Marc Geese

PRNU and DSNU Maximum Likelihood Estimation Using Sensor Statistics

Abstract Image sensors come with a spatial inhomogeneity, known as Fixed Pattern Noise, that degrades the image quality. In this paper a known maximum likelihood estimation method [1] is extended in a way that it allows to estimate the two parameters DSNU and PRNU of a sensors fixed pattern noise. The methods input are the averaged sensor responses and the corresponding pairwise sensor covariances. First results show a significant performance increase compared to related methods. Zusammenfassung Bildsensoren besitzen eine räumliche Inhomogenität, auch als Fixed Pattern Noise bekannt, das die Bildqualität herabsetzt. In diesem Paper wird eine bekannte Maximum-Likelihood-Methode [1] erweitert, so dass eine kombinierte Schätzung der beiden Parameter DSNU und PRNU des Fixed Pattern Noise möglich ist. Die neue Methode benutzt die gemittelten Sensor-Antworten und die dazugehörigen paarweisen Sensor-Kovarianzen. Erste Ergebnisse zeigen eine signifikante Performancesteigerung gegenüber vergleichbaren Methoden.

CNN based dark signal non-uniformity estimation

Image sensors come with a spatial inhomogeneity, known as Fixed Pattern Noise, that degrades the image quality. Especially the dark signal non uniformity (DSNU) component of the FPN drifts with time and depends highly on temperature and exposure time. In this paper we introduce a cellular neural network (CNN) to estimate the DSNU from a given set of recorded images. Therefore the foundations of a previously presented maximum likelihood estimation method are used. A rigorous mathematical derivation exploits the available sensor statistics and uses only well motivated statistical models to calculate the CNNs synaptic weights. The advantages of the resulting CNN-method are continuous DSNU updates and a reduction of the computational complexity. Furthermore, a comparison based on ground truth correction patterns shows a significant performance increase to related methods.