Rudiger von der Heydt PhD
Professor of Neuroscience
Professor of Neuroscience
Border-ownership coding in the visual cortex. The bar graph on the right shows the responses (mean firing rate) of an area V2 neuron to the visual stimuli shown on the left. Ellipses mark the conventional receptive field of the neuron. In each case, the receptive field is stimulated with the same local edge, but in a different image context. Note that the neuron responds more strongly if that edge perceptually belongs to a figure on the lower left. We find border-ownership selectivity in about 50% of the cells in V2. Thus, although neurons of V2 have tiny receptive fields and signal local contour features (orientation, color gradient), they also code the global figure-ground relationship. The elusive ‘gestalt principles’ of perceptual organization have their correlates in the neural mechanisms of image segmentation.
The visual sense, as much as the sense of touch, is a spatial sense. We see the world three-dimensional even though vision is based on just two-dimensional images, the optical projections on the retinae. In perception we can segregate figure and ground although objects often occlude one another so that foreground and background structures are cluttered up in the image. Perception decomposes a scene into objects and represents them in their proper dimensions in spite of varying image size and perspective distortion, and in their proper colors in spite of changing illumination. While the interpretation of images is technically very difficult, the visual system performs its task with great ease, so that we hardly become aware of it. The investigation of the neural processes underlying visual perception is the theme of research of this laboratory. The methods are micro-electrode recording from cells of the visual cortex in awake behaving monkeys, psychophysical experiments in humans and monkeys, and mathematical modeling. We are currently working on two main projects. One relates to the mechanisms of 3D form perception from binocular and monocular cues, the other to the coding of features as parts of objects. We use subjective perceptual phenomena as tools in relating the neural signals to the hypothetical representations and processes underlying perception. For example, the changing depth of an ambiguous display, such as the Necker cube, can be used to probe for the neural mechanisms of depth perception, the tendency in figure/ground displays to perceive contours as part of the figure (rather than the ground) can be used to probe for mechanisms of object-related feature representation.