A self-similar stack model for human and machine vision
Biological Cybernetics
Representation of local geometry in the visual system
Biological Cybernetics
Histograms of Infinitesimal Neighbourhoods
Scale-Space '01 Proceedings of the Third International Conference on Scale-Space and Morphology in Computer Vision
Features in Scale Space: Progress on the 2D 2nd Order Jet
Scale-Space '01 Proceedings of the Third International Conference on Scale-Space and Morphology in Computer Vision
Image and Vision Computing
A color neuromorphic approach for motion estimation
WMVC'09 Proceedings of the 2009 international conference on Motion and video computing
Using representations of the dihedral groups in the design of early vision filters
ICASSP'93 Proceedings of the 1993 IEEE international conference on Acoustics, speech, and signal processing: image and multidimensional signal processing - Volume V
Image features and the 1-D, 2nd
Scale-Space'05 Proceedings of the 5th international conference on Scale Space and PDE Methods in Computer Vision
| Hi-index | 0.00 |
A local receptive field assembly, composed of receptive fields of limited structural complexity, cannot uniquely encode the retinal irradiance distribution (or ''picture'' for short). We study the structure of the equivalence classes (called ''images'') of pictures that elicit identical observations. We show that there are images that cannot be explained in terms of any picture at all and that certain singular images have a unique explanation, whereas most images could have been caused by any member of a large class of distinct pictures. We study the segmentation of observation space in terms of these basic possibilities. In those special cases in which the observations uniquely specify the picture, we proceed to find the exact structure of these special pictures. Every observation allows an interpretation in terms of such a special picture modulo an arbitrary attenuation. We refer to these equivalence classes of pictures as ''icons''. The space of icons can again be divided into regions of qualitatively distinct configurations; these are the possible ''local features.'' In machine vision and image processing the ''neighborhood operators'' play the same role as the receptive fields in vision. The analysis applies immediately to the representation of pictures via collections of neighborhood operators and may be regarded as a principled theory of ''features''.