Selectivity and tolerance for visual texture in macaque V2

Abstract

As information propagates along the ventral visual hierarchy, neuronal responses become both more specific for particular image features and more tolerant of image transformations that preserve those features. Here, we present evidence that neurons in area V2 are selective for local statistics that occur in natural visual textures, and tolerant of manipulations that preserve these statistics. Texture stimuli were generated by sampling from a statistical model, with parameters chosen to match the parameters of a set of visually distinct natural texture images. Stimuli generated with the same statistics are perceptually similar to each other despite differences, arising from the sampling process, in the precise spatial location of features. We assessed the accuracy with which these textures could be classified based on the responses of V1 and V2 neurons recorded individually in anesthetized macaque monkeys. We also assessed the accuracy with which particular samples could be identified, relative to other statistically matched samples. For populations of up to 100 cells, V1 neurons supported better performance in the sample identification task, whereas V2 neurons exhibited better performance in texture classification. Relative to V1, the responses of V2 show greater selectivity and tolerance for the representation of texture statistics. These responses are typically characterized by measuring selectivity for specific visual attributes, such as light intensity or color, and local structural properties, such as spatial position, orientation, and spatial frequency. Stimulus selectivity, along with the complementary notion of "invariance" or "tolerance" to irrelevant variation, provides a de facto language for describing the functional roles and relationships of neurons in visual areas. For example, simple cells in the primary visual cortex, area V1, are selective for orientation (1) and spatial frequency (2-4). Complex cells exhibit similar selectivity, but are also more tolerant to changes in spatial position (1,5,6). Component cells in area MT (or V5) exhibit selectivity for orientation and speed, but (relative to their V1 inputs) are more tolerant of changes in location and spatial frequency, whereas MT pattern cells are tolerant to changes in orientation (and, more generally, spatial structure) (7).Neurons in the inferotemporal visual cortex (IT) are selective for visual images of particular objects, but are tolerant to identitypreserving transformations, such as translation, rotation, or background context (8,9). This tolerance increases from area V4 to IT (10), suggesting that an increase in selectivity is balanced by an increase in tolerance, preserving overall response levels and their distribution across neurons (11). However, the selectivity and tolerance of visual representations in midventral areas, particularly area V2, have been more difficult to establish because we lack knowledge of the relevant visual attributes. V2 neurons receive much of their afferent drive from V1, have receptive fields that are r...

Cross-orientation suppression in visual area V2

“…The biphasic pooling was prominent in the data set (34% of neurons), with uniform pooling observed in the remaining cases. Overall, the observed patterns of selectivity based on locally orthogonal excitatory and suppressive features that are repeated across a range of spatial position could mediate the observed selectivity of V2 responses to textures1718 and texture boundaries132021.…”

Local Order within Global Disorder: Synaptic Architecture of Visual Space

“…For example, the population in Figure 6a is best described using five PCs but only three can be readily visualized at once. In order to visualize PC coefficients in a two-dimensional (2-D) coordinate system we employed the t -distributed stochastic neighboring embedding algorithm ( t -SNE) (Maaten and Hinton, 2008; Ziemba et al, 2016), as shown in Figure 6b ( left ), which differentiate groupings of spines. We then ran a supervised k-means clustering algorithm on the synaptic PCA coefficients in order to cluster spines (see Methods), where the optimal number of clusters was chosen using the Calinski Harabasz score (Caliński and Harabasz, 1974) as the criterion (Fig.…”

On texture, form, and fixational eye movements

“…1 for an example). Recent studies have invigorated the debate in visual neuroscience as to whether form or texture are the primary drivers for visual perception [6, 7**, 8]. For example, often shapes defined by textures are perceived more readily than those based on outlines [8].…”

“…In the primary visual cortex (V1), neural responses are tuned to specific combinations of angles at specific positions. Presumably this explains why V1 neurons are better at discriminating individual samples with a shared texture than different texture types from each other [7**]. This situation changes in the secondary visual cortex (V2) where neurons trade the ability to distinguish individual samples for their ability to distinguish between different texture types [6,7**].…”

“…Presumably this explains why V1 neurons are better at discriminating individual samples with a shared texture than different texture types from each other [7**]. This situation changes in the secondary visual cortex (V2) where neurons trade the ability to distinguish individual samples for their ability to distinguish between different texture types [6,7**]. This review will first discuss recent results on the neural mechanisms for detecting edges defined in changes in luminance or texture.…”