Jan Kotera

Blind Deconvolution Using Alternating Maximum a Posteriori Estimation with Heavy-Tailed Priors

Single image blind deconvolution aims to estimate the unknown blur from a single observed blurred image and recover the original sharp image. Such task is severely ill-posed and typical approaches involve some heuristic or other steps without clear mathematical explanation to arrive at an acceptable solution. We show that a straightforward maximum a posteriory estimation combined with very sparse priors and an efficient numerical method can produce results, which compete with much more complicated state-of-the-art methods.

Convolutional Neural Networks for Direct Text Deblurring.

In this work we address the problem of blind deconvolution and denoising. We focus on restoration of text documents and we show that this type of highly structured data can be successfully restored by a convolutional neural network. The networks are trained to reconstruct high-quality images directly from blurry inputs without assuming any specific blur and noise models. We demonstrate the performance of the convolutional networks on a large set of text documents and on a combination of realistic de-focus and camera shake blur kernels. On this artificial data, the convolutional networks significantly outperform existing blind deconvolution methods, including those optimized for text, in terms of image quality and OCR accuracy. In fact, the networks outperform even state-of-the-art non-blind methods for anything but the lowest noise levels. The approach is validated on real photos taken by various devices.

PIZZARO: Forensic analysis and restoration of image and video data.

This paper introduces a set of methods for image and video forensic analysis. They were designed to help to assess image and video credibility and origin and to restore and increase image quality by diminishing unwanted blur, noise, and other possible artifacts. The motivation came from the best practices used in the criminal investigation utilizing images and/or videos. The determination of the image source, the verification of the image content, and image restoration were identified as the most important issues of which automation can facilitate criminalists work. Novel theoretical results complemented with existing approaches (LCD re-capture detection and denoising) were implemented in the PIZZARO software tool, which consists of the image processing functionality as well as of reporting and archiving functions to ensure the repeatability of image analysis procedures and thus fulfills formal aspects of the image/video analysis work. Comparison of new proposed methods with the state of the art approaches is shown. Real use cases are presented, which illustrate the functionality of the developed methods and demonstrate their applicability in different situations. The use cases as well as the method design were solved in tight cooperation of scientists from the Institute of Criminalistics, National Drug Headquarters of the Criminal Police and Investigation Service of the Police of the Czech Republic, and image processing experts from the Czech Academy of Sciences.

Improving the blind restoration of retinal images by means of point-spread-function estimation assessment

Retinal images often suffer from blurring which hinders disease diagnosis and progression assessment. The restoration of the images is carried out by means of blind deconvolution, but the success of the restoration depends on the correct estimation of the point-spread-function (PSF) that blurred the image. The restoration can be space-invariant or space-variant. Because a retinal image has regions without texture or sharp edges, the blind PSF estimation may fail. In this paper we propose a strategy for the correct assessment of PSF estimation in retinal images for restoration by means of space-invariant or space-invariant blind deconvolution. Our method is based on a decomposition in Zernike coefficients of the estimated PSFs to identify valid PSFs. This significantly improves the quality of the image restoration revealed by the increased visibility of small details like small blood vessels and by the lack of restoration artifacts.

Blind Deconvolution With Model Discrepancies

Blind deconvolution is a strongly ill-posed problem comprising of simultaneous blur and image estimation. Recent advances in prior modeling and/or inference methodology led to methods that started to perform reasonably well in real cases. However, as we show here, they tend to fail if the convolution model is violated even in a small part of the image. Methods based on variational Bayesian inference play a prominent role. In this paper, we use this inference in combination with the same prior for noise, image, and blur that belongs to the family of independent non-identical Gaussian distributions, known as the automatic relevance determination prior. We identify several important properties of this prior useful in blind deconvolution, namely, enforcing non-negativity of the blur kernel, favoring sharp images over blurred ones, and most importantly, handling non-Gaussian noise, which, as we demonstrate, is common in real scenarios. The presented method handles discrepancies in the convolution model, and thus extends applicability of blind deconvolution to real scenarios, such as photos blurred by camera motion and incorrect focus.

The World of Fast Moving Objects

The notion of a Fast Moving Object (FMO), i.e. an object that moves over a distance exceeding its size within the exposure time, is introduced. FMOs may, and typically do, rotate with high angular speed. FMOs are very common in sports videos, but are not rare elsewhere. In a single frame, such objects are often barely visible and appear as semitransparent streaks. A method for the detection and tracking of FMOs is proposed. The method consists of three distinct algorithms, which form an efficient localization pipeline that operates successfully in a broad range of conditions. We show that it is possible to recover the appearance of the object and its axis of rotation, despite its blurred appearance. The proposed method is evaluated on a new annotated dataset. The results show that existing trackers are inadequate for the problem of FMO localization and a new approach is required. Two applications of localization, temporal superresolution and highlighting, are presented.

Understanding image priors in blind deconvolution

Removing blurs from a single degraded image without any knowledge of the blur kernel is an ill-posed blind deconvolution problem. Proper estimators together with correct image priors play a fundamental role in accurate blind de-convolution. We demonstrate a superior performance of the variational Bayesian estimator and discuss suitability of automatic relevance determination distributions as image priors. Restoration of real photos blurred by out-of-focus and motion blur, and comparison with a state-of-the-art method is provided.

Feature Selection on Affine Moment Invariants in Relation to Known Dependencies

Moment invariants are one of the techniques of feature extraction frequently used for pattern recognition algorithms. A moment is a projection of function into polynomial basis and an invariant is a function returning the same value for an input with and without particular class of degradation. Several techniques of moment invariant creation exist often generating over-complete set of invariants. Dependencies in these sets are commonly in a form of complicated polynomials, furthermore they can contain dependencies of higher orders. These theoretical dependencies are valid in the continuous domain but it is well known that in discrete cases are often invalidated by discretization. Therefore, it would be feasible to begin classification with such an over-complete set and adaptively find the pseudo-independent set of invariants by the means of feature selection techniques. This study focuses on testing of the influence of theoretical invariant dependencies in discrete pattern recognition applications.

PSF accuracy measure for evaluation of blur estimation algorithms

Given the large amount of blur estimation and blind deconvolution methods just in the last decade, there is an increasing need to compare the performance of a particular method with others. Unlike in other fields in image processing, there are very few well-established benchmark databases of test data and, more importantly, no standard way of performance evaluation. In this paper, we focus on the latter. We propose a new error measure for the blur kernel - a method for comparison of the blur estimate with the ground truth - which correctly reflects how inaccuracies in the blur estimation affect the subsequent image restoration, without the necessity to perform the actual deconvolution.

Blind deconvolution of images with model discrepancies using maximum a posteriori estimation with heavy-tailed priors

Single image blind deconvolution aims to estimate the unknown blur from a single observed blurred image and recover the original sharp image. Such task is severely ill-posed and typical approaches involve some heuristic or other steps without clear mathematical explanation to arrive at an acceptable solution. We show that a straight- forward maximum a posteriori estimation incorporating sparse priors and mechanism to deal with boundary artifacts, combined with an efficient numerical method can produce results which compete with or outperform much more complicated state-of-the-art methods. Our method is naturally extended to deal with overexposure in low-light photography, where linear blurring model is violated.