Bell A.J. & Sejnowski T.J., Vision Research 37: 3327-3338 (1997)  Home/Search   Document Not in Database   Summary   Related Articles   Check

....the experimental observation that cells in V1 are edge detectors. From this perspective, the expansion in eq. 12) should be generalized to include an orientational degree of freedom. Filters of this type have been proposed in [17] and found from an independent component analysis of natural images [18] . These studies should be combined with the use of overcomplete basis, this introduces redundancy but the representation becomes stable under small changes in the images [19] 7. ....

Bell A.J. & Sejnowski T.J., Vision Research 37: 3327-3338 (1997)

....loss of predictability, because probably the two points belong to two different objects, with almost independent luminosities. Since they cannot be predicted easily, it is frequently said that edges are the most informative structures in the image and that they represent their independent features [Bell and Sejnowski 1997]. Anyone s subjective experience is consistent with this: most people would start drawing a scene by tracing the lines of the obvious contours in the scene. Even more, the picture can be understood by anyone observing that sketch. Only afterwards, one starts adding texture to the light flat areas ....

....component contains highly informative pixels of the images that are responsible for the epigenetic development leading to the adaptation of receptive fields. Learning of the statistical regularities [Barlow 1961] present in this portion of visual scenes (using for instance the algorithm in [Bell and Sejnowski 1997]) would give relevant predictions about the structure of receptive fields of cells in the early visual system. Acknowledgments We are thankful to Jean Pierre Nadal, Germ an Mato and Dan Ruderman for useful comments and to Angel Nevado for many fruitful discussions that we had during the ....

Bell and Sejnowski, Vision Research 37, 33273338 (1997).

....to its definition, ffl l;r (x) l = h; v ) is the local linear edge variance along the direction l at scale r. Let us remark that edges are well known to be important in characterizing images. A recent numerical analysis suggests that natural images are composed of statistically independent edges [18]. We have analyzed the scaling properties of the local linear edge variances in a set of 45 images taken into a forest, of 256 Theta 256 pixels each (the images have been provided to us by D. Ruderman; see [16] for technical details concerning them) An analysis of the image resolution and of ....

Bell and Sejnowski, Vision Research 37 3327-3338 (1997).

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC