Connectionists: time dependant non-linearity in motion opponency, and the rivalry of LRAM and SRAM

[Eero Simoncelli]

Dear Pooya,

The main nonlinearity in the Adelson/Bergen energy model is the  
squaring (i.e., the "energy" computation), which comes before the  
subtraction.   They also described a "normalized" version, in a 1986  
conference paper, in which the opponent energy signal is divided by  
the static energy, pointing out the relationship to least-squares  
computations of motion from gradient measurements (Lucas & Kanade,  
1981).

     The Extraction of Spatio-Temporal Energy in Human and Machine  
Vision
     Adelson, E. H., and Bergen, J. R.,
     Proc. Workshop on Motion: Representation and Analysis (pp.  
151-155), Charleston, SC; May 7-9 (1986).

Related "divisive normalization" descriptions of the physiology are  
presented in

     A Model of Neuronal Responses in Visual Area MT
     E P Simoncelli and D J Heeger,
     Vision Research, vol.38(5), pp. 743--761, Mar 1998.

The normalization is idealized as an instantaneous computation in this  
paper, but the real system presumably accomplishes it using feedback.

Best,

Eero Simoncelli
    Investigator, Howard Hughes Medical Institute
    Professor,
       Center for Neural Science, and
       Courant Institute for Mathematical Sciences
       New York University


On 14 Aug2008, at 6:06 AM, Pooya Pakarian wrote:

> Hi All
> Please let me ask two questions. I very much appreciate your  
> feedbacks.
>
> 1- the last stage of both the elaborated Reichardt Detector and also  
> the
> Adelson-Bergen model for motion detection is the motion opponency  
> stage that
> is a simple subtraction without any non-linearity or time-dependency  
> or
> mutual inhibition, etc. Is there any article elaborating this last  
> stage by
> for example adding some time-dependant non-linearity to it?
>
> 2- what are the major articles in which I can learn about the  
> computational
> models of rivalry between the short-range and the long-range  
> apparent motion
> system. this rivalry is hypothesized to occur in visual stimuli like  
> the
> missing fundamental grating or the Ternus effect, or the reversed  
> Phi, if
> the viewers' distance is adjusted properly. this rivalry is  
> interesting
> because the two rivals are not at the same level of processing.
>
> Best thanks for your attention
> Pooya
> --
>
>
> -- 
> This message has been scanned for viruses and
> dangerous content by MailScanner, and is
> believed to be clean.
>