8

The reading for this class was radically different from our past readings, but I appreciated the change and the "behind the scenes" look at the technological methods behind 3D facial reconstruction. Of course, many explanations were pretty obscure to me, but I appreciated the article when I finished it. Overall, what I took away from this piece was that while it is not a perfected method, it is definitely a plausible way of re-constructing the deceased.

“Reanimating the Dead: Reconstruction of Expressive Faces from Skull Data” – Kahler, Haber, Seidel

- Purpose: provide facial reconstruction approach that fits an anatomy based virtual head model

- facial reconstruction for postmortem identification of humans is challenging

- one of most popular tissue reconstruction methods bears many resemblances to surface fitting techniques on the computer

- manual reconstruction builds on the tight shape relationships between skull and skin

- 1935 – first documented case using 3D facial reconstruction

- not only correlation between skull and skin, but also muscle activity and skull shape

- the preparation of the skull is the beginning of the process of facial reconstruction

o also, dental analysis

- face reconstruction proceeds with approaches: 1) anatomical method, 2) tissue depth method

o anatomical method is through sculpting and takes a lot of time. Tissue depth method preferred, in which standard sets of tissue thickness measurements at specific points of the face are used.

o Computer graphics perspective

– surface interpolation problem

- Computer graphics in forensic application are still limited/simple

- Facial template chosen from a database of scanned faces

o Tissue thickness values

- Variety of techniques exist to create face models from images or scan data

o Deforming generic head model using radial basis functions

- 3D skull data acquired from volume scans and extraction of bone layers

o have physical skull on hand as reference

- skull model has landmarks, which can be manipulated

- head skull model has: 1) skin surface 2) virtual muscles (to control animation) 3) mass-spring system 4) landmarks

- tissue depth values at marker positions define the basic shape of the reconstructed head

- to keep user interface uniform, there is placement of vertical and horizontal line guides on the frontal view of the skull

- rules that influence the shape of the nose and mouth

- neutral pose of the face is preferred as the most generic facial expression in manual facial reconstructions

o however, virtual reconstruction allows for various expressions

o need a controllable tongue, eyeballs, and eye lids to allow for animation

- when actual appearance of individual is unknown, the accuracy of the reconstruction can only be guessed at

o Reconstruction plausible

- Results – overall good reproduction of facial shape and proportions, with well-matched details

o Takes less than a day to create reconstructed face model

- 3D scans – contact free interaction with physical skull, but scanning process is not so detailed

- tissue depth values should not be taken at face value in 3D facial reconstruction, but act as guides instead final facial reconstruction still relies heavily on artistic skills and intuition