CVNet - Ideal observers wanted

CVNet (cvnet@skivs.ski.org)
Mon, 6 May 96 01:31:01 PDT

Date: Fri, 3 May 1996 13:09:54 -0700
To: hchan@well.sf.ca.us
From: tecohn@mindseye.berkeley.edu (Theodore E. Cohn)
Subject: a proposed CVNET inquiry
Cc: thibos@indiana.edu

subject: Ideal Observers Wanted

The concept of the Ideal Observer was introduced into psychophysical
research about 40 years ago and since then many forms of the Ideal Observer
have been derived for various sensory tasks and experimental paradigms. As
a service to the vision science community, we are seeking to collect
references to all ideal observer configurations (e.g. what the Ideal
Observer should do) known to vision science. We are also looking for
closed form predictions of ideal observer performance expressing d' as a
function of signal and noise parameters in the various tasks. These are
ordinarily derived by examining the likelihood ratio for alternative
hypotheses in a given task, finding its distributions, and then finding ROC
parameters.

The point of the compilation would be to build a catalog of Ideal Observers
for the future use of others. An analogy here is Bracewell's useful
appendix which lists common waveforms and their Fourier transforms in his
textbook "The Fourier Transform and its Applications". To be most useful,
of course, solutions need to be referenced as to source, or documented if
there is no published source.

As is usual, a full listing of responses will be made available.

L. Thibos and T. Cohn

Please send replies to TECOHN@mindseye.berkeley.edu