While different optical flow techniques continue to appear, there has been a lack of quantitative evaluation of existing methods. For a common set of real and synthetic image sequences, we report the results of a number of regularly cited optical flow techniques, including instances of differential, matching, energy-based and phase-based methods. Our comparisons are primarily empirical, and concentrate on the accuracy, reliability and density of the velocity measurements; they show that performance can differ significantly among the techniques we implemented.

We have implemented a model of contrast gain control in human vision which incorporates a number of key features, including a contrast sensitivity function, multiple oriented band-pass channels, accelerating nonlinearities, and a divisive inhibitory gain-control pool. The parameters of this model have been optimized through a fit to the recent data that describe masking of a Gabor function by cosine and Gabor masks [Foley, J. M. (1994). Human luminance pattern mechanisms: masking experiments require a new model. Journal of the Optical Society of America A 11(6), 1710-1719]. The model achieves a good fit to the data. We also demonstrate how the concept of recruitment may accommodate a variant of this model in which excitatory and inhibitory paths share a common accelerating non-linearity, but which include multiple channels tuned to different levels of contrast [Teo, P. C. & Heeger, D. J. (1994). Perceptual image distortion. Proceedings,

This thesis describes some new approaches to the representation and analysis of visual motion, as perceived by a biological or machine visual system. We begin by discussing the computation of image motion fields, the projection of motion in the three-dimensional world onto the two-dimensional image plane. This computation is notoriously difficult, and there are a wide variety of approaches that have been developed for use in image processing, machine vision, and biological modeling. We show that a large number of the basic techniques are quite similar in nature, differing primarily in conceptual motivation, and that they each fail to handle a set of situations that occur commonly in natural scenery. The central theme of the thesis is that the failure of these algorithms is due primarily to the use of vector fields as a representation for visual motion. We argue that the translational vector field representation is inherently impoverished and error-prone. Furthermore, there is evidence that a ...

ii About the cover The figure on the cover shows a visualization of a symmetric tensor in three dimensions, G = λ1ê1ê T 1 + λ2ê2ê T 2 + λ3ê3ê T 3 The object in the figure is the sum of a spear, a plate and a sphere. The spear describes the principal direction of the tensor λ1ê1ê T 1, where the length is proportional to the largest eigenvalue, λ1. The plate describes the plane spanned by the eigenvectors corresponding to the two largest eigenvalues, λ2(ê1ê T 1 + ê2ê T 2). The sphere, with a radius proportional to the smallest eigenvalue, shows how isotropic the tensor is, λ3(ê1ê T 1 + ê2ê T 2 + ê3ê T 3). The visualization is done using AVS [WWW94]. I am very grateful to Johan Wiklund for implementing the tensor viewer module used. This thesis deals with filtering of multidimensional signals. A large part of the thesis is devoted to a novel filtering method termed “Normalized convolution”. The method performs local expansion of a signal in a chosen filter basis which

this article relate most directly to a specific hypothesis on the nature of this representation - the so-called `ripple analysis model' (Shamma and Versnel 1995; Shamma et al. 1995; Versnel et al. 1995). Briefly, the model postulates that the acoustic spectrum is encoded in AI at varying degrees of resolution by the activity of units with a range of response area bandwidths, asymmetries and best frequencies (BF's). Furthermore, it is assumed that this multi-scale decomposition can be characterized to a very good approximation as a linear process. Thus, if a complex spectral profile is decomposed into a weighted sum of simpler spectra, then linearity implies that responses to the complex profile can be predicted from a weighted superposition of the responses to the simpler spectra. Note further, that if the basic set of simple spectra is taken to be sinusoidally modulated envelopes or ripples, then the decomposition of an arbitrary profile into ripples with different amplitudes, phases, and densities corresponds simply to a Fourier decomposition of the spectral profile. J0856 5 (1 of 2) 3 The above postulates were extensively investigated and validated for stationary spectra in the ferret AI (Shamma et al. 1995; Shamma and Versnel 1995; Versnel et al. 1995; in cat: Schreiner and Calhoun 1995). For instance, it was shown that an AI unit could be fully characterized by its responses to ripples with a range of ripple frequencies and ripple phases, that is by its ripple transfer function. It was also shown that inverse Fourier transforming this function generates a response field (RF) - a function that is analogous to the response area of the unit obtained with single tones. The RF's of AI units exhibited a range of bandwidths and asymmetries, as required by the multi-scal...

... This article presents one alternative to the computer metaphor suggesting that brains are organized into independent modules. Evidence is reviewed that brains are organized into parallel processing streams with complementary properties. Hierarchical interactions within each stream and parallel interactions between streams create coherent behavioral representations that overcome the complementary deficiencies of each stream and support unitary conscious experiences. This perspective suggests how brain design reflects the organization of the physical world with which brains interact. Examples from perception, learning, cognition, and action are described, and theoretical concepts and mechanisms by which complementarity is accomplished are presented.

Many models for stereo disparity computation have been proposed, but few can be said to be truly biological. There is also a rich literature devoted to physiological studies of stereopsis. Cells sensitive to binocular disparity have been found in the visual cortex, but it is not clear whether these cells could be used to compute disparity maps from stereograms. Here we propose a model for biological stereo vision based on known receptive field profiles of binocular cells in the visual cortex and provide the first demonstration that these cells could effectively solve random dot stereograms. Our model also allows a natural integration of stereo vision and motion detection. This may help explain the existence of units tuned to both disparity and motion in the visual cortex. 1 Introduction It is well known that binocular disparity forms the basis of stereoscopic depth perception. There have been many physiological investigations on the mechanisms of stereopsis (see Freeman and Ohzawa, 19...

Abstract not found

We previously proposed a physiologically realistic model for stereo vision based on the quantitative binocular receptive field profiles mapped by Freeman and coworkers. Here we present several new results about the model that shed light on the physiological processes involved in disparity computation. First, we show that our model can be extended to a much more general class of receptive field profiles than the commonly used Gabor functions. Second, we demonstrate that there is, however, an advantage of using the Gabor filters: Similar to our perception, the stereo algorithm with the Gabor filters has a small bias towards zero disparity. Third, we prove that the complex cells as described by Freeman et al. compute disparity by effectively summing up two related cross products between the band-pass filtered left and right retinal image patches. This operation is related to cross-correlation but it overcomes some major problems with the standard correlator. Fourth, we demonstrate that as...

Abstract not found