arXiv subjects

Michael Zibulevsky

Publications and source records attributed to Michael Zibulevsky.

25 records · Page 2Linked to original sources

arXiv2016

Trainlets: Dictionary Learning in High Dimensions

Sparse representations has shown to be a very powerful model for real world signals, and has enabled the development of applications with notable performance. Combined with the ability to learn a dictionary from signal examples, sparsity-inspired algorithms are often achieving state-of-the-art results in a wide variety of tasks. Yet, these methods have traditionally been restricted to small dimensions mainly due to the computational constraints that the dictionary learning problem entails. In the context of image processing, this implies handling small image patches. In this work we show how to efficiently handle bigger dimensions and go beyond the small patches in sparsity-based signal and image processing methods. We build our approach based on a new cropped wavelet decomposition, which enables a multi-scale analysis with virtually no border effects. We then employ this as the base dictionary within a double sparsity model to enable the training of adaptive dictionaries. To cope with the increase of training data, while at the same time improving the training performance, we present an Online Sparse Dictionary Learning (OSDL) algorithm to train this model effectively, enabling it to handle millions of examples. This work shows that dictionary learning can be up-scaled to tackle a new level of signal dimensions, obtaining large adaptable atoms that we call trainlets.

cs.CV↗

arXiv2016

Patch-Ordering as a Regularization for Inverse Problems in Image Processing

Recent work in image processing suggests that operating on (overlapping) patches in an image may lead to state-of-the-art results. This has been demonstrated for a variety of problems including denoising, inpainting, deblurring, and super-resolution. The work reported in [1,2] takes an extra step forward by showing that ordering these patches to form an approximate shortest path can be leveraged for better processing. The core idea is to apply a simple filter on the resulting 1D smoothed signal obtained after the patch-permutation. This idea has been also explored in combination with a wavelet pyramid, leading eventually to a sophisticated and highly effective regularizer for inverse problems in imaging. In this work we further study the patch-permutation concept, and harness it to propose a new simple yet effective regularization for image restoration problems. Our approach builds on the classic Maximum A'posteriori probability (MAP), with a penalty function consisting of a regular log-likelihood term and a novel permutation-based regularization term. Using a plain 1D Laplacian, the proposed regularization forces robust smoothness (L1) on the permuted pixels. Since the permutation originates from patch-ordering, we propose to accumulate the smoothness terms over all the patches' pixels. Furthermore, we take into account the found distances between adjacent patches in the ordering, by weighting the Laplacian outcome. We demonstrate the proposed scheme on a diverse set of problems: (i) severe Poisson image denoising, (ii) Gaussian image denoising, (iii) image deblurring, and (iv) single image super-resolution. In all these cases, we use recent methods that handle these problems as initialization to our scheme. This is followed by an L-BFGS optimization of the above-described penalty function, leading to state-of-the-art results, and especially so for highly ill-posed cases.

cs.CV↗

arXiv2013

Speeding-Up Convergence via Sequential Subspace Optimization: Current State and Future Directions

This is an overview paper written in style of research proposal. In recent years we introduced a general framework for large-scale unconstrained optimization -- Sequential Subspace Optimization (SESOP) and demonstrated its usefulness for sparsity-based signal/image denoising, deconvolution, compressive sensing, computed tomography, diffraction imaging, support vector machines. We explored its combination with Parallel Coordinate Descent and Separable Surrogate Function methods, obtaining state of the art results in above-mentioned areas. There are several methods, that are faster than plain SESOP under specific conditions: Trust region Newton method - for problems with easily invertible Hessian matrix; Truncated Newton method - when fast multiplication by Hessian is available; Stochastic optimization methods - for problems with large stochastic-type data; Multigrid methods - for problems with nested multilevel structure. Each of these methods can be further improved by merge with SESOP. One can also accelerate Augmented Lagrangian method for constrained optimization problems and Alternating Direction Method of Multipliers for problems with separable objective function and non-separable constraints.

math.NA↗

arXiv2013

Spatially-Adaptive Reconstruction in Computed Tomography using Neural Networks

We propose a supervised machine learning approach for boosting existing signal and image recovery methods and demonstrate its efficacy on example of image reconstruction in computed tomography. Our technique is based on a local nonlinear fusion of several image estimates, all obtained by applying a chosen reconstruction algorithm with different values of its control parameters. Usually such output images have different bias/variance trade-off. The fusion of the images is performed by feed-forward neural network trained on a set of known examples. Numerical experiments show an improvement in reconstruction quality relatively to existing direct and iterative reconstruction methods.

cs.CV↗

arXiv2012

Phase retrieval combined with digital holography

We present a new method for real- and complex-valued image reconstruction from two intensity measurements made in the Fourier plane: the Fourier magnitude of the unknown image, and the intensity of the interference pattern arising from superimposition of the original signal with a reference beam. This approach can provide significant advantages in digital holography since it poses less stringent requirements on the reference beam. In particular, it does not require spatial separation between the sought signal and the reference beam. Moreover, the reference beam need not be known precisely, and in fact, may contain severe errors, without leading to a deterioration in the reconstruction quality. Numerical simulations are presented to demonstrate the speed and quality of reconstruction.

physics.optics↗

arXiv2012

Designing and using prior knowledge for phase retrieval

In this work we develop an algorithm for signal reconstruction from the magnitude of its Fourier transform in a situation where some (non-zero) parts of the sought signal are known. Although our method does not assume that the known part comprises the boundary of the sought signal, this is often the case in microscopy: a specimen is placed inside a known mask, which can be thought of as a known light source that surrounds the unknown signal. Therefore, in the past, several algorithms were suggested that solve the phase retrieval problem assuming known boundary values. Unlike our method, these methods do rely on the fact that the known part is on the boundary. Besides the reconstruction method we give an explanation of the phenomena observed in previous work: the reconstruction is much faster when there is more energy concentrated in the known part. Quite surprisingly, this can be explained using our previous results on phase retrieval with approximately known Fourier phase.

physics.optics↗

arXiv2010

Spatially-Adaptive Reconstruction in Computed Tomography Based on Statistical Learning

We propose a direct reconstruction algorithm for Computed Tomography, based on a local fusion of a few preliminary image estimates by means of a non-linear fusion rule. One such rule is based on a signal denoising technique which is spatially adaptive to the unknown local smoothness. Another, more powerful fusion rule, is based on a neural network trained off-line with a high-quality training set of images. Two types of linear reconstruction algorithms for the preliminary images are employed for two different reconstruction tasks. For an entire image reconstruction from full projection data, the proposed scheme uses a sequence of Filtered Back-Projection algorithms with a gradually growing cut-off frequency. To recover a Region Of Interest only from local projections, statistically-trained linear reconstruction algorithms are employed. Numerical experiments display the improvement in reconstruction quality when compared to linear reconstruction algorithms.

cs.CV↗