Alex, my answer would be from a clinician's point of view or more on a theoretical level. So if you are ok with that and with no research to back it up I would say that it depends on whether the discrimination is being done on a conscious level or not.

Example: When walking on an unevenly textured surface where there is little risk of falling, conscious attention may be on the goal out in space, yet below consciousness processes may be selecting "grippy" areas to step on based on texture variations of the surface below. In this instance global.

Second Example: A person is doing a similar thing but now the the surface being traversed is very slippery in some areas (black ice for example). Now a huge amount of conscious resources are going to selecting where to step next and the texture variations may be very much in a figural aspect of seeing and then, I feel (conjecture on my part) would be processed more locally.

Paul Harris, OD

Texture is a structural frequency phenomena, neurons are uniquely suited to detection of structural frequency because of their own structural frequency as noted by the index of fractality. As such texture at least tends to be a local phenomena in that it is the pattern of activation of different neurons that tends to indicate the structural frequency of the surface.

Figure and ground on the other hand can be inverted as is suggested by Eschers work, where he played with figure and ground. Thus it is a less localized and therefore more global phenomena and responds to the preconcieved concept of what is figure and what is ground. The exception seems to be certain types of perception where the figure is arbitrarily set by the mind, and can't be adjusted such as red berries on a tree.

Thanks Graeme. The Escher point is one for me to follow up on. In my case however I am not too concerned with which becomes figure or ground, but instead the 'rules' by which the distinction is made. I am trying to explore the relative contributions of local and global processes to the segmentation of natural image.

I have not heard the term 'structural frequency' before, do you have some papers on it? A simple database search yielded only superfluous results for me. Is it an engineering term?

Hi Alex,

You talk about "texture discrimination (and in turn figure-ground segmentation)", as though texture discrimination is the primary input into figure-ground segmentation. Although a neural 'tag' of texture may play a role to determine whether parts of a visual scene belong to the same surface, figure-ground segmentation can also be guided by the category of objects that are present in a scene (see for example the famous dalmatian dog in which there is actually very little texture to be discriminated, but segmentation still occurs; attached to this post). One hypothesis is that object categorization precedes figure-ground segmentation, or works in parallel with it (See the work by Mary Peterson, psychological science 1994 and a recent publication by me: Fahrenfort et al, 2012 in PNAS).

In this sense, figure-ground segmentation is not a local process at all, but very much a distributed effort by the visual cortex. Neurons high up in the ventral stream have receptive fields that are large enough to encapsulate entire objects / object categories, showing category specific responses as a result of feedforward computations, in parallel with or maybe even prior to figure-ground segmentation. Feedback responses from these higher areas 'reach back' to earlier visual areas (V1, V2), tagging neurons with small receptive fields as belonging to the same object/surface/border (e.g. see Lamme in Journal of Neuroscience, 1995; papers by Von der Heydt and colleagues: Qiu et al, 2007 in Nature Neuroscience and Craft et al in Journal of Neurophysiology in 2007; but also the work by Appelbaum, e.g. Journal of Neuroscience 2006).

Other mechanisms that are likely to influence segmentation are neural correlates of the age old 'Gestalt' rules of perceptual grouping, such as proximity and connectedness, mediated by horizontal connections (Kapadia, Neuron, 1995). I recommend an especially good review about perceptual grouping by Roelfsema, 2006 in Annual Review of Neuroscience, outlining how feedforward and lateral/feedback connections may work together to achieve image segmentation.

Because of the considerations above, I would question an approach in which you search for some 'generic' neural correlate of texture discrimination as a primary source for figure-ground segmentation.

I hope this helps, let me know if you have any questions, cheers,

Johannes

Hello Alex:

As always, the answer to the question depends on what do you exactly mean by "texture", which is in a very difficult question to answer. There is tactile texture, food texture, etc, (wikipedia cites more than 10 different contexts for tactile texture alone) which are most probably not what you are after, I imagine you are interested in visual texture or image texture. Even when you restrict it to images or volumes it is hard to define it strictly as something more than a function of the spatial variation of pixel intensities, and of course, there is the issue of scale, that a brick from a close distance will show the texture of its material, but from a distance will show the texture of a wall, for instance. I wrote a book chapter on Volumetric Texture Analysis, if you are interested, let me know.

Hope this helps,

Carlos

http://scholar.google.co.uk/citations?view_op=view_citation&hl=en&user=DLbXRxoAAAAJ&sortby=pubdate&citation_for_view=DLbXRxoAAAAJ:8k81kl-MbHgC

Hi,

you could have a look at the work of Lamme (1995) Neurophysiology of Figure-Ground Segregation. It is still a good example of how translating in neuropsychological terms some classical claims about the global nature of figure-ground segregation that characterize Rubin's seminal work and the Gestalt psychology too