Abstract
The paper reports results of preliminary research on automotive paint dopant concentration assessment based on microscopic image segmentation. The considered task is illconditioned due to the richness and diversity in contents of images to be analyzed. The proposed procedure involves two main phases: image segmentation, where focal-plane paint addons are extracted from the background, and object analysis and classification. The results of experimental verification of the proposed method on a set of eighteen paint pigmented images (black and yellow) show that the estimation can be done with approximately 5% accuracy for paints doped with only single addon type. For add-on mixtures, the results were strongly dependent on pigment color and mutual add-on proportions.