Improvement of small target detection based on tensorial filtering

摘要

Target detection is a key issue in processing hyperspectral images (HSIs). However, current spectral-identification-based algorithms are sensitive to noise during acquisition of the data. In most cases, the denoising algorithms cannot preserve small targets. In this paper, to overcome this problem, we propose a new algorithm which reduces noise to improve the target detection efficiency of HSI with small targets. First, a three-dimensional wavelet packet transform (3D-WPT) is used to decompose the HSI into several coefficient sets and models each coefficient set as a tensor. Then we exploit a powerful multilinear algebra model named parallel factor analysis (PARAFAC) to filter each tensor. The experiments conducted in both simulated and real-world hyperspectral images demonstrated the performance of the proposed method.