Neural population signals recorded with magneto- (MEG) or electroencephalography (EEG) contain information about viewed natural images and their categories. However, it remains unclear which specific visual features contribute to this information. To address this, we applied spectral manipulations and different texturization algorithms to natural images of various categories and recorded MEG in human subjects presented with these stimuli. The employed stimulus space allowed us to attribute components of visual information to specific low-, mid- and high-level visual features. Furthermore, this enabled us to investigate the effect of image manipulations on the geometry of the representational space. We found distinct contributions and time-courses of five visual features shaping neural representation: the retinotopy and spatial correlation structure of local visual features, spectral power as well as local and global shapes. These features differentially contributed to image and category information, with retinotopy and global shape as strongest features underlying image and category information, respectively. Furthermore, global shape preceded local shape information. Our results elucidate the contribution of different levels of image features to large-scale visual information in the human brain.