digital photography & imaging

24 Forensic Science International

tained from the car are also important to determine the driver. In thisstudy, three examples of traffic accident cases, which were sent tothe Third Specialization Board of the Council of Forensic Medicineby the attorneys in order the driver to be determined by examiningthe medical reports written by the hospitals just after the accident,are given. In two of the cases, the vehicles crashed head on. Thepictures of the cars and the lesions of the occupants were used todetermine the drivers in these cases. The accused people were drunkso they refused to be the driver. In one case, one of the occupantsdied during the accident. In the other case, the driver of the hit carwas dead. Occupants blamed each other as being the driver. Bloodexamples were collected from the windshield, sun visor region, thedashboard and it was determined that these examples belong to thesame person. We asked the attorney to get sample of blood fromthe living occupant for cross examination. All these evidences wereused to determine the drivers.

Keywords: Traffic accident, Forensic, Driver

CRIM-FO-06The Opris Case

Giancarlo Santacroce and Lars Schmidt, Polizia Scientifica delCanton Ticino, Bellinzona/TI, Switzerland.

This article is about the dealing of a complicated case of homi-cide with connections with different European countries such asLiechtenstein, Germany, Romania, Moldavia and Russia.

On the night of the 3rd December 2002, in Ponte Capriasca, asmall village in the south of Switzerland, a 31 years old pregnantgirl was brutally murdered by two young men from Eastern Eu-rope, as a revenge to a provisional arrest led by the girl’s husband,a customer officer in Chiasso, some months prior to the facts. Af-ter the murder the two killers left Switzerland by car directed toVaduz/Liechtenstein where they left their vehicle and subsequentlydisappeared. Ten days after the case, a suspect Klaus Ingo Opris,was arrested in Romania on a charge of being the murder of theyoung girl. His accomplice had since vanished and an internationalwarrant for arrest has been issued against him.

We will present the forensic work on the crime scene with par-ticular regard on how to protect and preserve the evidence in suchcomplex cases. Also the methods of collection of biological evi-dence used in Switzerland and the results of the subsequent DNAanalysis on difficult supports will be an argument of discussion. Fi-nally, practical aspects of the searching and collection of fingerprintswith particular attention to the cyanoacrilate fuming of big surfaceswill be discussed.

The main tasks of our service were to co-ordinate the informa-tions between the different countries and resolve the problems of dif-ferent legal procedures in the countries involved. Laboratory workwill be going on until at least February 2003. The suspect will beprobably judged in the Romanian Court in Summer 2003.

Keywords: Homicide, Crime Scene Investigation, Collectionand Preservation Of Evidence

Digital Photography & Imaging

DIG-WP-01Photographic Image Authentication: A Case Study

*Levent Guner, Ekrem Malkoc, Serkan Karagoz and Serkan Erdem,Jandarma Kriminal Daire Bsk.ligi 06580 Anittepe, Ankara,TURKEY.

With the advance in information technologies, there has been anincrease in number and types of crime. Forensic science practitionershave to keep up with this advance closely. No matter in which area orfield he/she studies, every researcher has his/her hands upon an im-

age or photograph. This case is also valid for ordinary people and/orpeople having a tendency towards committing a crime. These, usu-ally computer edited images can be enhanced, cropped, multiplied orreduced of unwanted objects, or different images/photographs takenat different times could be brought together so as to make a genuinelooking impression, by using image editing softwares. When doneas a hobby or in advertising business, this issue has no consequencesother than making money probably. However if it becomes subjectof a crime, it constitutes an image authentication issue, the methodof which will be presented as a case study in this presentation. Thecase includes a 50 by 40,5 cm photograph of two people side-by-side. The court wants to know if these two people were really takenthis photograph at the same time. Pixel distribution and neighboringpixels were examined and unusual deviations in pixel values weredetermined, which shows a photograph tampered with. Pixel valuesbelonging to another, genuine photograph are also determined forthe purpose of presenting a comparative study.

Keywords: Digital image, Pixel value, Authentication

DIG-WP-02Determination of Consistency and Reliability of DiagnosticAccuracy Derived from Autopsy Photographs

*Cigdem Suner, State Institute of Forensic Medicine, Istanbul,Turkey; Can Gokdogan, Istanbul University, Cerrahpasa MedicalFaculty, Department of General Surgery, Istanbul, Turkey; BulentSam, Sadik Toprak and Mustafa Pac, State Institute of ForensicMedicine, Istanbul, Turkey.

Forensic photography plays a significant role in investigation andrecord purposes in criminological procedures. Forensic photographyis a branch of forensic sciences develops rapidly along with techno-logical improvements and subdividing into new branches requiringspecial knowledge and skills. Photography is used as a document inmany instances and sometimes becomes a part of an investigationmethod. Photographs taken at the scene, during the autopsy or exam-ination might be inspected again for education purposes or in orderto revise the details and obtain consultation from other specialists. Inthis study, the cases were chosen randomly among the hanging caseswith hanging marks, in the Morgue Specialization Department in theInstitute of Forensic Medicine, Ministry of Justice, admitted on ev-ery Wednesdays, Saturdays and Sundays between October 2001 andApril 2002. The cases were not included in the study if putreficationhad started, the patient had medical intervention, or the characteris-tics of the hanging mark could not have been differentiated. Forensicmedicine specialists evaluated lesions of 20 cases at the autopsy andphotographs were taken with a 35 mm camera using dia-positivefilms. The photographs were transferred to the digital medium andevaluated independently by four forensic specialists twice in a monthinterval and the evaluation results were recorded on a form. Thespecialists who defined the lesions at the autopsy and the specialistswho evaluated the photographs were asked to evaluate; the type ofhanging, the ascending and superficialization character, the numberof lines, if the mark had signs of characters of the hanging material,the ending characteristics, the existence of hyperemia, and the ex-istence of the knot. The specialists, who evaluated the lesions andthe photographs, had been working in the Morgue Department ofthe Institute of Forensic Medicine for at least two years, and carriedout at least 500 autopsies every single year. Self-consistency of eachevaluator was estimated with kappa test. The minimum kappa valuewas accepted as 40. In order to evaluate the difference between theevaluators, scores were constituted by pointing each photograph onappropriateness with the "gold standard". The scores were evalu-ated separately for the first sight, for the second sight and for thesummation of both sights. The differences of diagnostic accuracybetween the evaluators were analyzed with Cochran-Q test. The sig-nificance value was accepted as p< 0.05. The kappa median valuesof the four evaluators were found to be between 55.5 and 80.5. Av-

Scientific Sessions 25

erage rates of diagnostic accuracy of the specialists were, for thefirst evaluation 68 ± 0.01, for the second evaluation 69 ± 0.01 andfor the summation of both evaluations 69 ± 0.01. Thus, in hang-ing cases, the lack of accordance between the autopsy findings andphotograph evaluations was 31%. The high inconsistency was at-tributed to two dimensional evaluations, and contrary to autopsies,lack of manipulation on photographic examination. When the relia-bility ratios were evaluated - of the first, second evaluations and forsummation of both-, significant difference in diagnostic accuracypercentages was achieved (p<0.05). This difference appeared, be-cause the first evaluator had a quite high accuracy rate, whereas thefourth evaluator had a low accuracy rate among the four evaluators.Photography is only a part of the forensic investigation and it is not arealistic approach to make conclusions only based on photography.Multiple aspects of an event such as the course of the events, sceneinvestigation and autopsy findings should be interpreted together.Photographs containing adequate information and have high qualityare supportive documents in the course of enlightening the foren-sic events. The efficiency of the use of photography for forensicsciences has been widely emphasized and thus has been a part ofvarious methods. However it might lead to misinterpretations and in-correct conclusions when used alone without adequate informationand documents.

Keywords: Forensic photography, Autopsy, Hanging

DIG-WP-03Photogrammetric Techniques to Estimate a Suspect’s Size

*Loïc Canevet, Institut de Recherche Criminelle de la GendarmerieNationale (IRCGN) 1, boulevard Théophile Sueur, 93111Rosny-sous-Bois Cedex, France.

The poster will present the two techniques, used to the "Videoand Imagery" unit, named "technique of comparison" and "tech-nique of calibration", permitting to realize either a measure or acomparison of suspects sizes on the pictures of registrations videoor photographic.

These techniques have more interests when they concern uniden-tifiable people, because their face is concealed by a hood or a helmetfor example.

For each of the methods, it will present their conditions of real-ization, their methodologies, their advantages and their limits.

The technique of comparison:It is about doing a comparison of sizes of people on the pictures

of the offence and its reconstruction. By a technique of developmentand superposition of tracings, then it is possible to succeed in es-tablishing the compatibility of size between authors and suspectedpersons.

Conditions of realization:- During the reconstruction, it is necessary to get the full co-

operation of the suspects in order to position them at the chosenplaces, in concordance with the pictures of the offence, and in thesame stances.

- To get comparable pictures, the system of shootings must beidentical and its parameters (position, focus) must be kept. If thecamera has been turned, it is possible to re-establish the justification.As far as possible, to have the same conditions of lighting, it ispreferable to achieve the reconstruction to an identical timetable tothe offence.

Methodology:-The reconstruction must be recorded in the same conditions

(lighting, parameters of the system of shootings) as the offence.It agrees to position the suspects in the chosen places and in similarstances.

- After the digitalisation of the offence pictures to the similarplaces, it is necessary to elaborate a tracing including the silhouetteof the suspects (for example with Photoshop).

- After the digitalisation of the pictures of the reconstruction,

the previous tracing must be superimposed on the pictures of thereconstruction to do a survey of the size compatibility.

Advantages and limits:- This method is very fast to practice and very visual, but it is

approximate. Indeed, it doesn’t permit to have a measure of sizes.- Besides, it is only applicable if some suspects exist and want to

co-operate.The technique of calibration:It is about doing in a first time a metrological raised of the ele-

ments of the scene where the offence took place, in order to calibratethe system of shootings. In a second time, the offence pictures per-mit to value the author’s size and the reconstruction pictures give anestimation of the uncertainty of measure.

Conditions of realization:- It agrees to have at the minimum a visible position on the

offence pictures where the suspects are visible from the head to thefeet. Because, it will be necessary to seize the co-ordinates of thepixels corresponding to these points in the software of calibration(for example, the "Measure" module of Video Investigator).

- As for the technique of comparison, a conservation of the con-ditions of shootings is necessary.

Methodology:- It is about going to the scene of the offence in order to do a

metrological raised very precise of the elements of the scene and torecord this operation with the same system. A minimum seven pointsof measurements with a good geometric distribution (multiplicity ofplans) is necessary.

- After having seized the geometric co-ordinates and the co-ordinates in the picture of the previous points, it is necessary toselect the corresponding pixel to the suspect’s position and to indi-cate to the software that this point is in the plane Z=0. The softwarewill calculate the X and Y values and there giving so the positionin this plan. While selecting the corresponding pixel in the suspect’sheight and while seizing the previous values, the software will cal-culate the Z value of this point, from where the suspect’s size.

- While positioning a ruler to the previous position and whileapplying the previous technique in a known height, the softwarewill provide a calculated height. While comparing, the height calcu-lated with its true value, an uncertainty of measure is gotten in thisposition.

Advantages and limits:- This method provides an evaluation of a size as well as an

uncertainty of measure. But, the gotten size is not the person’s realsize, but in a given stance.

- It permits to give a size for a research of possible suspectedpersons.

In conclusion, these methods are used solely to discharge possiblesuspected persons and the legal value of a compatibility of sizes islet to the magistrate’s appreciation. Insofar as some suspects exist, itis counselled to apply the two methods in order to validate the sizeof the suspects.

Keywords: Photogrammetry, Height measurement,Reconstruction

DIG-WP-04Applying Multi-Frame Interfusion Processing to RestoreImages Blurred by Motion

*Feng Qingzhi, Wang Zhiqun and Yang Hongchen, Department ofForensic Science and Technology, Criminal Police College ofChina, Shenyang 110035 China.

Images blurred by motion are degenerative images frequently en-countered in the television supervision, and cause some problems inthe course of image processing and identification. It is known thatthere are many frames of image in a short fragment of video, andeach frame of image contains the primitive information about theobject. Multi-frame interfusion processing draws the motion infor-


mation from images in time series, constructs the restoration filter,and restores images by the reversal filtering in the frequency domain.

There is the interrelation between the adjoining frames of image,and the object in images changes steadily and regularly. It is an es-sential term to utilize multi-frame interfusion processing to restoreimages blurred by motion. Now, there are M frames of image in ashort fragment of video. In order to utilize statistical filtration to re-duce or eliminate the interference of random noises, let us translatethe primitive image into the frequency spectrum by Fourier trans-formation, and express the relation between the frequency spectrumand the degenerative transfer function in logarithm form.

M∑

i=1

ln [Gi (u,v)] = M ln [F(u,v)] +M∑

i=1

ln [Hi (u,v)] (1)

If M is a great numerical value, then

1

Mlim

M→∞ ln Hi (u,v)] = ln [H (u,v)] (2)

Thus the frequency spectrum of the restored image is derived asfollows:

F(u,v) = exp[H (u,v)]M∏

i=1

[Gi (u,v)]1/M (3)

In formula 3, H (u,v) is the degenerative transfer function ofimages blurred by motion, and derived from the motion informationknown by image analysis; Gi (u,v) is the frequency spectrum ofprimitive image i; F(u,v) is the frequency spectrum of the restoredimage; Finally, we can successfully restore degenerative images byreversal Fourier transformation.

In most of cases, image restoring is based on the priori knowledgethat causes images’ degeneration, Many frames of image can offerus more information than one of them, so that we can obtain thepriori knowledge and construct the restoration filter easily. This isexactly the advantage of multi-frame interfusion processing. It caneffectively enhance Images blurred by motion in the television su-pervision. And is also suitable in the digital video image processing.

Keywords: Images blurred by motion, Multi-Frame interfusionprocessing, Fourier transformation

DIG-WP-05Study on the Factor Affecting Likeness Degree of CombinedImage

*Dai Xuejing, Shan Daguo and Yang Hongchen, Criminal PoliceCollege of China, China.

As we all know, investigator usually takes use of the computer todraw an image according to the victim or witness who can point outthe like face organ on computer. For a variety of factor, the victimor witness only remembers the suspect’s special carriage-movementor some broad outline. Whether the image, which is drawn on com-puter, can be recognized depends on the likeness degree of combinedimage. Therefore, the factor affecting likeness degree of combinedimage is studied in this paper. The factor of likeness degree that isstudied in this paper is about face image, which is made up of thecast of face, the facial features, hairstyle and facial special char-acteristics. All these are individual image characteristics that areoutstanding mark of image difference.

A person’s face has several-angle visual sense that gives a syn-thetically impression of visual sense. However, once fixed on generalappearance of a picture, the face has only one-angle visual sense.If we can capture some characteristics of face, the whole effect andaim can be shown. In this test, first the original image’s organs aretransformed and accordingly the image is from like to unlike, thenby test, the factor affecting likeness degree of combined image isgot. The numerical value of the factor affecting likeness degree ofcombined image is the reciprocal of percentage of likeness degree.

In the test, first, find twenty persons of different characteristic andtake use of digital cameral to photograph their face. Then take use ofour computer drawing image system to combine the photographedimage and separately change the combined image’s eye, eyebrow,hair style, the cast of face, nose, mouth and ear, being from liketo unlike. Last, find fifteen to twenty persons to test and let themcompare the changed image with the original image, then get thepercentage of likeness degree .The selected persons who take part inthis test all have very good educational background and accomplish-ment; their observation ability, remembrance and recall ability arevery high and on the same level, which guarantees the representationcharacter and believeness degree.

From the result of the test, the following conclusion can be drawn.When the combined image’s eye is changed, the image’s character-istic changed quite remarkably. So the eye makes the quite deepimpression on person, that is to say, the eye is the organ that isremembered quite deeply by the witness. When the combined im-age’s hairstyle is changed, though the image’s characteristic hardlychanged, the hairstyle is the outer outline of the image and is deepin the person’s recall. When the cast of face is changed, the imagechanged the most remarkably. Because the cast of face occupiesmost of the image outline, it is the organ which is remembered themost deeply by the witness and it is also the most important whendrawing a image.

When the combined image’s eyebrow, nose, mouth or ear ischanged separately, the image’s characteristic hardly changed. Thenumerical value affecting likeness degree of combined image hasbeen got by taking use of the reciprocal of percentage of likenessdegree. The numerical value of the cast of face is1.69, the eye is1.54, the eyebrow is 1.45, the mouth is 1.41,the nose is 1.33,theear is 1.30,the hair style is 1.25. Therefore, the sequence of factoraffecting likeness degree of combined image can be easily got. Thesequence of factor affecting likeness degree of combined image hasbeen got in this paper. Namely, the first is the cast of face, the secondis the eye, the third is the eyebrow, the fourth is the mouth, the fifthis the nose, the sixth is the ear, the seventh is the hair style. Whendrawing an image, according to the above sequence, the efficiencyand quality can be improved.

Keywords: Combined Image, Likeness, Degree Factor

DIG-WO-01Review of Calculating a Vehicle’s Speed from CCTV Footage

*David Compton and David Hague, Forensic Science Service, 109Lambeth Road, London, United Kingdom.

The use of video camera surveillance systems has reached im-mense levels in the United Kingdom both in private and public areas(>3M cameras). When an incident has been captured on video, typi-cally the footage is replayed to gain an understanding of the event(s)or to identify a particular information. In the field of Road TrafficAccidents there is particular interest in calculating the speed of a carfrom time information recorded on images captured on video. In thispaper an overview of the correct procedure for estimating a vehicle’sspeed from surveillance footage is presented. The process for cal-culating speed of a vehicle from video footage has been employedin road traffic accidents for many years. The estimation is a simpletime distance calculation. However, on reviewing data produced bysome RTA investigators in the UK it was established that the methodthey have adopted for calculating the time element was inaccurate.Therefore, in this paper we present and justify a method for calcu-lating the time difference between two Closed Circuit Television.And demonstrate its validity with experimental data.

Keywords: Video, Speed, RTA


DIG-WO-02The Use of 3D Computer Models in Forensic Investigations

Bart Hoogeboom, *Jurrien Bijhold, Derk Vrijdag and MirelleGoos, Netherlands Forensic Institute, Volmerlaan 17, 2288 GDRijswijk, Netherlands.

The value of the use of 3D computer models in the forensicinvestigation of images is discussed. Examples of cases are given inwhich 3D computer models are used to get more information aboutthe position and/or size of an object seen in a picture.

Photographs and video footage are a 2D representation of the 3Dworld. One way to get more information from these 2D images is theuse of 3D computer models. With software like 3D Studio Max it ispossible to construct a 3D model and place a virtual camera. Thiscamera can be placed anywhere in the scene, so it’s also possibleto place a virtual camera at the location of the original camera thatshot the original image. The positioning of this camera can be doneautomatically by the software since it is provided with a camera-match algorithm that calculates the camera position and Field OfView on the basis of corresponding points in the original image andthe 3D model. By superimposing the original image over this virtualcamera view of the 3D model, it is now possible to place 3D objectsin the model and see them in the original image at the same time. Ifone of the degrees of freedom of the object in the original image isknown, the 3D object can be placed at the position of the object inthe original image. By this, the position and/or size of an object canbe determined.

The uncertainties in the extracted information using this methodare caused by:

Errors made in the 3D modelLens distortion of the camera that shot the original imageErrors made in position and FOV of the virtual cameraPositioning of the 3D object

The last two errors are not just caused by measuring errors, but arealso depending on the interpretation of the image by the investigator.For this reason, it is recommended that a measurement is alwaysrepeated by a second independent investigator.

An estimation of those errors can be done by making referenceimages with the original camera system. When taking these refer-ence images one should place objects of known size at a known placein the vicinity of the place of interest from the original image. Bymeasuring them later in the 3D model one can estimate the errors.To keep these errors small it is important to avoid using informationfrom outside the boundaries of the 3D model, since the propagationof errors outside this volume can be enormous.

This technique is used for height estimation in several cases of arobber from a robbery taped on a CCTV-recording. For the heightestimation a thin cylinder was placed at the position of the robber.The length of the cylinder is now an estimation of the height ofthe robber. One of the problems is that the pose of the robber isunknown. CCTV-recordings usually consist of only a few imagesper second. So, when a robber is walking/moving it is hard to tellwhich phase of his walking pattern is captured in an image. Justthis fact alone introduces an uncertainty of approx. 8 cm [1. DavidCompton et al.]. Together with the previously mentioned errors thereported uncertainty is approx. 11 cm. This makes this techniqueonly useful for excluding a suspect.

Another example of use of this technique is the estimation of thespeed of a motorcycle or a car. For speed estimation the averageframe rate and the variation in it should be measured. In time lapserecordings, the average frame rate and the variation in it depends onthe number of frames in the time span of the recording. Estimation ofthese parameters is done by making a copy of a real time recordingof a stopwatch with the time lapse recorder. In this copy, a numberof times the time interval in the number of frames that correspondsto the disputed recording is measured.

Further, we dealt with cases from CCTV-recordings of shooting

incidents in which one could not see the shooting itself, but theimpact of the bullet was visible. With the reconstruction of the bullettrajectory it was possible to point out the position of the shooter. Theadvantage of the use of a 3D computer model in these kinds of casesis that it can also be used as a visualization tool. Uncertainties in theestimation of the trajectory can also be visualized.

3D computer models can also be of help in the visualization ofcomplex incidents. By generating a sequence of images or evenmake an animation it is possible to visualize a scenario. Howeverprecaution should be taken because of the generated reality. It isvery hard to leave options open when you have shown an animation.

Finally, we have used 3D computer models as overlays on darkimages from a multiplex time lapse recording (images from dif-ferent camera’s in a changing order on one recording) as a tool toreconstruct what had happened during the recording of a burglary atnight.

Literature: David Compton, Clair Prance, Mark Shears andChristophe Champod, Systematic Approach to Height Interpreta-tion from Images, Proceedings of SPIE Vol. 4232 (2001)

Richard Medeck, Photogrammetric evaluation and computer ani-mation of a crime scene, Proceedings of the 6th International Rollei-Metric Police Workshop 2001

Keywords: The Use of 3D Computer Models, ForensicInvestigations

DIG-WO-03Novel Techniques for Image Enhancement of Faces and LicensePlates with 3-D Accurate Registration.

*Lenny Rudin, Pascal Monasse and Ping Yu, Cognitech., Inc.

We introduce a novel technique specifically designed to enhanceimages of human faces. This technique explicitly utilizes theoreti-cally justified 3-D shape and a 3-D motion models. We show that inorder to enhance image of a human face derived from a sequence ofvideo frames a new theory of a fully 3-D model-based registrationis required. It is also demonstrated that the technique of frame aver-aging is bound to fail to enhance facial images when the individualin the video is not stationary. We also show that our Frame Fusiontechnique applied to enhancing License Plates is physically basedon general 3-D motion of a planar structures and the related accurateplanar projective registration techniques

Keywords: 3-D model-based registration, Novel technique,Image Enhancement.

DIG-WO-04Interactive Forensic3D

*Vojin Mastruko, Forensic Institute in Zagreb, Croatia and DjaniBodlovic, Protech Ltd, Zagreb, Croatia.

Forensic3D is essentially application of different 3D technolo-gies on reconstruction of homicide cases. 3D approach in foren-sic field demand multilevel knowledge, which include engineeringskills within geodetic (either by total station or photogrametry),cartography, digital video editing, 3D modeling and finally 3D visu-alization on one hand and Forensic Ballistic Examination on otherhand. Different computer aided design (CAD) and Computer AidedModeling (CAM) software allows us to create 3D model of partic-ular crime scene. Depending on the software and hardware config-uration used we can observe the 3D model of crime scene insidethat software in different qualities. Normally on our monitor wecan have it in wireframe, flat-shaded, anisotropic, lambert, Blin,Phong, Radiosity, The stronger hardware configuration we have thebetter quality of the picture will be on our screen. If our scene istoo complex we will not be able to show it on our screen withoutsoftware rendering. Complex animations also cannot be shown onthe screen directly. Instead, we have to use software renderer and


typical output will be a bitmap (bmp, tiff, jpeg,..) for single framerenderings (still pictures) or if we render animations our output willbe in some video-format (typically avi file format with or withoutcompression). Analog output is not covered in this presentation. Ifwe speak in terms of animation we will have a big file showing howdifferent items on crime scene change in time and space. We callit animation. Animation was created by some specific 3D softwarewe use (AutoCad, Arhicad, Lightwave, 3D Studio Max, Houdini,Maya,...). To play the animation we do not need that original soft-ware, we need just player. There are a lot of different digital-videoplayers we can use and most of them are free (Windows MediaPlayer, Quick Time Player,...). With those players we can play ouranimations and this is a principle how we can transfer our animationfrom our computer we used to create animation to other computerswhere we want to show it. Advantage of this approach is that wecan transfer our animations from computer to computer and com-press it if necessary. Disadvantage is that we can only repeat thatanimation all over again. The animation we created was baked inform that we cannot change any more. We cannot change angle ofview of animated camera, we cannot dolly, pan or zoom the camera.We are passive spectators and we can only watch the animation. Ifwe want to be able to move inside 3D model of crime scene, watch3D models of a suspect and a victim play their roles, inspect everydetail on crime scene from every angle we desire and stay there aslong as we wish, than we need Interactive Forensic3D. This presen-tation will show differences between non-interactive and interactiveForensic3D technology. In order to get interactivity we must prepareour animations in different way. We have been experimenting withVRML and ShockWave3D technologies. VRML is an acronym forVirtual Reality Modeling Language and is open source technology.ShockWave3D is owned by Macromedia. Those technologies allowinteractivity to our animations. The same animations now have in-teractivity in a sense that we can move inside the scene in everydirection and interactively inspect every detail on the crime scene.

Several reconstruction of shooting cases will be discussed. Everyreconstruction of shooting case will be presented by use of Foren-sic3D animation and by use of Interactive Forensic3D technique.Advantages and disadvantages of those techniques will be discussedas well as possible future development and framework of Foren-sic3D techniques inside Forensic Institutes, taking into account dif-ferent advanced techniques like photogrametry, total station, cartog-raphy, digital video editing, 3D modeling, animation and simulation.

Keywords: Interactive Forensic3D technique, Crime Scene.

DIG-WO-05Police Photography into the 21st Century

*John Yearnshire, Centrex NTC, Harperley hall, Fir Tree, Crook,Co. Durham DL15 8Ds, United Kingdom.

Police Photography into the 21st Century now in this the 21stcentury technology is forcing the hand of the modern police serviceto change from traditional to digital photography. The traditionalimage captured by photography is an analogue image. The imagesare recorded as a variation in some physical property, for examplefrequency, amplitude, or activation state of a photochemical emul-sion (of colour dyes and silver halides, for example, as transparencyor a negative). This varies with the brightness, colour and contrastof the scene being recorded. A digital image however is a numericalrepresentation recorded simply as a series of binary digits (bits):either one or zero with no values in between. The image is capturedby being focused onto an electronic sensor (a charged coupled de-vice, CCD), which is made up of individual light sensitive elementscalled pixels (picture elements). These act as switches modifying anelectrical current on or off and a computer processes the informa-tion. Images may be either moving or still, the data can be storedon a variety of media (e.g. compact disc, computer hard disk, ordigital tape), and the image may be displayed either on an electronic

display (computer monitor, TV screen or computer projection) oras a hardcopy print. You may or may not be aware that for thelast two years, research involving/developing the evidential use of’digital images’ has been actively conducted by the Police Scien-tific and Development Branch. In conjunction with the Associationof Chief Police Officers, steering group headed by Paul Garvin aChief Constable in the United Kingdom. This research has nowbeen disseminated throughout the U.K. police forces (Via two CDs.The most recent in March 2002 together with a comprehensive buteasily understandable booklet titled ’Digital Imaging Procedure’.Good communication of information is most important to the futuresuccess of the system. The research focused on the integrity and ev-idential chain. The research was instigated after numerous enquiriesfrom police practitioners who realised that digital technology had arole in scientific evidence gathering. There were clearly many bene-fits such as instant photographic images and new digital processingprocedures such benefits were expected to be an important tool inpolice efforts to fight crime. Many forces were already dabblingwith this new technology. The main weakness was the perceptionthat it could not be used in the Criminal Justice System becauseof the fact that images could be manipulated or altered after cap-ture by the camera and these may lead to allegations of corruptionof images and subsequently loss of criminal cases at court. PoliceForces had no procedures, policy or codes of practice to supportthe use of digital images as evidence. Nor was there any credibledepth of knowledge as to what was best value equipment to pur-chase or appropriate quality of images. The images in use at thistime were gathered for, and focused on, intelligence purpose only.The images would never be scrutinised in a court of law. In 1998, theHouse of Lords Select Committee in London considered the issueof manipulation of images and ruled that the courts would have toaccept the integrity of the person taking the digital images. Theyemphasised to the Government Department with responsibility forpolicing (Home Office) that the responsibility of ’proving’ the relia-bility and authenticity of data, including digital imaging, falls to thebody carrying out the capture, processing and modification. Thusan audit trail, showing the continuity of evidence collection eitherelectronic or traditional paper based records or preferably both areessential. While any digital image can be considered for presenta-tion as evidence, the court would be well advised to place greaterweight on evidence, which can be rigorously demonstrated to havebeen derived from authenticated original images.

The procedure developed by the Police scientific and develop-ment branch for maintaining the integrity, capturing the image andcontinuity are as follows:

Preparation, Capture and Protection of the images and use.Preparation1 Obtain authority from the appropriate bodies.2 Start audit trail either electronic or paper base scene of crime job

sheet.3 Check digital operational equipment including batteries.Capture the images4 Take images DO NOT delete images.

a Capture still imagesb Captures video image

5 a Non removable medium cd-romb Reusable removable medium Flash cardc Reusable tape medium – mini dv datd Non removable mechanism e.g. Hard drive

Protection6 a Close the write once read many times medium

b Copy to write once read many times mediumc Activate write protect mechanismd Download to removable medium

5 a, b, c7 Retain master copy and produce a working copy8 Document and secure storage of the master copy9 Retain the master copy as an exhibit


10 Produce working copy images for court presentation11 Prepare prosecution file12 Prepare exhibit for court13 Retain exhibit for the statutory period14 Dispose of exhibits and complete audit trails.

To date this procedure has not been tested at law but it is believedprovides a good guide or framework.

Nevertheless the steering group commend it to forces and otherorganisations for adoption as current ’best practice’.

Some of the questions commonly asked about ’the use of digitalimaging.’ Will going digital mean deskilling in the police photogra-pher? No, in many respects, the process of capturing the image usinga digital camera will be similar to that for a conventional camera andthe same expertise is needed by the photographer. The main differ-ence is in the image storage medium in the camera, whether it is astill or video camera. In this respect, it is not so much a matter of ’de-skilling ’ but more one of retraining in the use of equipment, as thesame level of photographic knowledge will still be required. How-ever, the photographer using the equipment must be fully trained inits operation and in the use of down loading the images to a WORMdisc. (Write Once Read Many Times) For example Agfa E-Box sys-tem has been developed by Agfa to allow Images to be downloadedfrom the media onto a worm disc in seconds. Is the use of digitalimaging too difficult to comprehend, finance and achieve? In, gen-eral the answer to this question is ’No’. Whilst digital imaging doesincorporate new technology, through education, training, guidanceand awareness the principles and operation of digital photographycan easily become the norm. The initial outlay for digital equipmentis quite high, however with the advances in modern technology theprices will come down. To gain best value, it is probably worth con-verting to digital photography gradually. Will there be resistance tothe introduction of new technology from the Criminal Justice Sys-tem? Should this occur, I believe any problems can be minimised bydemonstration, explanation, training and awareness in the relevantareas, soon, like most new innovations in scientific evidence gather-ing, it is expected that, there will be a general acceptance of the newtechnology.

Keywords: Photography, Integrity, Training

DIG-SO-01Facial Image Comparison Using a 3D Laser Scanning System

Arnout Ruifrok, Mirelle Goos, Bart Hoogeboom, Derk Vrijdag and*Jurrien Bijhold, Netherlands Forensic Institute, Volmerlaan 17,2288 GD Rijswijk, Netherlands.

The face is still the ’biometric of first choice’ in a lot of verifica-tion and identification situations, making reliance on facial recogni-tion, and facial comparison, more and more common. The increasein traveller movement and number of documents to be compared,and the increasing use of safety-cameras, still cause an increasingreliance on facial comparison for forensic applications.

To reliably perform comparisons of facial images, it is importantto position the head corresponding to the facial images available.Several techniques have been proposed to solve the problem of dif-ferent positioning of the head and camera in reference and disputedpictures (1,2). Techniques using three or more landmark points onthe face have been proposed for matching the face and camera posi-tions to the available photographs (1). The use of ’structured light’as a form of one-shot 3D photography, used in combination withdedicated software, has been proposed to refine the results of three-point analysis of 3D reconstruction (2). However, these methods canbe cumbersome, and require the cooperation of the subject.

3-D photography using laser-scanning technology, together with3-D modelling software, offer the possibility of flexible and repro-ducible positioning of the head of a person corresponding to theface and camera position of the 2-D facial comparison images. Be-side that, the use of 3-D laser scanner data for facial comparison

offers the possibility of mathematical matching of reference anddisputed material.

We will present our experiences with a non-contact 3-D laser-scanning system (Minolta VI-900), especially with respect to ease-of-use, reproducabilty, and performance for facial comparison ap-plications.

References: Maat, G.J.R. The positioning and magnification offaces and skulls for photographic superimposition. Forensic ScienceInternational, 41: 225-235, 1989.

Heuvel, H. van den. The positioning of persons or skulls forphoto comparison using three-point analysis and one-shot 3-D pho-tographs. Proc of SPIE, 3576:203-215, 1998.

Keywords: Facial Image, 3D Laser Scanning System

DIG-SO-02Comparison of Clothing to Video Images

*Johanna Morley, The Forensic Science Service, 109 Lambeth Rd.,London SE1 7LP, United Kingdom.

See FIB-TO-07

DIG-SO-03Semi-Automatic Jigsaw Puzzle Reconstruction of FragmentedDocuments

*Patrick De Smet, Johan De Bock and Els Corluy, GhentUniversity, Dep. TELIN/TW07, Sint-Pietersnieuwstraat 41, B-9000Gent, Belgium.

Automatic jigsaw puzzle reconstruction is an interesting toy prob-lem that is traditionally studied in the area of computer vision re-search. Recently, this topic has been receiving some additional at-tention from two major areas of applications, i.e., archeology andforensic investigation. In this paper we discuss several computervision techniques and tools that can be used to assist forensic inves-tigators in recomposing torn or ripped-up documents. More specifi-cally, we briefly discuss a framework of techniques that enables therealization of each required processing step, i.e., image and fragmentacquisition and segmentation, fragment matching and repositioning,and global document recomposition. However, we do not only dis-cuss the required image processing algorithms, but we also studythe interactive GUI tools that were developed in order to assist orcorrect the automated processes involved in the proposed recompo-sition framework. Our research demonstrates the functionalities ofthe proposed framework for several real examples of ripped-up doc-uments, and also discusses the obtained results, techniques and toolsthat could benefit from further research and development. More in-formation is also available at: http://telin.rug.ac.be/ipi/foco/ The re-sults reported in this work are based upon an image acquisition andsegmentation step that uses a flatbed scanner set-up and several GUIcomponents. For the image segmentation step the GUI tools offera so-called chroma-keying region growing method [1], gradient-based watershed image segmentation [2], and marker-based imagesegmentation using a graph-based watershed post-processing algo-rithm [3]. Each of these components can be set up to operate eitherfully automatically or with required user-interaction and verifica-tion of the segmentation results. The most important functionalitiesof the interactive tools consist of the use of, e.g., click-and-mergeoperations, rejection or selection of certain results, tools for paintingregion growing and marker areas on top of the image, adjustmentof algorithmic parameters, etc. These tools allow us to obtain veryaccurate fragment delineations and result in a list of segmentedfragments. Next, we use a contour matching technique in orderto match and reassemble the segmented fragments. This requiresbuilding lists of optimal local contour correspondences that are sub-sequently used to arrive at a global optimization and repositioningof all the fragments. The "curvature" feature vectors are built by


tracing around each fragment using a chain code contour follow-ing algorithm, and the shape of the fragment is approximated andannotated using linear polygon interpolation and resampling. Then,we apply our global optimization procedure, which is (currently)based on a technique reported earlier by Bunke and Kaufmann [4].While this procedure is computationally expensive due to its ex-haustive best-first combinatorial search, it yields the possibility offinding global solutions even in situations where several fragmentsare delineated by locally almost identical contour descriptions. Also,we adapted and extended the basic algorithm in order to allow usto keep track of lower-ranked, i.e., higher cost, global solutions.These ordered results could then be browsed through quickly. Addi-tionally, we include the possibility for progressively building globalreconstructions, i.e., the user can arrive at a global solution by auto-matically but iteratively recomposing smaller multi-fragment piecesinto larger fragments. The translational and rotational reposition-ing information for each fragment is first computed by calculatinga set of raw positioning parameters from the contour correspon-dence matching results. Subsequently, the local neighborhood ofthese parameters (in the proper parameter space) is evaluated in or-der to determine if better positioning results can be obtained. This isdone by also computing actual fragment overlap costs and remainingcontour fitting gaps. Again, each of the above components allowscost-function results to be presented to the user and the user canalso assist and/or correct the obtained results. Generally speaking,the obtained experimental results indicate that while all these toolsyield very interesting and valuable tools for forensic document re-composition, considerable room for further research still remains; inaddition to further optimization, several interesting possibilities forfuture research could include, e.g., detection of cross-fragment linesand figures in the document, layout and paper type issues, OCR andtext line detection and orientation, etc.

References: [1] P. De Smet, J. De Bock, E. Corluy, "Computervision techniques for semi-automatic reconstruction of ripped-updocuments", submitted to SPIE AeroSense 2003, Investigative Im-age Processing III. [2] P. De Smet, R. Pires, D. De Vleeschauwer,I. Bruyland, "Activity Driven Non-linear Diffusion for Color Im-age Watershed Segmentation", Journal of Electronic Imaging, SPIE,vol. 8, no. 3, pp. 270-278, July 1999 [3] P. De Smet, "Segmen-tation and analysis of digital image sequences", PhD dissertation,Ghent University, 2002. [4] H. Bunke, G. Kaufmann, "Jigsaw puz-zle solving using approximate string matching and best-first search",Lecture Notes in Computer Science, Computer Analysis of Imagesand Patterns, D. Chetverikov, W. G. Kropatsch (Eds.), pp. 299-308,Springer-Verlag.

Keywords: Reconstruction fragmented documents, Computervision

DIG-SO-04Restoration of Corrupt JPEG Data

Stefan Ott and *Bernd Rieger, Bundeskriminalamt, Section KI 22,D-65173 Wiesbaden, Germany.

Viewing the undamaged part of a partially destroyed image is atask, the success of which, among other things, is highly dependenton the nature of the image information carrier. To look at the remain-ing portions of a cut up or scratched paper photograph, for instance,does not usually pose any real problems. One can often restore partsof a disturbed digital image as long as the image information isstored uncompressed and is pixel orientated. The restoration tasktends to become complicated as soon as digital image informationis stored compressed without explicitly referring to pixel data. Thelatter situation especially holds true for the widespread JPEG fileformat.

Current preliminary procedures involving corrupted JPEG files,which have been sent to the Bundeskriminalamt’s image enhance-ment service, have forced us to examine the situation more closely.

As there was no commercially available program on the market, theBundeskriminalamt decided to develop their own software tool forthe analysis and restoration of partially damaged JPEG files. Thistool, named JLab, is now available for police applications and willbe presented here.

The paper will first give a brief introduction into the area of digitalimage coding, paying special attention to some JPEG algorithm andfile format details. A summary of practical problems concerning thehandling of digital images in connection with police investigationswill also be listed, including hints and tips relating to such aspectsas backup, authentication, viewer-related problems, long-term docu-mentation, scalable quality, correctness and completeness, steganog-raphy, effects of faulty data, data loss and the analysis of large datablocks with an unknown structure. Emphasis will be placed on ex-plaining the functions and features of the JLab tool.

The JPEG standard consists of an entire series of different com-pression procedures for diverse applications. This is a very com-plex matter, and so it was initially necessary to find a solution fora limited area that would nevertheless be of great practical signifi-cance. JLab is, therefore, an analysis and reconstruction program forJPEG/JFIF files (JPEG File Interchange Format) with DCT (DiscreteCosine Transform) and Huffman coding. This type of compressionis by far the most commonly used at present. The program runs oncomputers with the MS Windows operating system. JLab combinesthe "viewer" and the "hexeditor" functions. In three sub-windowsit shows three views of a JPEG data stream: a structural view, animage view, and a hexadecimal view with a simultaneous ASCIIinterpretation. The three views are linked so that the user can makea dedicated analysis of structural elements and image areas.

The program can also handle several JPEG data streams in asingle data file, such as when one large JPEG image contains asmall thumbnail image, which is also in the JPEG format. Thisis frequently the case with images produced by the well-knownPhotoshop program. Since JLab provides a detailed representation ofthe complete structure of a data file, one is able to draw conclusionsabout the origins of the image from both the comments and thoseinputs, normally not displayed, that are specific to the application.Even if other viewer programs do not accept a JPEG file, JLab canat least recognize whether it has any structures in it, which conformto JPEG.

The quantization tables and the Huffman tables are among themost important parameters of the compression procedure. Insteadof faulty or missing tables, standard tables and tables from correctsample files may be used. Databases from different tables can easilybe created and extended. Tables that are currently being analyzedcan be compared with those from the database on the basis of a briefcharacterization, so that similar tables can be found quickly. Manualrepairs of tables and the restoration of other destroyed marker inputsrequire an exact knowledge of the JPEG standard.

The more theoretical description of the underlying principles ofJLab will be supplemented by a live presentation showing test im-ages and also some examples of practical casework.

Besides the fields where JLab provides successful support, thelimitations must also be considered. There is, for instance, absolutelyno chance of performing repairs when faulty data covers the entireimage. If, for example, the data contained on one of the four discsof a RAID system is lost (that is, 2 bits from each byte, withoutredundancy), the remaining 6 bits can no longer be used to showpictorial information. This contrasts with image data that is notcompressed and which can, at least in part, be made visible. Alsothe loss of major parts of a Huffman table can generally not becompensated, because such a table is frequently characteristic of animage, and other tables do not produce anything that could be useful.

The next version of JLab, which might already be available by thetime the conference takes place, will considerably extend the capa-bilities of this program to provide support in the repair of Huffmantables and other structures. Suggestions on the further developmentof the program’s functions are highly welcome.


Keywords: Data corruption, Restoration of JPEG files, Digitalimage storage

DIG-SO-05Identification of Gun Shot Injury by Digital Imaging

*A.K. Pal and N.P. Waghmare, Central Forensic ScienceLaboratory, 30, Gorachand Road, Kolkata, 700014, India; M.S.Rao, Directorate of Forensic Science, Ministry of Home Affairs,New Delhi, India; A. Manna, Jadavpur University, Kolkata,700032, India.

The firearm injuries play very vital role in the commission ofcrime to ascertain the distance, angle, direction of firing etc. bycomparing the injury patterns. In most shooting incident firearms& ammunitions may not be received from scene of crime becausethey were deliberately thrown away or hidden by the criminals. Insuch cases except the gun shot injury on the body of victim no otherrelated clue material found at the crime spot. In such cases the In-vestigating Officers invariably want information relating to distance,angle, direction of firing from the gun shot injury. Photographs ofgun shot wound pattern may give definite opinion whether the giveninjury is gun shot wound, distance of firing, angle / direction of fir-ing, type of firearms and ammunitions were used for specific crimespot. The forensic Scientists make manual search in books exclu-sively dealing with wound patterns for comparison. This is a timeconsuming process. For speedy searching authors have developed asemi-automated computerized retrieval system to create and updatedata bank with the help of published data from books and Journalson wound Ballistics and also to access the required information tothe suspected injury after comparing the injury patterns from thedata bank in a short time as well as at a low cost by using image-processing technique. Present study can be utilized for speedy casesdisposal relating to firearms injuries, forensic medicines and ballis-tics experts professionals.

Keywords: Digital Photography, Computerized Data RetrievalSystem, Firearm Expert

DNA Analysis

DNA-TP-01Turkish Population Data for the 15 STR Loci (ContainingCodis Multiplex) of the AmpFISTR Identifiler Multiplex Kits

*Faruk Asıcıoglu, Bestami Colak, Fatih Akyuz, OmerMuslumanoglu, Department of Biology, The Council of ForensicMedicine, Cerrahpasa, 34300, Istanbul, Turkey and Ugur Ozbek,Department of Biology, The Council of Forensic Medicine,Cerrahpasa, 34300, Istanbul, Turkey and Institute for ExperimentalMedicine (DETAE), University of Istanbul, Turkey.

Short tandem repeat (STR) loci are highly represented in thehuman genome and, most of them considered very useful for in-dividual identification and paternity testing. Increasing number ofmultiplex PCR systems have been available. The results of STRstudies in different populations may show significant differences inallele frequencies. A population study for 15 tetrameric short tandemrepeat (STR) loci containing CODIS System: CSF1PO, D18S51,D3S1358, D21S11, D5S818, FGA, D7S820, HUMTH01, D8S1179,TPOX, D13S317, VWA, D16S539 and D2S1338, D19S433 was per-formed on 250 unrelated Turkish blood samples. Our donors weresampled from all different regions of Turkey. The DNA was ex-tracted by the salting out method according to the procedure de-scribed by Miller et al., and Chelex extraction was done for blood-stain. The quantity of extracted DNA was estimated using ethidiumbromide stained agarose yield gels and measured by spectropho-tometer. Allele and genotype frequencies for the 15 loci were de-

termined by PCR using the manufacturer’s recommended protocol,and using the commercially available AmpFlSTR Identifiler multi-plex kits. PCR amplification was carried out in a GeneAmp PCRSystem 2700 (PE Applied Biosystems, USA). Typing was doneby ABI Prism 3100 Genetic Analyzer equipped with 16 capillariesand Performance Optimized Polymer 4 (POP-4) with 28 min runtimes and run conditions as followed the manufacturers recommen-dations. The allele frequencies in Turkish population computed, het-erozygosity rate, power of exclusion, power of discrimination, poly-morphism information content, matching probability was calculated(Table 1). All 15 loci met Hardy-Weinberg equilibrium expectations.D2S1338 locus displays highest power of discrimination (PD) valueand TPOX, the lowest (0.971 and 0.793 respectively). Present studygives the data of highest number of individuals studied for these lociever published on Turkish population. In conclusion, the allele fre-quency data can be used in identity testing to estimate the frequencyof a multiple PCR-based profile in the Turkish population.

Keywords: STR, Turkish Population, CODIS

DNA-TP-02DNA Fingerprinting of Cannabis Sativa L. Accessions UsingRAPD and AFLP Markers

*Erdogan E. Hakki, Selcuk University, Faculty of Agriculture, FieldCrops Department, 42031 Kampus, Konya, Turkey; Elif Uz, BilkentUniversity, Molecular Biology and Genetics Department, Ankara,Turkey; Ayla Sag, Sevil Atasoy, Istanbul University, Institute ofForensic Sciences, Istanbul, Turkey and Mahinur S. Akkaya, MiddleEast Technical University, Department of Chemistry, Ankara,TR-06531, Turkey.

Climatic and geographical conditions in Turkey favor the naturalgrowth of Cannabis sativa L. (hemp plant). While hemp is a goodcandidate, for producing seed oil, feed, fiber for textile and pulpindustries and for therapeutic purposes, the main reason of its plan-tation is more for its use as narcotics. The most psychoactive naturalcannabinoid is D9-tetrahydrocannabinol (THC), which varies dra-matically within C. sativa varieties, also affected by environmentalconditions. DNA fingerprinting of the Cannabis samples will bevaluable in facilitating the identification of its origin, often neededsince the cultivation of hemp is strictly controlled in many countries.Morphological, chemical and biochemical studies do not result insufficient information to identify the origin of a seized sample, how-ever, DNA fingerprinting using molecular markers is not only verypowerful but also relatively an easy technique for genotype iden-tification. In this study, Randomly Amplified Polymorphic DNAs(RAPDs) and Amplified Fragment Length Polymorphisms (AFLP),genotypic markers not affected by environment like that of the phe-notype, were used to fingerprint the 18 different individuals from fivedifferent locations representing 3 geographical regions of Turkey.

Keywords: Cannabis Sativa, DNA Fingerprinting, ForensicPlant Identification

DNA-TP-03Modified Method of Differential Lysis for Mixed Biomaterialsafter Sexual Assaults

Abror Ikramov, Sh. Nurmatov, D. Akhmedova, A. Nuritdinov,Republican Scientific Research Criminalistic Centre after H.Sulaymonova, Ministry of Justice. 29, Chilanzarskaya Street,Tashkent City, 700115, Republic of Uzbekistan; *RustamMukhamedov, Institute of Genetics and Experimental Biology ofPlants Academy of Sciences, Republic of Uzbekistan and S.Atakhodjaev, Republican Scientific Research Criminalistic Centreafter H. Sulaymonova, Ministry of Justice. 29, ChilanzarskayaStreet, Tashkent City, 700115, Republic of Uzbekistan.

Detection of male specific DNA from evidence with sperm cells