Abstract
Within natural images there is substantial spatial variation in both the local contrast and the local mean luminance. Understanding the statistics of these variations is important for understanding the dynamics of the receptive field stimulation that occur under natural viewing conditions. Methods — Local mean luminance and contrast were measured in a large set of calibrated 12-bit gray-scale natural images (van Hateren et al 1998), for a number of analysis-patch sizes. For each image, and each patch size, we measured the range of contrast, the range of mean luminance, the correlation in contrast (and in mean luminance) as a function of the distance between patches, and the correlation between contrast and mean luminance. In addition we hand segmented the images into regions containing different kinds of “stuff”—sky, ground, foliage, backlit foliage—and performed the same analyses for each kind of region. Results — Within the typical image, mean luminance varies by more than one log unit and rms contrast varies by more than 0.5 (e.g., 95% confidence interval = 0.05 − 0.55). Contrast tends to increase as the patch size increases, whereas mean luminance remains relatively constant. The correlation in contrast diminishes rapidly with distance (r-squared = 0.06 within 1.0–2.5 deg). The correlation in mean luminance diminishes more slowly (r-squared = 0.06 within 1.5–4.0 deg). The typical correlation between mean luminance and contrast is −0.2 (compared to −0.8 for 1/f noise). Discussion — The fact that there are large variations in local contrast and luminance within an image and that local contrast (and luminance) is uncorrelated at relatively small spatial separations has important consequences for neural gain control mechanisms. Eye movements are frequently large enough that there is little correlation in the contrast or luminance within a receptive field from one fixation to the next, and thus rapid components of contrast and luminance gain control are essential.