Biomedical subjects
A Hurlbert
Publications and source records attributed to A Hurlbert.
Visual perception: Learning to see through noise.
New studies show that perceptual learning does not reduce the noise inherent to the neuronal mechanisms of perception. Rather, learning boosts the brain's capability to extract and make use of the relevant outside signal - but where and how the neuronal changes occur is still unknown.
Illusions and reality-checking on the small screen.
Explore the source record for details and available documents.
Interactions between colour and motion in image segmentation.
BACKGROUND: . An important early stage in visual processing is image segmentation, in which similar regions are grouped together and segregated from dissimilar regions, so that distinct objects ultimately may be located and recognized. In the natural world, objects are simultaneously characterized by colour, motion, texture and other visual attributes. How does the human visual system combine these attributes to segment the image? Although colour and motion information are conveyed by distinct functional streams from retina to visual cortex, there is increasing evidence for early and substantial cross-talk between the streams. Here, we explore psychophysical evidence for interactions between colour and motion in image segmentation. RESULTS: . Observers performed forced-choice segmentation tasks on random-dot stimuli. The dots in the vertical target figure were distinguished from the background dots by a different distribution of speeds or colours. To explore interactions between motion and colour segmentation, we added motion noise to the colour signal (or vice versa) by assigning all dots speeds (or colours) drawn from one of several noise distributions. Motion noise severely affects segmentation by colour. Motion noise defined by a broad distribution of speeds degrades colour segmentation, but a two-speed motion distribution (half moving up, half moving down) facilitates colour segmentation. Control experiments prove that the facilitatory effect is not caused by integrating colour information over different frames, nor can it be explained by probability summation over the two planes of moving dots. Colour noise also affects motion segmentation, but under a more restricted range of conditions, and not in a facilitatory way. CONCLUSIONS: . Colour and motion information interact at early stages during image segmentation, before decisions based on either cue in isolation are made. The robust bipolar effects of motion information on segmentation by colour indicate that the establishment of motion-defined surfaces takes primacy, and that such surfaces constitute important primitives for further processing.
Colour vision: putting it in context.
Our perception of a surface's colour depends in a complex way on the colours of surrounding surfaces; this "colour contrast' is linked to the "colour constancy' which is important for object recognition, but in ways more complex than we yet understand.
Visual perception. Deciphering the colour code.
Explore the source record for details and available documents.
Formal connections between lightness algorithms.
The computational problem underlying color vision is to recover the invariant surface-spectral-reflectance properties of an object. Lightness algorithms, which recover an approximation to surface reflectance in independent wavelength channels, have been proposed as one method to compute color. This paper clarifies and formalizes the lightness problem by proposing a new formulation of the intensity equation on which lightness algorithms are based and by identifying and discussing two basic subproblems of lightness and color computation: spatial decomposition and spectral normalization of the intensity signal. Several lightness algorithms are reviewed, and a new extension (the multiple-scales algorithm) of one of them is proposed. The main computational result is that each of the lightness algorithms may be derived from a single mathematical formula, under different conditions, which, in turn, imply limitations for the implementation of lightness algorithms by man or machine. In particular, the algorithms share certain limitations on their implementation that follow from the physical constraints imposed on the statement of the problem and the boundary conditions applied in its solution.
Colour vision: Is colour constancy real?
Colour constancy is typically weaker in the laboratory than it seems in our everyday experience. New measurements using real-world stimuli show that colour constancy is in fact almost perfect, and that several different perceptual mechanisms contribute to achieving it.
View-based models of 3D object recognition: invariance to imaging transformations.
This report describes the main features of a view-based model of object recognition. The model does not attempt to account for specific cortical structures; it tries to capture general properties to be expected in a biological architecture for object recognition. The basic module is a regularization network (RBF-like; see Poggio and Girosi, 1989; Poggio, 1990) in which each of the hidden units is broadly tuned to a specific view of the object to be recognized. The network output, which may be largely view independent, is first described in terms of some simple simulations. The following refinements and details of the basic module are then discussed: (1) some of the units may represent only components of views of the object--the optimal stimulus for the unit, its "center," is effectively a complex feature; (2) the units' properties are consistent with the usual description of cortical neurons as tuned to multidimensional optimal stimuli and may be realized in terms of plausible biophysical mechanisms; (3) in learning to recognize new objects, preexisting centers may be used and modified, but also new centers may be created incrementally so as to provide maximal view invariance; (4) modules are part of a hierarchical structure--the output of a network may be used as one of the inputs to another, in this way synthesizing increasingly complex features and templates; (5) in several recognition tasks, in particular at the basic level, a single center using view-invariant features may be sufficient.