Adaptive Weighted Variable Step Size Matching Tracking Algorithm for Dice Coefficient Matching

Aiming at the problem that the variable-step-length class matching tracking algorithm affects the reconstruction accuracy due to the insufficient selection of atoms during the iteration process of, this paper proposes an algorithm from the pre-selection of atoms with Dice coefficient matching and combines the secondary screening and variable-step-length principle. Firstly, the algorithm of adaptive selection of atoms by Dice coefficient matching is used to improve the correlation between the current residuals and the selected atoms by adding the angular characteristics of the signals while retaining the length characteristics of the signals. At the same time, for the presence of certain atoms in the atoms entering the iteration due to the primary selection in the reconstruction process that can still provide a contribution to the original signal, a secondary selection of atoms is carried out using an adaptive weighting strategy, and then the atoms that have already been selected into the support set are screened out using the projection backtracking principle. The balance between reconstruction accuracy and reconstruction time is achieved by adaptively selecting the step size at different stages of the algorithm operation through the noise value and the signal value estimated at each iteration. The reconstruction experiments on one-dimensional signals and two-dimensional images show that the algorithm can effectively improve the reconstruction effect on signals and images.