Dispersion assessment in the location of facial landmarks on photographs

Abstract

The morphological assessment of facial features using photographs has played an important role in forensic anthropology. The analysis of anthropometric landmarks for determining facial dimensions and angles has been considered in diverse forensic areas. Hence, the quantification of the error associated to the location of facial landmarks seems to be necessary when photographs become a key element of the forensic procedure. In this work, we statistically evaluate the inter- and intra-observer dispersions related to the facial landmark identification on photographs. In the inter-observer experiment, a set of 18 facial landmarks was provided to 39 operators. They were requested to mark only those that they could precisely place on 10 photographs with different poses (frontal, oblique, and lateral views). The frequency of landmark location was studied together with their dispersion. Regarding the intra-observer evaluation, three participants identified 13 facial points on five photographs classified in the frontal and oblique views. Each landmark location was repeated five times at intervals of at least 24 h. The frequency results reveal that glabella, nasion, subnasale, labiale superius, and pogonion obtained the highest location frequency in the three image categories. On the contrary, the lowest rate corresponds to labiale inferius and menton. Meanwhile, zygia, gonia, and gnathion were significantly more difficult to locate than other facial landmarks. They produced a significant effect on the dispersion depending on the pose of the image where they were placed, regardless of the type of observer that positioned them. In particular, zygia and gonia presented a statistically greater variation in the three image poses, while the location of gnathion is less precise in oblique view photographs. Hence, our findings suggest that the latter landmarks tend to be highly variable when determining their exact position.

Modeling Facial Soft Tissue Thickness for Automatic Skull-Face Overlay

“…As has already been explained in this article, placing landmarks on a photograph is a difficult task that presents a particular degree of imprecision [18], [19]. Thus, the imprecision in the location of landmarks has a negative effect on the final SFO achieved (best transformation).…”

“…The uncertainty related to the location of facial landmarks refers to the difficult task of placing landmarks on a photograph [18], [19]. Among other reasons, the definition of many anthropometric landmarks is imprecise in nature.…”

“…These approaches are based on overlaying a skull 3D model on a facial photograph by minimizing the distance among pairs of landmarks as well as handling the imprecision due to the facial landmarks' location [18], [19]. This minimization process involves the search for the specific projection of the skull 3D model that reduces all the distances between every pair of corresponding landmarks as much as possible.…”

Photo anthropometric variations in Japanese facial features: Establishment of large-sample standard reference data for personal identification using a three-dimensional capture system

“…Although we can find these obscure landmarks on the surfaces of 3D facial images, we cannot identify them exactly without palpation. Previous studies [22,23] showed that there were comparatively large errors in the positioning of some landmarks such as (se), (zy), and (go) in photo anthropometry.…”

Modeling the consistency between the bony and facial chin outline in craniofacial superimposition

“…The works developed by authors such as [10,11,12,13,14] serve as examples of how computer algorithms can automate SFO and accommodate the uncertainty/fuzziness of some facial landmarks [15] and of the soft tissues [16,17]. These methods represent a clear step forward since they have managed to reduce time and subjectivity inherent to manual approaches.…”