Pregled bibliografske jedinice broj: 229257
Neural network for perceptual grouping and lightness perception
Neural network for perceptual grouping and lightness perception // Twenty-second European Conference on Visual Perception - Abstracts / Gregory, Richard (ur.).
London : Delhi: Pion, 1999. (poster, međunarodna recenzija, sažetak, znanstveni)
CROSBI ID: 229257 Za ispravke kontaktirajte CROSBI podršku putem web obrasca
Naslov
Neural network for perceptual grouping and lightness perception
Autori
Domijan, Dražen
Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni
Izvornik
Twenty-second European Conference on Visual Perception - Abstracts
/ Gregory, Richard - London : Delhi : Pion, 1999
Skup
Twenty-second European Conference on Visual Perception
Mjesto i datum
Trst, Italija, 22.08.1999. - 26.08.1999
Vrsta sudjelovanja
Poster
Vrsta recenzije
Međunarodna recenzija
Ključne riječi
Lightness perception; Lightness illusions; Perceptual grouping; Neural networks; Boundary and surface computation
Sažetak
The Munker - White illusion, Benary's cross, checkerboard contrast, and second-order simultaneous contrast are examples of perceptual phenomena that could not be explained by classical concepts such as Wallach's ratio rule and centre - surround antagonism. Since such illusions involve two or more aligned borders with different magnitudes of contrast, it is proposed that a low-contrast contour receives stronger support if it is aligned with a high-contrast contour and therefore performs contrast negation in a filling-in layer. To implement this hypothesis the bipole cell model of perceptual grouping has been revised. The new model exhibits analog sensitivity and operates as a statistical MAX gate rather than an AND gate. Both features are consequences of presynaptic inhibition embedded in cooperative interactions. A MAX gate means that the bipole cell inherits activity level from the lobe that samples stronger contrast. Computer simulations have been performed with a neural network for lightness perception formulated in the tradition of filling-in theories. The network has four stages: (1) cells with centre - surround receptive fields ; (2) simple and complex cells ; (3) bipole cells ; (4) filling-in, which combines signals from stages 1 and 3. Activity distributions in the filling-in layer show that the model correctly predicts appearance of grey patches in all illusions mentioned here.
Izvorni jezik
Engleski
Znanstvena područja
Psihologija
