Independence of luminance and chromatic edge orientations in natural scenes
56.533, Tuesday, May 14, 2:45 - 6:45 pm, Vista Ballroom
Alik Mokeichev1,2, Ohad Ben-Shahar1,2; 1Department of Computer Science, Ben-Gurion University of the Negev, 2Zlotowski Center for Neuroscience, Ben-Gurion University of the Negev
Recent findings show that luminance and chromatic edge contrasts are statistically independent in natural scenes and it has been argued that a linear combination of these sources is optimal for a reliable detection of object boundaries [Hansen and Gegenfurtner 2009]. In a linear model, however, an underlying assumption is that the orientations of the luminance edges and of the chromatic edges always overlap. Here we examine the local joint statistics of chromatic and achromatic edge orientations in natural images and find a significant degree of statistical independence. Additionally, the statistical independence grows higher through the transformations of color representation from the retinal ganglion cells and LGN thalamic cells to the perceptually uniform Lab color space and its putative neural correlates. Our results provide an ecologically plausible explanation both for the non-linear color transformation from the cone-opponent space to the perceptual Lab color space as well as for recent findings of cells in the visual cortex that respond best to isoluminant-oriented patterns [Johnson et al. 2001, Bushnell et al. 2011].