| Abstract : | In lower-depth astronomy images (where $\lambda * t_{exposure}$ typically ranges from 0.1 to 10 or $\sigma$ \approx O($\lambda)$), conventional methods relying on 3-sigma clipping and median/mode estimation often fail to accurately determine the true background level. This failure can be attributed to the Poisson distribution characteristics of these images, where the sample mean provides a more reliable background estimate than the median or mode. Additionally, in these images, effectively excluding faint sources from the background estimation process is challenging, resulting in an overestimation of the background and obscuring the detection of very faint sources. To tackle this problem, we introduce an algorithm based on the "fellwalker" approach, designed to identify minima in the data. This algorithm effectively reduces the confusion between sources and background in the images, ensuring a better background estimation and preserving the detection of the maximum number of faint sources. |
