Plenary Lecture

Multidimensional Signal Processing Methods: Classification and Target Detection Methods Based on Tensor Decompositions

Professor Salah Bourennane
Aix Marseille University

Abstract: A hyperspectral image is a multidimensional array also named as a tensor and it normally consists of hundreds of spectral bands. So, HSI data, for instance,airborne hyperspectral images HYDICE (Hyperspectral Digital Imagery Collection Experiment), has two spatial dimensions and one spectral dimension. While acquired images in hyperspectral imagery are disturbed by additive noise, which can degrade classifcation and target detection results. To reduce the noise, HSI is commonly split into vectors or matrix so any 2D filtering method could be applied, but this splitting way does not consider the related information between image planes. So, some new approaches, such as tensor decomposition methods, have been used to denoise those images and showed some prospects in this field. There are two main decomposition models for multidimensional arrays: TUCKER3 (Three-mode factor analysis) decomposition and PARAFAC/CANDECOMP (Canonical Decomposition / Parallel Factor Analysis) decomposition. A multiway Wiener filter (MWF) is proposed to process a HSI as a whole entity based on TUCKER3 decomposition. In MWF, the filter in each mode is computed as a function of the filters in other modes, which reflects its capability in integrally utilizing the information in each mode of the multidimensional data. In practice, HSIs are always disturbed by hard-removed non-white noise, but this MWF method could not deal with the cases with non-white noise. So a pre-whitening procedure for HSIs to change the non-white noise to a white one is proposed in this paper. After that MWF can be used to filter the prewhitened result (PW-MWF). Then we can get the denoised images by an inverse processing of prewhitening. Though PW-MWF preserves the data structure of HSI, it also has some negative side effects in preserving small targets in the denoising process. In fact, PW-MWF is essentially an optimal low-pass filter while small targets are high frequency signals in Fourier basis, therefore PW-MWF might remove small targets in the denoising process. A multidimensional wavelet packet transform decomposes the prewhitened HSI into different coefficient tensors (components) by wavelet packet transform, and jointly filter each component by MWF. Since small target detection is an important issue in the HSI processing field, in this paper, PW-MWPT-MWF is proposed to reduce non-white noise in HSI with small targets and hence improve the target detection performances. The experiments of simulated and real-world images are given to present the performances of target detection after denoising by PW-MWPT-MWF.

Brief Biography of the Speaker: Salah Bourennane is currently a full Professor and he held also the position of the Dean of Research at the Ecole Centrale de Marseille, France. He is also the head of the Multidimensional Signals Group at Institut Fresnel, Marseille. He has over 30 years of research experience in the field of signal and image processing. His current research interests include statistical signal processing, array processing, image processing, remote sensing, tensor signal processing, and performance analysis. He authored over 350 research papers in various top-tier international journals and conferences, and edited many books and served as a guest editor of several special issues. He served on the editorial boards of many international journals and proceedings including the International Journal of Signal Processing, Image Processing and Pattern Recognition, The International Journal of Image and Signal Systems Engineering, Journal of Remote Sensing and Technology, among others. He has served on the technical program committees for numerous premier conferences and workshops including Advanced Concepts for Intelligent Vision Systems, International conference on latent variable analysis and signal separation, International Conference on Vision, Image and Signal Processing, and many others. He was an organizer of several international conferences such as the 6th European Workshop on Visual Information Processing at Marseille, 2016. He received a Ph.D. degree from Institut National Polytechnique de Grenoble, France, in signal processing.

Bulletin Board


The Conference Guide is online.

The paper submission deadline has expired. Please choose a future conference to submit your paper.

Plenary Speakers

WSEAS Main Site

Publication Ethics and Malpractice Statement