#computationalmodelling

[dgmc90-blog]

A nice example of the usefulness of using somewhat 'abstract' stimuli to investigate midlevel vision.

[dgmc90-blog]

How do mid-level visual system mechanisms combine the outputs of low-level mechanisms? The functional objective of low-level mechanisms is, largely, to encode information pertaining to edges and contours falling within their receptive fields. So, the role of mid-level mechanisms would be to combine low-level outputs such that more complex information about the environment is being encoded. One possibility is that some mid-level mechanisms operate like an AND- gate for processing particular combinations of information at their input. For example, by orthogonally superimposing two sinusoidal gratings that share the same spatial frequency and contrast (i.e. they are ‘matched’), a coherent plaid is formed that is perceived as one checkerboard pattern, rather than two overlapping gratings (see picture above - left: grating component, middle: coherent plaid, right: non-coherent plaid). However, when two non-matched gratings are orthogonally superimposed, the two gratings are much more readily perceived. Psychophysical investigations into this has suggested that the coherent plaid may be processed as more than the sum of its parts (e.g. Nam, Solomon, Morgan, Wright & Chubb, 2009; Hancock, McGovern & Peirce, 2010). As such, this contrast between coherent and non-coherent plaid perception provides an interesting and fairly simple way of investigating mid-level (as well as low-level) mechanism operations. My PhD work, under the supervision of Dr. Jonathan Peirce, utilises grating and plaid stimuli to investigate these operations with a variety of research methodologies; EEG, psychophysics and computational modelling.