Comparison of Dictionary-Based Image Reconstruction Algorithms for Inverse Problems

2020-10-07
Dogan, Didem
Öktem, Sevinç Figen
Many inverse problems in imaging involve measurements that are in the form of convolutions. Sparsity priors are widely exploited in their solutions for regularization as these problems are generally ill-posed. In this work, we develop image reconstruction methods for these inverse problems using patchbased and convolutional sparse models. The resulting regularized inverse problems are solved via the alternating direction method of multipliers (ADMM). The performance of the developed algorithms is investigated for an application in computational spectral imaging. Simulation results suggest that the convolutional sparse model provides similar reconstruction performance with the patch-based model; but the convolutional method is more advantageous in terms of computational cost.

Suggestions

Improved performance of Bayesian solutions for inverse Electrocardiography using multiple information sources
Serinağaoğlu Doğrusöz, Yeşim; MACLEOD, Robert (2006-10-01)
The usual goal in inverse electrocardiography (ECG) is to reconstruct cardiac electrical sources from body surface potentials and a mathematical model that relates the sources to the measurements. Due to attenuation and smoothing that occurs in the thorax, the inverse ECG problem is ill-posed and imposition of a priori constraints is needed to combat this ill-posedness. When the problem is posed in terms of reconstructing heart surface potentials, solutions have not yet achieved clinical utility; limitation...
DEEP LEARNING-BASED UNROLLED RECONSTRUCTION METHODS FOR COMPUTATIONAL IMAGING
Bezek, Can Deniz; Öktem, Sevinç Figen; Department of Electrical and Electronics Engineering (2021-9-08)
Computational imaging is the process of forming images from indirect measurements using computation. In this thesis, we develop deep learning-based unrolled reconstruction methods for various computational imaging modalities. Firstly, we develop two deep learning-based reconstruction methods for diffractive multi-spectral imaging. The first approach is based on plug-and-play regularization with deep denoisers whereas the second one is an end-to-end learned reconstruction based on unrolling. Secondly, we con...
Comparison of Kalman filter and Bayesian-MAP approaches in the spatio-temporal solution of the inverse electrocardiography Ters elektrokardiyografinin zaman-uzamsal çözümünde Kalman filtre ve Bayes-MAP yöntemlerinin karşilaştirilmasi
Aydin, Ümit; Serinağaoğlu Doğrusöz, Yeşim (2010-07-15)
In this study some of the spatial and spatio-temporal methods for the solution of the inverse problem of electrocardiography (ECG) are compared with each other. Comparisons are also made for the cases with geometric errors, where the location of the heart is shifted for 10mm and the size of the heart is reduced by 5%. The compared methods are the Kalman filter and Bayesian maximum a posteriori estimation (MAP). Two different Bayesian-MAP algorithms are used. While one uses only spatial information the other...
Singular inverse Sturm-Liouville problems with Hermite pseudospectral methods
Altundag, H.; Taşeli, Hasan (2021-10-01)
A numerical approximation to recover certain symmetric potentials in the singular inverse Sturm-Liouville problems over (-infinity,infinity) is presented. A Hermite pseudospectral method is employed to cope with the corresponding direct problem on the real line, which is encountered in the iterative procedure proposed for the inverse problem. The usual but unwelcome ill-posed structure of the resulting numerical algorithm has been treated to some extent by the help of a flexible optimization parameter and r...
Evaluation of Sparsity-based Methods for Parameterized Source Separation
Baskaya, Hasan Can; Öktem, Sevinç Figen (2020-10-07)
Parametric reconstruction problems arise in many areas such as array processing, wireless communication, source separation, and spectroscopy. In a parametric recovery problem, the unknown model parameters in each superimposed signal are estimated from noisy observations. Sparsity-based methods used in compressive sensing are also applied to parametric recovery problems. These methods discretize the parameter space to form a dictionary whose atoms correspond to candidate parameter values, represent the data ...
Citation Formats
D. Dogan and S. F. Öktem, “Comparison of Dictionary-Based Image Reconstruction Algorithms for Inverse Problems,” 28th Signal Processing and Communications Applications Conference (SIU) (2020), 2020, Accessed: 00, 2021. [Online]. Available: https://hdl.handle.net/11511/71585.