Visual analysis starts with the parallel extraction of different attributes at different spatial frequencies (SF) following a predominant and default coarse-to-fine processing sequence. Low spatial frequencies (LSF) would be processed faster than high spatial frequencies (HSF), allowing an initial coarse parsing of the visual input, before the analysis of the finer information. Studies investigating SF processing commonly use SF-filtered images as stimuli. Since LSF and HSF differ not only in terms of SF, but also in terms of luminance and contrast, recent studies normalized RMS contrast of images (i.e. the standard deviation of luminance values) in order to avoid any confound between SF components and known effects of luminance and contrast.In the present study, we investigated whether RMS contrast normalization of filtered scenes would induce bias in the default coarse-to-fine processing strategy during a categorization task (indoor vs. outdoor). We used dynamic sequences as stimuli composed of bandpass-filtered versions of a scene, presented from LSF to HSF or from HSF to LSF, allowing us to impose a coarse-to-fine (CtF) or a fine-to-coarse (FtC) processing of the scenes. In one condition (LUM), only the mean luminance of the filtered scenes was equated. In the other condition (RMS), both the mean luminance and the RMS contrast were equalized across the filtered scenes. In the LUM condition, CtF sequences were categorized faster than FtC sequences, consistent with a considerable number of studies. However, in the RMS condition, there was no advantage of CtF over FtC processing. The present study thus suggests that RMS contrast normalization of filtered scenes induces changes in the default SF processing strategies. We argue that such manipulation modify scene spectral properties that are exploited by the visual system to enable recognition. RMS contrast normalization should thus be used with caution when investigating SF processing of natural scenes.