Objective evaluation for segmentation of microscope images of materials

In this paper, we propose a new objective evaluation method to evaluate segmentation of microscope images. We incorporate this method into two existing measures of similarity from the statistics community to compare segmentations with the associated ground truth. This paper focuses on weighting object boundaries more than other quality considerations, since object boundary usually provide useful information for characterizing materials property. Experimental results using datasets from materials science are presented visually and numerically to showcase the agreement between true segmentations and objective evaluation scores. The proposed method is expected to be widely useful for incorporating various quality considerations into evaluation measures.