On the estimation of optical flow: relations between different approaches and some new results
Artificial Intelligence
Integration of visual modules: an extension of the Marr paradigm
Integration of visual modules: an extension of the Marr paradigm
Computing perceptual organization in computer vision
Computing perceptual organization in computer vision
Use of monocular groupings and occlusion analysis in a hierarchical stereo system
Computer Vision and Image Understanding
Inferring global perceptual contours from local features
International Journal of Computer Vision - Special issue on computer vision research at the University of Southern California
IEEE Transactions on Signal Processing
Hi-index | 0.00 |
In this work we investigate the multi-modal statistics of natural image sequences looking at the modalities orientation, color, optic flow and contrast transition. It turns out that second order interdependencies of local line detectors can be related to the Gestalt law collinearity. Furthermore we can show that statistical interdependencies increase significantly when we look not at orientation only but also at other modalities.The occurrence of illusionary contour processing (in which the Gestalt law 'collinearity' is tightly involved) at a late stage during the development of the human visual system (see, e.g., [3]) makes it plausible that mechanisms involved in the processing of Gestalt laws depend on visual experience about the underlying structures in visual data. This also suggests a formalization of Gestalt laws in artificial systems depending on statistical measurements. We discuss the usage of statistical interdependencies measured in this work within an artificial visual systems and show first results.