Evaluation locale de la performance basée sur un critère Mission/Qualité
Résumé
The development of an image processing chain remains a delicate task and requires many adjustments difficult to achieve for a non-specialist. The processing chain is cut into four sub-systems and a method is proposed to guide the user to the sub-system which requires the tuning of its parameters to improve the quality of the final result. To evaluate local performance, we developed a test mission/quality associated with each sub-systems. A measure based on a criterion of separability between regions by calculating the Manhattan distance allows a macroscopic adjustment. This same criterion calculated with the Earth Mover's Distance(EMD) allows then a finer adjustment. The feasibility of this approach is used on 3D tomographic images for the quality control of industrial parts.