Recent models of visual perception suggested that scene recognition is processed in terms of spatial frequencies: low spatial frequencies (LSF) rapidly reach high-order cortical areas to allow initial scene recognition and high spatial frequencies (HSF) subsequently carry fine details analysis. In the present MRI study we studied for the first time in children the relationship between the cortical thickness and behavioral performances to LSF and HSF processes. Sixteen children (10 years old) were presented with natural scenes and they had to indicate as rapidly and as accurately as possible for each trial whether the scene was outdoor or indoor. Each scene was filtered either in LSF or in HSF. Children were scanned with a 3-Tesla MRI scanner, and mean cortical thickness values were extracted with Freesurfer software according to the Destrieux Atlas. Regression analyses, which included age as covariable, were computed to determine the degree of relationship between the cortical thickness and behavioral performances to LSF and HSF filtered scenes. The results indicated that children exhibited (1) faster reaction times to LSFs associated to a decreased cortical thickness in higher order cortical areas constituted of middle frontal, lateral orbital and insula regions and (2) faster reaction times to HSFs associated to an increased cortical thickness in parietal and calcarine regions. These findings constitute the first structural study that supports the hypothesis that both synaptic pruning (i.e., decreased cortical thickness) and expansion mechanisms (i.e., increased cortical thickness) co-occur to allow healthy children to develop an efficient perception of the visual world.