The natural environment around us, which is often crowded, cluttered or even foggy, is subject to a dynamically changing composition of objects and events. The human brain is continuously perceiving, recognizing and evaluating this dynamic scene composition. If the perception of degraded visual objects is important, e.g. in the case of potential threat stimuli, the brain needs to be more sensitive in detecting these objects from the natural environment. It is therefore hypothesized that reacting to the dynamically changing environment involves a robust and quick processing of salient information, which can be either with or without conscious awareness. We investigated the dynamics and robustness of perception using pictures of three salience levels, i.e. fearful faces (most salient), neutral faces (salient) and houses (nonsalient), which appear from dynamically decreasing random visual noise. Stimuli were matched for luminance, contrast, brightness and spatial frequency information. Reaction times show a significantly earlier response for faces than for houses. Fearful faces were significantly more quickly detected than neutral faces. The neural correlates sustaining robust perception were investigated with event-related functional magnetic resonance imaging (fMRI). The amygdala showed a significant perception-related response for faces, as compared to houses, that was further enhanced for fearful faces as compared to neutral faces. Our data indicate that emotionally salient information processing is (i) mediated by the amygdala and (ii) more robust than for nonsalient stimuli as it shows a significantly lower perceptual threshold.