In this paper, we present a novel approach to the estimation of strongly varying backgrounds in astronomical images by means of small objects removal and subsequent missing pixels interpolation. The method is based on the analysis of a pixel local neighborhood and utilizes the morphological distance transform. In contrast to popular background estimation techniques, our algorithm allows for accurate extraction of complex structures, like galaxies or nebulae. Moreover, it does not require multiple tuning parameters, since it relies on physical properties of CCD image sensors - the gain and the read-out noise characteristics. The comparison with other widely used background estimators revealed higher accuracy of the proposed technique. The superiority of the novel method is especially significant for the most challenging fluctuating backgrounds. The size of filtered out objects is tunable, therefore the algorithm may eliminate a wide range of foreground structures, including the dark current impulses, cosmic rays or even entire galaxies in deep field images.

The impulsive noise in astronomical images originates from various sources. It develops as a result of thermal generation in pixels, collision of cosmic rays with image sensor or may be induced by high readout voltage in Electron Multiplying CCD (EMCCD). It is usually efficiently removed by employing the dark frames or by averaging several exposures. Unfortunately, there are some circumstances, when either the observed objects or positions of impulsive pixels evolve and therefore each obtained image has to be filtered independently. In this article we present an overview of impulsive noise filtering methods and compare their efficiency for the purpose of astronomical image enhancement. The employed set of noise templates consists of dark frames obtained from CCD and EMCCD cameras working on ground and in space. The experiments conducted on synthetic and real images, allowed for drawing numerous conclusions about the usefulness of several filtering methods for various: (1) widths of stellar profiles, (2) signal to noise ratios, (3) noise distributions and (4) applied imaging techniques. The results of presented evaluation are especially valuable for selection of the most efficient filtering schema in astronomical image processing pipelines.

Shift-and-add is an approach employed to mitigate the phenomenon of resolution degradation in images acquired through a turbulent medium. Using this technique, a large number of consecutive short exposures is registered below the coherence time of the atmosphere or other blurring medium. The acquired images are shifted to the position of the brightest speckle and stacked together to obtain high-resolution and high signal-to-noise frame. In this paper we present a highly efficient method for determination of frames shifts, even if in a single frame the object cannot be distinguished from the background noise. The technique utilizes our custom genetic algorithm, which iteratively evolves a set of image shifts. We used the maximal energy of stacked images as an objective function for shifts estimation and validate the efficiency of the method on simulated and real images of simple and complex sources. Obtained results confirmed, that our proposed method allows for the recovery of spatial distribution of objects even only 2% brighter than their background. The presented approach extends significantly current limits of image reconstruction with the use of shift-and-add method. The applications of our algorithm include both the optical and the infrared imaging. Our method may be also employed as a digital image stabilizer in extremely low light level conditions in professional and consumer applications.