Statistical signal processing & content-based information retrieval

Modern high-resolution sensing devices, with signal processing and communication capabilities largely based on the seminal Shannon and Nyquist studies, have enabled the acquisition, storage, and transmission of ever increasing amounts of data. To address the problem of handling and processing efficiently such a data deluge, SPL has given special focus on the design and implementation of powerful signal processing algorithms, which are based on the principles of compressive sensing (CS), in conjunction with multiresolution decompositions and accurate modelling of the underlying statistical behavior of the analyzed signals. Targeted applications include signal and image reconstruction from highly reduced sets of linear random projections, jointly sparse signal ensemble recovery in sensor networks, lightweight video processing, secure indoor localization, and signal classification.

Compressive sensing (CS) and its applications

  • Develop Bayesian CS approaches for signal and image reconstruction from a set of CS measurements, where the sparsity prior belief is enforced by means of heavy-tailed (alpha-stable) distributions.
  • Design Bayesian matching pursuit methods for reconstructing an ensemble of multiple signals, acquired by the nodes of a sensor network, in a distributed way by exploiting the joint sparsity structure among the signals of the ensemble.
  • Design and implement CS-based video compression architectures for remote sensing systems (UAVs and terrestrial-based sensor networks).
  • Develop robust algorithms for accurate and secure localization of mobile devices in indoor environments, by employing a highly reduced set of CS measurements generated from the received signal strength fingerprints.

Related Projects: PHySIS, CS-ORION.

Selected Publications:

Tsagkatakis, Grigorios; Woiselle, Arnaud; Tzagkarakis, George; Bousquet, Marc; Starck, Jean-Luc; Tsakalides, Panagiotis, Multireturn Compressed Gated Range Imaging, Optical Engineering, Vol. 54, No. 3, 031106, March 2015.

Tzagkarakis, George; Tsakalides, Panagiotis; Starck, Jean-Luc, Compressive Video Sensing with Adaptive Measurement Allocation for Improving MPEGx Performance, in Proc. International Conference on Computer Vision Theory and Applications (VISAPP ’15), Berlin, Germany, March 11-14, 2015.

Milioris, Dimitris; Tzagkarakis, George; Papakonstantinou, Artemis; Papadopouli, Maria; Tsakalides, Panagiotis, Low-dimensional Signal-Strength Fingerprint-based Positioning in Wireless LANs, Ad Hoc Networks, Vol. 12, pp. 100-114, 2014.

Tzagkarakis, George; Achim, Alin Marian; Tsakalides, Panagiotis; Starck, Jean-Luc, Joint Reconstruction of Compressively Sensed Ultrasound RF Echoes by Exploiting Temporal Correlations, in Proc. 2013 IEEE International Symposium on Biomedical Imaging (ISBI ’13), San Francisco, CA, USA, April 8-11, 2013

Tzagkarakis, George; Tsagkatakis, Grigorios; Starck, Jean-Luc; Tsakalides, Panagiotis, Compressive Video Classification in a Low-Dimensional Manifold with Learned Distance Metric, in Proc. 20th European Signal Processing Conference (EUSIPCO ’12), Bucharest, Romania, August 27-31, 2012.

Tzagkarakis, George; Tsakalides, Panagiotis, Bayesian Compressed Sensing Imaging using a Gaussian Scale Mixture, in Proc. 2010 International Conference on Acoustics, Speech, and Signal Processing (ICASSP ’10), Dallas, TX, March 14-19, 2010.

 

 

 Non-Gaussian modeling and multiscale Bayesian processing for various signal modalities 

  • Design statistical methods based on alpha-stable distributions to model the degree of sparsity in multiresolution decompositions of distinct signals and real world images (such as, medical, SAR, texture).
  • Develop novel techniques for inverse problems in underwater acoustics. The transform coefficients of underwater acoustic signals are modelled by using members of the alpha-stable family, whereas the task of classification is addressed by employing statistical similarity measures.

  • Design novel Bayesian algorithms exploiting the efficiency of alpha-stable models, and validate our processing philosophy to various application domains including:
    • image retrieval, fusion, and watermarking
    • SAR image denoising and autofocus
    • underwater acoustic signal classification
    • biomedical (ultrasound, microarray, miRNA) signal enhancement and classification

Related Projects: CS-ORION 

 

Selected Publications:

Achim, Alin Marian; Basarab, Adrian; Tzagkarakis, George; Tsakalides, Panagiotis; Kouame, Denis, Reconstruction of Ultrasound RF Echoes Modelled as Stable Random Variables, IEEE Trans. on Computational Imaging, Vol. 1, No. 2, pp. 86-95, June 2015.

Tzagkarakis, George; Beferull-Lozano, Baltasar; Tsakalides, Panagiotis, Rotation-Invariant Texture Retrieval via Signature Alignment based on Steerable sub-Gaussian Modeling, IEEE Trans. on Image Processing, Vol. 17, No. 7, pp. 1212-1225, June 2008.

Milioris, Dimitris; Tzagkarakis, George; Papakonstantinou, Artemis; Papadopouli, Maria; Tsakalides, Panagiotis, Low-dimensional Signal-Strength Fingerprint-based Positioning in Wireless LANs, Ad Hoc Networks, Vol. 12, pp. 100-114, 2014.

Tzagkarakis, George; Beferull-Lozano, Baltasar; Tsakalides, Panagiotis, Rotation-Invariant Texture Retrieval with Gaussianized Steerable Pyramids, IEEE Trans. on Image Processing, Vol. 15, No. 9, pp. 2702-2718, September 2006.

Taroudakis, Michael; Tzagkarakis, George; Tsakalides, Panagiotis, Classification of Shallow-Water Acoustic Signals via Alpha-Stable Modeling of the One-Dimensional Wavelet Coefficients, Journal of Acoustical Society of America (JASA), Vol. 119, No. 3, pp. 1396-1405, 2006