—Few evaluation protocols were suggested to assess quality of skeletonization methods of grayscale images. Most of these protocols employ criteria and images both devoted to target application. No common image databases are available and the validation of skeleton structural properties under grayscale object variability suffers from a lack of standardized procedures. These properties are namely the preservation of geometry, topology and extremities respectively related to skeleton location and morphological quality. In this paper, we propose an evaluation protocol for skeletonization applied to grayscale curvi-linear structures that focuses on skeleton structural properties, regardless of application specificities. We first identify challenging situations for skeletonizing grayscale images and then, construct a synthetic image database of objects with varying contrast, curvature and width. Secondly, we focus on criteria that reflect skeleton structural properties to assess its quality and noise robustness. We apply the proposed protocol on skeletonization methods within differential geometry framework that highlights good skeleton location and morphological thinning category that promotes skeleton connectivity. Experimental results indicate that the proposed protocol is able to describe the behavior of the criteria regarding the structural rendering of skeletonization methods.