| MAYO 2013 |
  |
|
 |
Día con actividad |
 |
Día actual |
|
Síguenos en:
|
|
 |
Reflections on Image Quality
|
23|06|2006
|
Fecha: 23|06|2006 - Hora: 12:30
Lugar: CVC
Autores: Dag Waaler - Gjøvik University College
|
Quality is a frequently used term by most of us to rate properties of products, systems and services. Obviously high quality is better than low quality. When asked to define quality, however, we are often unable to do so - and thus not nor able to agree with others as to the proper assessment of it. Image quality is no exception.
The present talk is a result of my own inquiries into some aspects of image quality, in particular in relation to medical diagnostics and non-destructive testing (NDT) of materials.
The first thing to notice is that quality has several dimensions to it, and that in order to assess and discuss quality we need to specify the particular dimension(s) in question. In the scientific imaging context these might be features, performance, reliability and/or conformance.
Secondly, in addition to quality features such as spatial resolution and/ or contrast sensitivity one may need to specify the exact purpose or task of the imaging system, which may relate to classification tasks such as medical diagnostic accuracy in medicine and condition assessments in NDT.
The imaging performance quality may then be objectively rated by a model observer sign al-to-noise-ratio as a combined result of the imaging task and the feature qualities of the system.
One important consequence of this is that the ranking of two systems in terms of quality thus might depend on the nature of the task. This might also suggest a strategy for local image quality weighting in automated estimation tasks such as edge tracking.
|
|
|
|
|
