Range Image Segmentation Based on Differential Geometry: A Hybrid Approach
IEEE Transactions on Pattern Analysis and Machine Intelligence
An Experimental Comparison of Range Image Segmentation Algorithms
IEEE Transactions on Pattern Analysis and Machine Intelligence
Prior Learning and Gibbs Reaction-Diffusion
IEEE Transactions on Pattern Analysis and Machine Intelligence
Make3D: Learning 3D Scene Structure from a Single Still Image
IEEE Transactions on Pattern Analysis and Machine Intelligence
Learning independent causes in natural images explains the spacevariant oblique effect
DEVLRN '09 Proceedings of the 2009 IEEE 8th International Conference on Development and Learning
| Hi-index | 0.00 |
Natural vision systems still outperform artificial vision systems in terms of generalization. Therefore, many researchers turned to investigate biological vision systems in order to reverse engineer them and implement their principles into artificial vision systems. An important approach for developing a theory of vision is to characterize the visual environment in statistical terms, because this may provide objective yard sticks for evaluating natural vision systems using measures such as, for example, the information transmission rates achieved by natural vision systems. Most such studies focused on characterizing natural luminance images. Here we propose to investigate natural luminance images together with corresponding depth images using information-theoretical measures. We do this using a database of natural images and depth images and find that certain oriented filter responses convey more information about relevant depth features than other oriented filters. More specifically, we find that vertical filter responses are much more informative about gap and orientation discontinuities in the depth images than other filters. We show that this is an inherent property of the investigated visual scenes, and it may serve to explain parts of the oblique effects.