Decomposition of stimuli into a set of low- and high-level features.

<p>For each category (<i>first</i> row, <i>green</i> background, <b>Natural</b>; <i>second</i> row, <i>red</i> background, <b>Fractal</b>; <i>third</i> row, <i>black</i> background, <b>Urban</b>; <i>last</i> row, <i>magenta</i> background; <b>Pink Noise</b>) a representative stimulus (left upper corner) and its associated low-level feature maps are shown. <b>LM</b>: Mean Luminance; <b>RGM</b>: Mean Red-Green Intensity; <b>YBM</b>: Mean Yellow-Blue Intensity; <b>SATM</b>: Mean Saturation; <b>LC</b>: Luminance Contrast; <b>RGB</b>: Red-Green Contrast; <b>YBC</b>: Yellow-Blue Contrast; <b>SATC</b>: Saturation Contrast; <b>TC</b>: Texture Contrast; <b>S</b>: Surfaceness; <b>C</b>: Cornerness; <b>E</b>: Edgeness; <b>BiSymm</b>: Bilateral Symmetry; <b>Ra-Symm L/H</b>: Radial symmetry with high or low spatial frequency selectivity. In addition to these low-level features, interestingness ratings were collected with the help of a pointer device (see Material and Methods), the topographic distribution of this high-level feature is shown as interestingness maps (<b>iMap</b>: Interestingness Map). Please note that this data is not collected for the case of Pink Noise category (lowest row). In addition to click data, recorded eye-movements for these four images are also presented in the same topographic form (<b>fMap</b>: Fixation Map; <i>second</i> row in each panel, <i>last</i> entry). All these maps were are shown following the histogram equalization step therefore all values occur equally likely.