Abstract
The aim of this paper is to describe a novel non-parametric noise reductiontechnique from the point of view of Bayesian inference that may automaticallyimprove the signal-to-noise ratio of one- and two-dimensional data, such ase.g. astronomical images and spectra. The algorithm iteratively evaluatespossible smoothed versions of the data, the smooth models, obtaining anestimation of the underlying signal that is statistically compatible with thenoisy measurements. Iterations stop based on the evidence and the $\chi^2$statistic of the last smooth model, and we compute the expected value of thesignal as a weighted average of the whole set of smooth models. In this paper,we explain the mathematical formalism and numerical implementation of thealgorithm, and we evaluate its performance in terms of the peak signal to noiseratio, the structural similarity index, and the time payload, using a batteryof real astronomical observations. Our Fully Adaptive Bayesian Algorithm forData Analysis (FABADA) yields results that, without any parameter tuning, arecomparable to standard image processing algorithms whose parameters have beenoptimized based on the true signal to be recovered, something that isimpossible in a real application. State-of-the-art non-parametric methods, suchas BM3D, offer slightly better performance at high signal-to-noise ratio, whileour algorithm is significantly more accurate for extremely noisy data (higherthan $20-40\%$ relative errors, a situation of particular interest in the fieldof astronomy). In this range, the standard deviation of the residuals obtainedby our reconstruction may become more than an order of magnitude lower thanthat of the original measurements. The source code needed to reproduce all theresults presented in this report, including the implementation of the method,is publicly available at https://github.com/PabloMSanAla/fabada