Figure S9: Correlation with basic image properties

To quantify the extent to which the IFP responses could be explained by low-level characteristics of the images, we considered a list of 15 basic image properties: mean pixel grayscale value, standard deviation of the pixel grayscale values, median of the pixel grayscale values, minimum pixel grayscale value, maximum pixel grayscale value, number of pixels different from the background gray, number of pixels below the background gray, number of pixels above the background gray, number of very dark pixels (grayscale value < 64), number of very bright pixels (grayscale > 192), number of gray pixels, number of boxes above background gray (box size = 20 pixels), number of boxes below background gray, number of very dark boxes (mean intensity within the 20x20 pixel box < 64), number of very bright boxes (mean intensity within the 20x20 pixel box > 192). For each electrode (n=912), and each image property (n=15), we computed the Pearson correlation coefficient between the IFP responses to each image and the corresponding property for each image. We considered the property that yielded the maximum correlation (that is, the image property that accounts for most of the variance in the IFP signals for that electrode). Here we show the distribution of the Pearson correlation coefficients for all the electrodes. Bin size = 0.02, the thick bar shows the average.