Recent Changes - Search:


edit SideBar


Performance Characterization In Computer Vision

R.M. Haralick, CVGIP-IE, 60, 1994, pp.245-249.

The title rather over-stretches the mark regarding the actual content, as this paper covers perhaps only a third of the topic with one particular slant. However, perturbation analysis is a fundamental tool in many areas of engineering for the assessment and characterisation of systems. The idea is linked to the full quantitative use of Likelihood for the estimation of Cramer-Rao Bound estimates of parameter covariances. It is an indispensable approach to understanding the information available in computed data and for the evaluation and testing of algorithms. What is left unsaid in this paper, but can perhaps be inferred, is that algorithms which are not predicted to produce stable outputs (or worse, that the stability can not be evaluated) are probably of little value in a larger system. Such a view leads to the conclusion that much more work is needed in the field to assess existing algorithms using these techniques. For many papers this is the missing link, which would take the work forward to the point that it could be practically used.

NAT 9/3/2005

Edit - History - Print - Recent Changes - Search
Page last modified on March 05, 2012, at 03:57 PM