Green, M.A. and Evison, M.P. (1999). Interpolating between computerised three-dimensional forensic facial simulations. Journal of Forensic Sciences, 44(6), 1224-8. ISSN 0022-1198 (Print).
TECHNICAL
OTE
Micha el A. Green, J M.B. and Mortin P. Evison,1 Ph.D.
Interpolating Between Com puterized Three-Di mensional Forensic Facial Sim ulations*
REFER · NCE: Green MA , Evison MP. Interpolating bet ween com pu terized three-dim ension al fo rensic facial simulation s. J Forensic Sci 1999;44(6) 1224-1228
tions produced using the " pl as ti c" technique. M orphing is achieved using qm.DIf, a demonstration program, and usi ng the Coorclinarc/l1 lerpo/alOr node of vutual reality modeling langu age or VRML (7).
ABSTRACT: The shape of th e skull is a lJmiled predictor of facial appeannce and fa cial recon slJ1.l ction methods ca n offer at best a re semblance to the indi vidual durin ~ Ii f~ . Vm-ious unkn own factors in addition to those relating to age, buiiJ, obesity, ethni c group, and the varying shape of th e eyes, nose, and mOllth mean th at a va.!'iety of fa cia] reconstl1.lctiDns could be produced from any single slcull. A d i sadvan t3 g~ of bo th traditional "plasric" fa cial reco nstruction melho ds and existing computerized equi valents is tlIat the fini shed reconstl1.l criolJ is fix ed and sub stantial furtll er work is requ ir~ J if a modified ou,come is to be presen ted. An advaI1lage of compu teri za ti on is thal if offers a greate r poteI1lial for easy revisio n of images. Here we describe a meth od for interpolating OJ "morpll ing" between virtual images 01 faci al reconstructions produced using the "plastic'; technique which enables the interacti ve fatten.ing of tile race in three dimensions. Morphing is acl1 ieveJ usin g qmolj, a demon strati on program, and usi ng lhe interpolator node of virtual rea lity modeling language (VRl\1L).
Materials and Methods
K EYWORDS: foren sic science, forensic antluopol ogy, physical anthrvpol ogy, l~ ci ui reco nstru ction, faci al reproduc ti on, th ree-di mensionalmodeling, virtual reality modelin g language
It is well known that th e shape of the slrulJ is a limited predictor of faci al appea.!-ance and that facial reconsl:illction methods can offer at best a semblance of the individual dm1ng life (1). The influence of various unknown fac tors in addition to tb ose relating to age, build, obesity, ethnic group , and the var)'ing sl1ape of the eyes, nose and mOLlth me3.J) that a vaIlety of faci al reconstructions could be gener ated [rom any sin gle skull. A disadvantage of botl1 traditional "plas tic" facial rec onstruction methods (2-4) and existing computel1zed equiva len l.'> (5 ,6) is that the finished rec onstruction is fixed and wb stantial furtller w ork is requ ired if a modified outcome is to be pre sented. An ad va ntage of com pu terization is tl1at it offers a greater po tential for easy revision of images. The Llse of computer graphic techniques further offers a means of malung available an international ser vice vi(l the Internet and of increasillf the fl exibility of presenta ti on of facial reconstructions. Here we present a m e tbod for interpo lating or "morph:ing" between virtual images of facial reconstrucI Department of Forensic Path ology. The University of Sheffield , The Nledico-Legal Centre, ShefDe ld. S3 7ES , United Ki ngd om . " The Sheffi e.ld i'acial reconsLruction project IS supported by the UK T-Tome Office. Re eived 2 Dec. 1998; an cl in revised form 26 Feb. 1999 ; accepted 1 March 1999.
Plastic rec onstmctions of typical male and female archaeologi cal skulls , produced us ing published "obese" and "emaciated " tis su e depth data (2), were scanned usi ng a Cyberware 3030RGBICN color laser scanner and Echo software (Cyberw are, Monterey, CA l l1lnning on a Silicon Graph ics Ind y Workstation and IRIX 5.3 op erating system (Silicon Graphics , Mountain View , CA l. Scans of the antel10r half of the head were produced . To ensure tha t "obese" and "em aciated" versions of rec onstructions produced from identi c al sku lls were captured in preci sely th e same orientation, th e re consul.lctions were positioned on the sca nn er platfonn in the Frank furt plane (8 ), positioned at a constant height on the scanner platform using telescopic stand s. As the s tands were supelfluous to the final image, they w ere covered w ith a black cloth- wmch is in vi sible to the laser scanner-and hence excluded . The 180 0 scan s w ere 0l1ginated at the dorsal marg in of tbe light external audi tory meatu s and tennin ated at the dorsal margin of the left. Images were cropped app roximately at the hair line using the crop comman d in Echo. Void s in the scanned image w ere filled using tbe fill com mand. Three-dimensional images in Echo format were convened into Silicon Graphi cs Open Ii1ventor™ file form at usi ng cy2iv, an IRIX conunand line utility program supplied by Cyberware, which generates ou tput in quadrilateral mesh (q uadmesh ) form al. Co-or dinate matri x sizes of 151 X 151 w ere generated fro m cy2il' and the Open Inventor (.iv) files were transferred to a Silicon Graphics O 2 works tation running IRIX 6.2 for morphing. Morphing was ach ieved Llsing qmolf, a demon s tr ation program whicb wi ll interpolate between any two Open In ventor three-di mensional im ages in quadmesh format. QmOlf was used to inter polate betw een "emaciated" and "obese" reconsllllctions prod uced fro m typical female and male skulls. As the m oving images gener ated by qmo lf cannot easil y be displayed via the Internet, the Co ordil1C1 leinlerpolalor node of VRML 2.0 was used to crea te a file which will interactively display morphing between 40 X 40 co-or dina te "emaci ated" and "obese" reconstruction s on the Internet. Tl1e CosmoWorldsH1 (Silicon Graphics, Mountain Vie w _ CAl V RML editor was used to r ender the face with a "w hite" skin-like su rface material and add text to the model!\, selection of images was captured from each morphing experiment at d ifferent stages of i nterpolation. .
1224
Green, M.A. and Evison, M.P. (1999). Interpolating between computerised three-dimensional forensic facial simulations. Journal of Forensic Sciences, 44(6), 1224-8. ISSN 0022-1198 (Print).
GREEN AND EVISON • INTERPOLATING BETWEEN FACIAL SIMULATIONS
12:
FIG. l-Images showing qmorf interpolation bel:\1leen two reconstructions from a female archaeo logical skull produced using the "emaciated" ane "obese " tissue dep th data.
1226
Green, M.A.SCI andENCES Evison, M.P. (1999). Interpolating between computerised three-dimensional forensic facial simulations. JOURNAL OF FORENSIC Journal of Forensic Sciences, 44(6), 1224-8. ISSN 0022-1198 (Print).
FIG . 2-lmages showing qrnorf interpolation between fWO recoils/rue/ions frOI11 a male a rchaeological skull produced using Ih.e "emacialed" and 'obese" lissue depth dala (hllp:lljorensic.shefac. uk/jfslfallen .mov).
Green, M.A. and Evison, M.P. (1999). Interpolating between computerised three-dimensional forensic facial simulations. Journal of Forensic Sciences, 44(6), 1224-8. ISSN 0022-1198 (Print).
GREEN AND EVISON . · INTERPOLATING BETWEEN FACIAL SIMULATIONS
Results Images showing the results of qmoifmorphing between "emaci ated" and " obese" extremes for reconstructions of a female and a male skull are shown in Figs . I and 2, respectively . Images of a VRML 2.0 file lo aded into a VRML browser (Silicon Graphics' Cosmo Playe r™) and showing interpolation between "emaciated" and "obese" extremes of a female reconstruc tion are shown in Fig. 3. Movement through a series of three preset viewpoints is also shown in this figure.
Discussion When supplied with file s representing "emaciated" and "obese" reconstruction s from the sa me skull , qmolf will interpolate smoothly between the extremes. The same res ult can be achieved in a simple VRML modeL The images can be examined interac tively ill three dimensions. As well as generating images of the re
1227
construction at different levels of obesity, the mean tissue depth dataset is also represented. The time taken to morph automatically between two 151 X 151 co-ordinate datasets is three seconds. In qmOlj, a slider control allows the L1ser to manipulate the interpola tion process manually or to stop at any particular point in the process . A plasti c facial reconstruction can take up to a week to produce, although the process can be reduced to one or two days in an emer gency. If the recons truction has to be amended some hours or days further work can be required . Amendments to finished tW'ee-di mensional reconstructions in VRML or using morphing programs sllch as qmoif can be achieved in seconds . The same approach can be used to age the face, to move between typical female and male tissue depth datasets-in the case of a robust female, gracile male or un-sexed skulls, for example-or to morph between typical datasets for the major ethnic categor ies where a skull cannot be clearly placed in one or other group . Since important features offa
FlO. 3-Emages p roduced llsing a VRM L model to interpolate hetwe en "emaciated" and "ob ese" reconstructions from. a female archaeological skull, as displayed on the Enternet (htlp://fore n s;c.s h ~fac.!lkJjjj/t(llten. wrl). /vEovemeJII fhrough three preset vie wpoints is shown in the bottom half of the figure .
Green, M.A. and Evison, M.P. (1999). Interpolating between computerised three-dimensional forensic facial simulations. Journal of Forensic Sciences, 44(6), 1224-8. ISSN 0022-1198 (Print).
1228 JOURNAL OF FORENSIC SCIENCES
cial appearance-the shape of the eyes, tip of the nose and the lips-cannot be predicted from the skull, VRML or programs such as qmorj can be used to allow the user to select from a variety of shape options. Since qmOli performs a simple linear interpolation between the start and end co-ordinates, future research is directed at the devel opment of VRML models and Java algorithms for non-linear inter polation which more appropriately reflect the anatorrtical processes underlying the fattening of the face. Acknowledgments We thank Mandy Holmes for producing model plastic facial re constructions during her M.Sc. dissertation research in the Depart ment of Archaeology & Prehistory , University of Sheffield. This research is supported by the UK Home Office. QmOliis an Open Inventor demon stration program written by Gavin Bell.
References I. Tyrre ll AI, Evison MP, Chamberlain AT, Green MA. Forensic three-di mensional facial reconstruction: histOlical review and contemporary de velopments I Forensic Sci I 997;42(4):653-61.
2. Prag J , Neave R. Making faces: using forensic and archaeological evi dence. London: British Museum Press 1997. 3. Lebedinskaya GV, Balueva IS , Veselovskaya EV. Principl es of facial reconstruction. In: Js~an MY , H elmer RP, editors. Forensic analysi, of the skull. New York: Wil ey Liss 1993; 183-98. 4. Fedosyutkin BA, Nainys JV. The relationship of skull morpho logy to fa cial features In: ls~an MY, Helmer RP, editors. Foren sic analy sis of the skull. New York: Wiley Li ss 1993;199-2 14. 5 . Vanezis P, Blowes RW , Linney AD, Tan AC, Richards R, Neave R. Ap pli ca tion of 3-D computer graphics for facial recon slnlctio n and com parison with sc ulpting techniques. Forensic Sci Int 1989;42:69- 84. 6. Ubelaker DH , O'Donnell G . Computer assisted facial reconsuuction. J Forensic Sci 1992;37: J 55-62. 7. Ames AL, Nadeau DR, Moreland JL. VRML 2.0 sourcebook. New York: Wiley 1997. 8. Farkas LG. Anthropometry of the head and face. New Yark: Raven Press 1994. Additional information and reprint requests: Martin Evison Department of Foren sic Pathology The Uni vers i[y of Sheffiel d The Medico-Legal Centre Sheffield, S3 7 ES United Kingdom E-mail:
[email protected] 
