SPIE Proceedings [SPIE SPIE Optical Metrology - Munich, Germany (Monday 23 May 2011)] O3A: Optics...

Hylemetry versus Biometry: a new method to certificate the lithography authenticity

Giuseppe Schirripa Spagnolo, Lorenzo Cozzella, Carla Simonetti Dipartimento di Ingegneria Elettronica – Università Degli Studi Roma Tre

Via della Vasca Navale, 84, 00146 Roma (Italy)

ABSTRACT

When we buy an artwork object a certificate of authenticity contain specific details about the artwork. Unfortunately, these certificates are often exchanged between similar artworks: the same document is supplied by the seller to certificate the originality. In this way the buyer will have a copy of an original certificate to attest that the “not original artwork” is an original one. A solution for this problem would be to insert a system that links together the certificate and a specific artwork. To do this it is necessary, for a single artwork, to find unique, unrepeatable, and unchangeable characteristics. In this paper we propose a new lithography certification based on the color spots distribution, which compose the lithography itself.

Due to the high resolution acquisition media available today, it is possible using analysis method typical of speckle metrology. In particular, in verification phase it is only necessary acquiring the same portion of lithography, extracting the verification information, using the private key to obtain the same information from the certificate and confronting the two information using a comparison threshold. Due to the possible rotation and translation it is applied image correlation solutions, used in speckle metrology, to determine translation and rotation error and correct allow to verifying extracted and acquired images in the best situation, for granting correct originality verification.

Keywords Biometry, Artworks Authentication, Speckle Metrology, Digital Image Processing, Lithography.

1. INTRODUCTION The main problem when we buy an artwork object consists in getting a certificate of authenticity, in particular for the artwork bought through a seller and not at first hand from artist. There is a tremendous abuse in the “certificate of authenticity” business, because unless a certificate of authenticity originates from and is signed by either the artist, the publisher of the art (in the case of limited editions), a confirmed dealer or agent of the artist (not a third party or reseller), or an acknowledged expert on the artist, that certificate is pretty much meaningless. A legitimate one must contain specific details about the artwork such as when and how it was produced, the names of people or companies involved in its production, dimensions, and the names of reference books or similar resources that contain either specific or related information about either that work of art and/or the artist. It should also state the qualifications and full contact information of the individual or entity that authored the certificate, and include his or her complete and current contact information.

Unfortunately, often these certificates are exchanges between similar artworks: the same document is supplied by the seller to certificate the originality. In this way the buyer will have a copy of an original certificate to attest that the “not original artwork” is an original one.

Certificates of authenticity are often problematic; many are just worthless. Unfortunately, most people believe that art with a certificate is automatically genuine, but that’s not even close to truth. It happens because no law govern who is or is not qualified to write certificates of authenticity, or what types of statement, information or documentation a certificate of authenticity must include. In other words, anyone can write a certificate whether they are qualified or not. As if that is not bad enough, unscrupulous sellers forge official looking certificates of authenticity and use them to either sell outright fakes or to misrepresent existing works of art as being more important or valuable than they actually are. A possible

O3A: Optics for Arts, Architecture, and Archaeology III, edited by Luca Pezzati, Renzo Salimbeni, Proc. of SPIE Vol. 8084, 80840S · © 2011 SPIE · CCC code: 0277-786X/11/$18 · doi: 10.1117/12.889387

Proc. of SPIE Vol. 8084 80840S-1

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 09/21/2013 Terms of Use: http://spiedl.org/terms

fraud can be put the following way into effect. An art merchant, starting from an original lithography and its certificate of authenticity, duplicates both the lithography and the certificate of originality.

A solution for this problem would be to insert a system that links together the certificate and a specific artwork. In this way the exchange will not be possible and the buyer will be able to verify the originality by himself. To do this it is necessary, for a single artwork, to find unique, unrepeatable, and unchangeable characteristics. If these characteristics are present, we have the possibility of to identify the artwork and to distinguish it from the other.

Actually, this procedure is used by biometrics authentication systems. Biometric identification relies on physical characteristics that are unique to each person to ascertain the identification of an individual. The most commonly known method of biometric identification is fingerprint, DNA iris scans, hand geometry, facial feature, and voice.

The term Biometrics derives from two Greek words: βιος that means “life” and μετρον “measurement”; so it is possible to literally translate the word as “life measurement”.

Since the biometric identification has given excellent results, it comes spontaneous to apply similar criterions for unique identification of “nonliving matters”, as for instance the artworks.

Obviously the identification will be possible if, for a specific work of art, we can choose, as for the biometrics systems, unique characteristics, unrepeatable and unchangeable ones. In other words, if for an object we are skilled to find unique characteristics and not repeatable ones we have the possibility to implement a system of recognition similar to the biometric one. As for biometric authentication, it is possible to coin a new word to identify this new method of authentication of nonliving matters. As for Aristotle the term υλη (hyle) means nonliving matters, the hylemetric authentication may identify the procedure for the identification of an inanimate object.

In theory, every casual and non-reproducible characteristic could be used in hylemetric identification. However they have to have the following desirable properties:

(i) uniqueness, which indicates that no two objects should be the same in terms of the characteristic,

(ii) permanence, which means that the characteristic should be invariant with time,

(iii) collectability, which indicates that the characteristic can be measured quantitatively,

(iv) non-invasive, which indicates that the characteristics can be measured without modifying anything.

Choosing the opportune characteristic, such identification can be applied to many types of artworks object.

The most counterfeit artworks are the lithography. Therefore, in this paper we will implement the hylemetric identification for authentication of ones.

2. STONE LITHOGRAPHY The technique of stone lithography (from the Greek word for "stone drawing") relies on "the principle of the antipathy of grease and water." Generally, the stone on which the image is initially created is limestone. The image is drawn on the stone with some greasy material. After the image is drawn, the stone is dampened and ink is applied with a roller. The greasy image repels the water and holds the oily ink while the rest of the stone surface does the opposite. The stone is chemically treated after the image is created in order to enhance this effect.

The finished stone is placed on a bed that carries it through the press. The paper is placed on top of the stone with some backing papers to protect it. A sheet of metal or plastic is placed on top of all the materials and they are braced together. A roller underneath that is turned by a handle that moves the bed. This is similar to the intaglio press except that a scraper bar instead of a roller applies the pressure from above. The scraper bar slides along the greased metal plate pressing the paper against the stone so that it lifts the ink from where the greasy drawing material holds it on the stone.

Color lithography is a more complex process that usually involves multiple pressings, one for each color in the image. This requires an extensive knowledge of color theory because the process requires the mixing of colors on the final image itself. According to a book on lithography technique, the original color drawing should be treated as a guide for the final print, not as a finished work to be duplicated exactly. Different stones are sometimes used for each color but the



same stone can be used for multiple colors. Because of the equipment used and the knowledge and skill required for the printing process, lithography lends itself to collaboration between an artist and a printer.

Working with a skilled printer (or being the artist himself a skilled printer) the artist can realize exactly how to achieve certain effects and how the print can benefit most from the chosen medium. In the second half of the 19th century, Edgar Degas and Édouard Manet worked in lithography, and Odilon Redon made it his principal means of expression. Besides, Degas had to say "If only Rembrandt had been familiar with Lithography what he would have achieved".

Although lithography after World War II was generally considered a commercial medium, in reality, the Lithography is an important artistic medium. Figure 1 shows a Stone Lithography realized with traditional method [1] by artist Giovanni Job.

Figure 1 – Stone lithography

3. UNIQUE CHARACTERISTICS DETERMINATION To certify a lithography authenticity by means of hylemetric identification it is necessary to find a unique, non-reproducible and immutable characteristic. For instance, we could use the distribution of colorful “stains”, which composes a feature of the image. In this way the method exploits ownership of lithography, because it is made using limestone’ porosity. The image is drawn on the print plate’ surface with an oil-based media (hydrophobic). The range of oil-based media is endless, but the dexterity of the image relies on the lipid content of the material being used; its ability to withstand water and acid. Following the placement of the image, it is applied an acid emulsified with Arabic gum. The function of this emulsion is to create a salt layer directly around the image area. The salt layer seeps into the pores of the stone, completely enveloping the original image. This process is called etching. Using lithographic turpentine, the printer



removes the greasy drawing material, leaving only the salt layer; it is this salt layer, which holds the skeleton of the image's original form. When printing the stone or plate is kept wet with water. Naturally the water is attracted to the layer of salt created by the acid wash. Ink that bears high lipid content is then rolled over the surface. The water repels the grease in the ink and the only place for it to go is the cavity left by the original drawing material. When the cavity is sufficiently full, the stone and paper are run through a press which applies even pressure over the surface, transferring the ink to the paper and off the stone. Every stone has different distribution of pores and it shines through the print. This is to distinguish lithographs of dissimilar runs, because the porosity changes with changing the stone; but if we have the same run, the differences between each print are in the corrosion of the stone or in the various piles of color that is deposited on the paper when the print is made. So the methods that we use to distinguish the lithography are the same because also in this case there will be different distributions of colorful "stains".

The idea consists to choose from the artwork a nuance, for instance the muzzle of the dog, and use it to identify the lithography. The portion of lithography is acquired by a digital high definition camera and it is on the certificate printed (in low resolution – without the details present on the lithographic press) to help the user to find what the right portion of the work is he has to analyze for the authenticity verification. Of course this portion is a copy of the original but we use it only for the localization; in fact after this step we drown from the original the same portion greater than last one and we match each other to be sure that it is the right zone; after we will derive a unique template that describes only the original lithography.

Due to possible rotation, translation, distortion and scale errors, which can lead to a misinterpretation, is necessary indicates on the image attached to the originality paper also the trust points (from two to six) to be used for correcting any possible image distortion.

The lithography zone under analysis is put on the authentication certificate, with the indication of trust points, as introduced before. In verification phase the verifier has to acquire a similar image zone and correct any possible distortion and acquisition error before correlate the two images for verifying the lithography authenticity.

Figure 2 – In (a) it is highlighted the area used for certification image (b) and reference image (c).

On certification and reference image are also highlighted the used trusted points.



With reference to Figure 2, we consider image shown in Figure 2(b) as the certification one and image shown in Figure 2(b) as acquired reference one.

The first step in the process is correct, as much as possible, any acquisition distortion, which can lead to a misidentification of original certificates. In fact, when acquired with a digital camera, the reference image A can contain a bigger portion of image around the same zone, can have some roto-translation on it and can be affected by some acquisition distortion, due to lens system used. After that, could also be possible having additional noise on it, due to not perfect cleaning on camera lens. The proposed method is able to correct all these problems and results robust to added noise of various origin (e.g. Gaussian, salt and pepper, median, etc.).

Due to the absence of standard reference point inside lithography, it is necessary define a number of trust points, used to correct transform the acquired reference image, with the aim to be the most possible similar to the certificate ones. In this way the final verification, due by means of threshold correlation methods, will be maximised.

We can call the reference image RI and the image presents on the certificate as CI . Due to the manual operations

necessary to acquiring the image RI , this can have different aspect ratio, can be centred in a different point, can be rotated or can have some noise or distortion due to acquiring media. Using a correct image transformation (e.g. affine, polynomial, reflective, etc.), based on the identification of a set of trust points, it’s possible reducing this difference between the two images and verifying if corresponding to the same lithography.

The approach used in this paper for correcting all the possible misalignments between RI and CI is based on choosing a correct image transformation, as introduced before. This choice is based on the fact that is very difficult correct non centred roto-translations without the definition of trust points. The usage of image transformation, such as affine or polynomial ones, allows to correctly identifying a translation along x and y axis, as well as a combined rotation on an angle α of the first image respect to the second one.

As described in [2], is possible correcting a great amount of errors selecting Trust points on the two figures. Trust points are points present in both the two images, which identifying clearly the same real point (e.g. the dog elbow).

The Table 1 reports a set of classical transformation with the related minimum numbers of trust points, with the related set of correctable distortions.

In this paper we have used affine, similarity and polynomial (second order) transformation. The selection of the transformation was based on the number of trusted point inserted by the verifier during verification phase. If the verifier defines only two trust points the proposed system applying a non-reflective similarity transformation for correcting only rotation and translation distortions. Otherwise, if are selected a different amount of trust points, a different transformation is applied, for correcting also other distortion.

After having aligned the two images, is possible verifying if the two are extracted from the same lithography. In fact the image transformation evaluating only geometrical differences among the two images. If we have two different lithography having the same subject, the two images are at first sight similar, and after correcting geometrical distortions, are pretty the same image. The presence of a hylemetric characteristic such the colour pattern previously described, having a randomly distribution, different from a lithography to another one, allows using a correlation approach, to determine if the two images are extracted from the same lithography or not.

The proposed approach is based on normalised correlation coefficient C a [3]:

*

1*

'

'

( ) ( )( , )

( ) ( )C R

C R

I F IC x y

I F Iα−⎡ ⎤⎢ ⎥Δ Δ =⎢ ⎥⎣ ⎦

FF

F (1)

Where ( , )x yΔ Δ are the correlation peak coordinates and F and 1−F are forward and backward Fourier transform

operators, respectively, and * means the complex conjugate. 'RI is the reference image after geometrical image transformation.



Due to the impossibility to obtain the same image to correlate, it is also defined a verification threshold, T a , so that:

false lithograpjytrue litography

C TC Tα α

α α

<⎧⎨ ≥⎩

(2)

The selection of the appropriate threshold is based on the minimization of False Acceptance Rate, such as the percentage of false lithography recognized as true; respect the total amount of verification tests.

Table 1 – Type of transformation in relation to distortions.

Transformation Type Min Trust Points Description

Non Reflective Similarity 2 pairs Use this transformation when shapes in the input image are unchanged, but the image is distorted by some combination of translation, rotation, and scaling. Straight lines remain straight, and parallel lines are still parallel.

Similarity 3 pairs Same as 'nonreflective similarity' with the addition of optional reflection.

Affine 3 pairs Use this transformation when shapes in the input image exhibit shearing. Straight lines remain straight, and parallel lines remain parallel, but rectangles become parallelograms.

Projective 4 pairs Use this transformation when the scene appears tilted. Straight lines remain straight, but parallel lines converge toward vanishing points that might or might not fall within the image.

Polynomial 6 pairs (second order)

10 pairs (third order)

15 pairs (fourth order)

Use this transformation when objects in the image are curved. The higher the order of the polynomial, the better the fit, but the result can contain more curves than the base image.

Picewise Linear 4 pairs Use this transformation when parts of the image appear distorted differently.

Local Weighted Mean 12 pairs Use this transformation, when the distortion varies locally and piecewise linear is not sufficient.

4. TEST RESULTS In this section are reported some experiments made on real lithography. For made them we have create a dedicated Matlab© program, which allows the verifier to select verification zones among the two images, to input the trust points, having also a help system for selecting the same points. This program is also able to add controlled noise on the reference image for testing noise strength with different noise parameters.



Figure 3 – Certificate Image (a) and Reference Image (b)

In Figure 3(a) is shown a peace of lithography reported in Figure 1 and Figure 2(b), that could be used inside the originality certificate. In Figure 3(b) it is reported a similar zone, acquired during verification phase, which represents the reference image. It can be easily viewed as the manual acquisition of the reference image leads to have the introduction of scale, area, rotation and translations respect to the certificate image. In Figure 4 Gaussian noise is added to reference image, which degrades the resulting value of C a , but still over the threshold Ta . Has to be noted that the reference image is also clockwise rotated and translated both horizontally and vertically.

Figure 4 – Example of verification result with Gaussian Noise addition. (a) Certificate Image; (b) Reference Image

with Gaussian Noise Added; (c) Correlation result after Reference Image geometrical correction.

The Table 2 reports the added noise, the related used noise parameters and the resulting statistical threshold Ta and

related correlation value C a . For each added noise are also reported the used noise parameters, for underlying the effects on correlation of different noise cases.



Table 2 – Correlation values for different added noise, with related noise parameters.

Noise Type Noise Parameters Threshold Value Correlation Value

Gaussian Mean Value equal to 0.10

Variance equal to 0.10 3.86 23.99


Variance equal 0.15 3.76 22.45


Variance equal to 0.10 3.84 23.73


Variance equal 0.15 3.73 22.13

Salt & Pepper Distribution equal 0.1 4.06 26.88



Poisson No parameter to be defined 4.23 30.40

Speckle Distribution equal to 1.0 3.49 19.22



It could be noted that the used threshold varies in any experiment. This is due to the fact that we have used an adaptive statistical threshold, based on statistical characteristics of the correlation function. The threshold T a used in this article is:

3 * CTa s= (3)

Where Cs is the standard deviation of the correlation function C a . It has to be noted that is possible define alternative statistical thresholds. We have used in our experiments also different thresholds, with similar results, such as three timers the mean value plus standard deviation, three times the standard deviation plus mean value and so on.

Starting from the previous Figure 2, it is possible synthetizing the process to verify lithography authenticity starting from two different images. The first one is the Certificate one; the second is the Reference one, acquired at verification phase. This last one has some geometrical distortions due to manual acquisition of the interested area. In particular can be easily view rotation and translation errors. In particular reference image is rotated of 12 degree clockwise respect the certificate one, with a horizontal translation of 34 pixels and a vertical one of 54 pixels.

Applying a second order polynomial correction, by means of a selection of two trust points, the result is shown in Figure 5, with a resulting correlation value after correction 20.46Cα = (with a threshold equal to 3.79Tα = ). It has to be noted that the value before correction is sensible minor ( 1.75Cα = with 1.38Tα = ), but still over statistical threshold, due to the accuracy used in acquisition phase.



Figure 5 – (a) Reference Image rotate of 12 degree clockwise; (b); Reference Image after Geometrical Correction;

(c) Resulting Ca equal to 1.75 in case of no geometrical transformation; (d) Resulting Ca equal to 20.46 in case of geometrical transformation.

5. CONCLUSIONS In this paper has been shown a new approach, based on Hylemetry, which allows granting the Lithography originality. The selected hylemetric characteristic allow to obtain high verification rates, thanks also to a geometrical correction preprocessing, made in semi-automatic way, for correcting acquisition errors and misalignments during verification phase.

The proposed approach can be easily used for other unlived matter, such as banknotes [4, 5], identification documents and so on, after having correctly found an intrinsic characteristic which can be used in the hylemetric approach.

REFERENCES

[1] Strauss V., [The Lithographers Manual: A Compendium in Two Volumes;. 20th Anniversary Edition], Waltwin Publishing Company, New York (1958).

[2] Vandome F.P., F. Mcbrewster A. F., Miller J., [Affine Transformation], Alphascript Publishing, Beau Bassin, Mauritius (2010).

[3] Sjodahl&& M., Digital speckle photography, [Trends in Optical Non-destructive Testing and Inspection], Elsevier Publishing, Amsterdam, 179-195 (2000)

[4] Schirripa Spagnolo G., Cozzella L. and Simonetti C., “Banknote security using a biometric-like technique: a hylemetric approach,” Meas. Sci. Technol. 21, 055501 (2010).

[5] Schirripa Spagnolo G., Cozzella L. and Simonetti C., “Currency verification by a 2D infrared barcode,” Meas. Sci. Technol. 21, 107002 (2010).



Date post:	17-Dec-2016
Category:	Documents
Upload:	renzo
View:	215 times
Download:	1 times

SPIE Proceedings [SPIE SPIE Optical Metrology - Munich, Germany (Monday 23 May 2011)] O3A: Optics...

Documents