A model of neuronal responses in visual area MT.

By Eero P. Simoncelli, David J. Heeger. 1998

Electrophysiological studies indicate that neurons in the middle temporal (MT) area of the primate brain are selective for the velocity of visual stimuli. This paper describes a computational model of MT physiology, in which local image velocities are represented via the distribution of MT neuronal responses. The computation is performed in two stages, corresponding to neurons in cortical areas V1 and MT. Each stage computes a weighted linear sum of inputs, followed by rectification and divisive normalization. V1 receptive field weights are designed for orientation and direction selectivity. MT receptive field weights are designed for velocity (both speed and direction) selectivity. The paper includes computational simulations accounting for a wide range of physiological data, and describes experiments that could be used to further test and refine the model.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

Perceiver: General Perception with Iterative Attention (2021, background): “Some attention maps exhibit spatiotemporal structure reminiscent of the ﬁlters seen in spatiotemporal image processing (Adelson & Bergen, 1985; Simoncelli & Heeger, 1998).”
End-to-end Optimized Image Compression (2016, methodology): “neurons (Heeger, 1992; Carandini and Heeger, 2012). Cascaded versions of such models have been used to capture multiple stages of visual transformation 2 Published as a conference paper at ICLR 2017 (Simoncelli and Heeger, 1998; Mante, Bonin, and Carandini, 2008). Some earlier results suggest that i”
Normalization as a canonical neural computation (2011, background): “Normalization in other cortical areas There is evidence for normalization downstream from area V1, and particularly in the visual cortical area MT”
Convolutional neural networks applied to house numbers digit classification (2012, background): “Lp pooling is a biologically inspired pooling layer modelled on complex cells [13, 5] who’s operation can be summarized in equation (1), where G is a Gaussian kernel, I is the input feature map and O is the output feature map.”
A Biologically Inspired System for Action Recognition (2007, methodology): “We compare 3 types of motion-direction sensitive S1 units: a) Gradient-based features (It/Ix, It/Iy); b) Optical flow based features; c) Space-time oriented filters [29], which have been shown to be good models of motion-sensitive simple cells in the primary visual cortex [15].”
Normalizing the Normalizers: Comparing and Extending Network Normalization Schemes (2016, background): “In this section we ﬁrst review related work on normalization, followed by a brief description of regularization in neural networks.”
A Canonical Neural Circuit for Cortical Nonlinear Operations (2008, background): “For example, a complex cell in V1 (or a motion-selective cell in MT, or a shape-selective cell in IT) may show a tuned response for a particular edge orientation (or velocity, or a face image), a sigmoidal response for the image contrast, and a translation-invariant response within its receptive fie”
Automated home-cage behavioural phenotyping of mice. (2010, background): “A hallmark of the system is its hierarchical architecture: The first processing stage corresponds to an array of spatio-temporal filters tuned to four different directions of motion and modelled after motion-sensitive (simple) cells in the primary visual cortex (V1”
Perceptual Video Coding Based on SSIM-Inspired Divisive Normalization (2013, background): “It is shown to be a useful framework that accounts for the masking effect in the HVS, which refers to the reduction of the visibility of an image component in the presence of neighboring components [35], [36].”
Metamers of the ventral stream (2011, unknown): “Finally, local correlations are compatible with models of cortical computation that propose hierarchical cascades of linear filtering, point nonlinearities and poolin”
Probabilistic decision making by slow reverberation in cortical circuits. (2002, unknown): “The total synaptic currents are given by future work, it will be interesting to consider in more detail how motion stimuli are represented in MT via the Isyn(t ) Iext,AMPA(t ) Irec,AMPA(t ) Irec,NMDA(t ) Irec,GABA(t ) distribution of correlated neuronal responses (Zohary et in which al., 1994; Shadl”
Motion illusions as optimal percepts (2002, background): “, and detailed models of the physiology of the motion pathwa”
Population coding under normalization (2009, background): “…& Sun, 2006), the lateral geniculate nucleus (Bonin, Mante, & Carandini, 2005, 2006), primary visual cortex (Carandini et al., 1997; Heeger, 1992; Ringach & Malone, 2007; Rust, Schwartz, Movshon, & Simoncelli, 2005), area MT (Simoncelli & Heeger, 1998), and area IT (Zoccolan, Cox, & DiCarlo, 2005).”
Dynamic scene understanding: The role of orientation features in space and time in scene classification (2012, background): “In addition to dynamic texture analysis [8], measures of spatiotemporal oriented energy have been used previously to capture a wide range of dynamic patterns, including image motion [1, 12, 32], semi-transparency [6] and human actions [5, 16, 7].”