difficulties to compute the mappings and therefore the related wavelet filters. This work aims to propose an efficient numerical scheme to compute empirical wavelet coefficients using the demons registration algorithm. Results show that the proposed approach gives a numerically robust wavelet transform. An application to texture segmentation of scanning tunnelling microscope images is also presented.