| Autor: |
Babenko VV; Rostov State University, 344006 Rostov-on-Don, 105 B. Sadovaya St., Kuraev GA, Kul'ba SN |
| Jazyk: |
ruština |
| Zdroj: |
Rossiiskii fiziologicheskii zhurnal imeni I.M. Sechenova [Ross Fiziol Zh Im I M Sechenova] 2003 Oct; Vol. 89 (10), pp. 1300-9. |
| Abstrakt: |
The analysis of the visual segmentation problem involved questions concerned with a signal description as a result of linear filtering, nature following nonlinear transformations, and possible role of a spatial attention in pooling. Different approaches to the visual segmentation problem are compared. Our data suggest a model of visual segmentation. It uses the universal principle of linear filters pooling at second-order mechanisms. The pooling mechanism is characterized by position, spatial frequency and orientation tuning an is selective in respect to pattern size. The most activated pooling mechanisms are "targets" for selective attention. It is shown that the information selected this way is necessary and sufficient for recognition. Multiple representation of a visual scene with different resolution and hierarchical principle of description organise the recognition as successive specification. |
| Databáze: |
MEDLINE |
| Externí odkaz: |
|
