News |  Sitemap |  Contact
PDF Export  | 

Benefits of ontologies in image processing workflows


In all domains and in particular in image processing applied in microscopy, the reuse is an important stake. It reflects the desire to capitalize the expertise and the knowledge of the domain's actors in order to optimize the community work. It is not uncommon that during the creation of an image processing workflow, the choice of the suitable algorithm poses a problem. For example, the choice of a deconvolution algorithm depends on a combination of factors which are not always easy to know and/or identify.

To solve this problem, we believe these factors can be formalized in an ontology. The most known definition of ontology is given by Gruber who defined it as a formal and explicit specification of a shared conceptualization. More practically, it is composed on one hand by a vocabulary which describes consensually a reality, and on the other hand by a set of hypothesis about the meaning of terms. But, due to the formal nature of ontologies, their modification is a difficult task, which requires the presence of an expert. In this way, if the ontology must help to choose a deconvolution algorithm, the ontology must be complete and composed by all the factors which have an influence on this choice. But in practice, this situation is not realizable because few algorithms are describes by their creator, and therefore the factors are unknown.

Our proposal consists in allowing users to describe these algorithms. In order to enable users to add their knowledge themselves in an ontology, we have recourse to a less formal knowledge form than an ontology. We use the social tagging, which qualifies the practice of describing objects using freely chosen keywords. In our solution, all algorithms are considered as independant boxes and are implemented in a workflows management system. These systems have a rich user interface which allows users to create image processing workflows by a simple drag and drop of algorithms in the workspace and linking them together. Broadly speaking for imageJ users, it is a graphical equivalent of the macro language. But the significant difference between the both lies in the ease to make an image processing workflow. With the macro language, users must know algorithms, when and how use them. If they choose a wrong algorithm, the final result will be skewed. With our approach, users must just know what they want to do. The ontology will propose algorithms corresponding with users needs. On the other hand, if the ontology does not include a suitable algorithm or ignores the factors which designate it as such, we will ask users to remedy it by a tagging. Our tagging system guides all the process of acquiring knowledge in order to capitalize all users answers, and enhance the ontology.

Thus, users are able to take advantage of knowledge resulting of previous users experiments and therefore, they will be able to work faster and more efficiently.


image processing workflow, workflow management system, ontology, social tagging, folksonomy

Administrative data

Presenting author: charles-georges.guillemot
Organisation: UHA - Laboratoire MIPS EA2332, groupe LSI, 12 rue des freres Lumiere, 68093 Mulhouse

co-authors: Frederic Fondement (1)

© Luxembourg Institute of Science and Technology | Legal Notice