EP3054429A1 - Photometric dovid comparison - Google Patents

Abstract

The invention relates to a method for comparing the surface of an object (10), in particular a banknote, a chip card or a badge, comprising a DOVID area (11) with a reference object (20), likewise comprising a DOVID area (21). . a) wherein the object (10) is taken by an image recording unit (30) and digital images (32a, 32b, ...) of the same size are created, each digital image (32a, 32b, ...) each under a lighting of a Illuminant (31 a, 31 b, ...) is created with each different angle of incidence (Õ, ¸), and such a picture stack (39) comprising the individual digital images (32 a, 32 b, ...) is created, b) wherein a feature image (33) is created, which has the same size as the created digital images (32a, 32b, ...), wherein for each pixel (34x, 34y, ...) of the feature image to be created (33) in each case an illumination distribution function (36x, 36y,...) is determined which indicates how strongly the part of the DOVID area (11) imaged in the pixel (34x, 34y,...) illuminates at the respective angle of incidence (Õ, ¸) light reflects, and one or more characteristic features (37x, 37y, ...) are extracted from the illumination distribution function (36x, 36y, ...) and assigned to the respective pixel (34x, 34y, ...) of the feature image (33), and c) wherein the feature image (33) is compared with a reference feature image (23) created by a reference object (20) according to the same criteria.

Description

The invention relates to a method for comparing the surface of an object with a reference object according to claim 1.

From the prior art, a variety of articles with imprints or applications are known, each showing different views under different directions. Such imprints are for example on bills and are referred to as DOVIDs with reference to the English term "Diffractive Optically Variable Image Device". Occasionally, DOVIDs are also referred to as holograms. Different methods are known from the prior art, as the authenticity of DOVID imprints can be easily determined. In this case, the respective banknote, the authenticity of which is to be checked, is in each case compared with a predetermined genuine banknote and it is examined whether there is a match.

From the prior art, it is known to microscopically examine the relevant area to which a DOVID imprint should be present or to determine pointwise refraction patterns of the respective surface area and then to compare them. Both approaches can be achieved only with great effort and in any case do not allow a rapid investigation of a variety of items, especially banknotes in a short time.

Therefore, it is an object of the invention to provide a quick and easy to perform method, with the objects can be easily checked for the presence of certain DOVID imprints and their correctness.

• a) der Gegenstand von einer Bildaufnahmeeinheit aufgenommen wird und derart Digitalbilder gleicher Größe erstellt werden, wobei jedes Digitalbild jeweils unter einer Beleuchtung eines Leuchtmittels mit jeweils unterschiedlichem Einstrahlungswinkel erstellt wird, und derart ein Bildstapel umfassend die einzelnen Digitalbilder erstellt wird,
• b) ein Merkmalsbild erstellt wird, das dieselbe Größe aufweist wie die erstellten Digitalbilder, wobei für jeden Bildpunkt des zu erstellenden Merkmalsbilds
  • jeweils eine Beleuchtungsverteilungsfunktion ermittelt wird, die angibt, wie stark der in dem Bildpunkt abgebildete Teil des DOVID-Bereichs bei einer Beleuchtung unter dem jeweiligen Einfallswinkel Licht reflektiert, und
  • aus der Beleuchtungsverteilungsfunktion ein oder mehrere charakteristische Merkmale extrahiert werden und dem jeweiligen Bildpunkt des Merkmalsbilds zugeordnet werden, und
• c) das Merkmalsbild mit einem nach denselben Kriterien von einem Referenzgegenstand erstellten Referenz-Merkmalsbild verglichen wird.

The invention solves this problem in a method of the type mentioned above with the features of claim 1. The invention provides that

• a) the object is taken by an image recording unit and digital images of the same size are created, wherein each digital image is created in each case under illumination of a light source having different angles of incidence, and such an image stack comprising the individual digital images is created,
• b) creating a feature image having the same size as the created digital images, wherein for each pixel of the feature image to be created
  • in each case an illumination distribution function is determined, which indicates how much the part of the DOVID region imaged in the pixel reflects light when illuminated at the respective angle of incidence, and
  • one or more characteristic features are extracted from the illumination distribution function and assigned to the respective pixel of the feature image, and
• c) the feature image is compared with a reference feature image created from a reference object according to the same criteria.

It is particularly advantageous in this case that an effective checking of objects for the presence and correctness of DOVID imprints is possible even with a small number of images taken by the relevant object and with a lighting unit which can be mounted relatively simply.

A particularly simple type of check, with which it is particularly easy to determine whether a DOVID imprint is even present, provides that the eccentricity is extracted as a characteristic feature from the illumination distribution function.

In order additionally or alternatively to be able to determine whether there are DOVID areas on the object whose reflectivity changes most when tilting at the same tilt angle, it can be provided that the orientation angle is extracted from the illumination distribution function as a characteristic feature that indicates according to which direction of tilting the DOVID range changes its reflectivity the strongest or least or indicates the dominant direction of the reflection distribution in the illumination distribution function.

• dass das Merkmalsbild und das Referenz-Merkmalsbild miteinander verglichen werden, indem
  ausgehend vom Merkmalsbild ein Histogramm erstellt wird, das die Häufigkeit des Vorkommens eines Merkmalswerts innerhalb eines bestimmten Wertebereichs angibt,
• dass für das Referenz-Merkmalsbild nach denselben Kriterien ein Referenz-Histogramm erstellt wird, und
• dass das Histogramm des Merkmalsbilds mit dem Referenz-Histogramm des Referenz-Merkmalsbilds auf Übereinstimmung verglichen wird und eine große Übereinstimmung der Histogramme einer großen Übereinstimmung der DOVID-Bereiche der Gegenstände gleichgehalten wird.

A particularly simple comparison, which is invariant to the rotation of the imaged object relative to the image acquisition unit, provides

• in that the feature image and the reference feature image are compared with each other by
  a histogram is created starting from the feature image, which indicates the frequency of occurrence of a feature value within a certain value range,
• that a reference histogram is created for the reference feature image according to the same criteria, and
• that the histogram of the feature image is compared with the reference histogram of the reference feature image for agreement and a large match the histograms of a large match of the DOVID areas of the objects is kept equal.

• dass zur Erstellung des Merkmalsbilds aus der Beleuchtungsverteilungsfunktion der Ausrichtungswinkel als Merkmal extrahiert wird, und
• dass vor dem Vergleich des Histogramms mit dem Referenz-Histogramm beide Histogramme Fourier-transformiert werden und die Fourier-Transformierten der beiden Histogramme miteinander verglichen werden.

In order to be able to advantageously determine an angular orientation of a DOVID area, it can be provided that

• in that the orientation angle is extracted as a feature to produce the feature image from the illumination distribution function, and
• in that prior to the comparison of the histogram with the reference histogram, both histograms are Fourier-transformed and the Fourier transforms of the two histograms are compared with one another.

• dass vom Gegenstand von einer Bildaufnahmeeinheit jeweils unter einer Beleuchtung eines Leuchtmittels mit jeweils unterschiedlichem Einstrahlungswinkel jeweils ein Mehrfarben-Digitalbild aufgenommen wird, und diese Mehrfarben-Digitalbilder untereinander jeweils gleiche Größe aufweisen,
• dass die einzelnen Bildkanäle der Mehrfarben-Digitalbilder separat behandelt werden und für jeden Bildkanal ein eigener Bildstapel erstellt wird,
• dass nach einem der vorangehenden Ansprüche jeweils für jeden der Bildstapel separat ein Merkmalsbild ermittelt und als Farbkanal-Merkmalsbild dem jeweiligen Farbkanal zugeordnet wird und die Farbkanal-Merkmalsbilder zum Vergleich mit auf dieselbe Weise erstellten Referenz-Farbkanal-Merkmalsbildern herangezogen werden.

To check DOVID ranges at several different wavelengths of light, it may be provided

• that in each case a multicolor digital image is taken of the object by an image recording unit under illumination of a light source having respectively different irradiation angles, and these multicolor digital images each have the same size,
• that the individual image channels of the multicolor digital images are treated separately and a separate image stack is created for each image channel,
• according to one of the preceding claims, a feature image is determined separately for each of the image stacks and assigned as a color channel feature image to the respective color channel and the color channel feature images are used for comparison with reference color channel feature images created in the same way.

A particularly simple performance of a comparison of color shots of DOVID areas provides that a feature image is created from the color channel feature images by the respective feature value by pixel-wise linking the feature values of the color channel feature images, in particular by forming the average, the sum or a Maximums, is formed.

A more detailed comparison, which takes into account, in particular, the detected orientation angle strength, provides that each pixel of each color channel feature image is assigned both eccentricity and orientation angle and a feature image is formed by pixel-wise taking the orientation angle of that color channel in the respective pixel as a feature whose eccentricity for this pixel is the largest.

Furthermore, it can be provided that the reference feature image is made on the basis of the measurement of a reference object with a DOVID area according to one of the preceding claims analogously to the creation of the feature image.

Advantageously, a computer program for carrying out a method according to the invention can be stored on a data carrier.

Several embodiments of the invention are illustrated in more detail with reference to the following drawing figures.

Fig. 1a shows a structure of a detection unit with an image pickup unit and a number of bulbs. Fig. 1b schematically shows the angles of incidence of the light of a light source. Fig. 2 shows a lighting distribution function for a specific pixel. Fig. 3 shows a first histogram, which was created on the basis of the determined eccentricities. Fig. 4 shows a second histogram created based on the orientation angles. Fig. 5 shows the determination of a feature image based on a monochrome recorded image stack. Fig. 6 shows the determination of a feature image based on a multicolored recorded image stack.

In a first embodiment of the invention, the following steps are performed for both a reference object 20 and the object 10 to be examined. Both objects 10, 20 each have an area with a DOVID imprint, hereinafter referred to as DOVID area 11, 21, on. The aim of the method now described is to determine whether the DOVID area 11 of the item 10 matches the DOVID area 21 of the reference item 20 and, since the reference item 20 is a known genuine item, is genuine.

In Fig. 1a an article 10 is shown having a DOVID area 11. On the DOVID area 11 of the article 10, both an image pickup unit 30 and a number of bulbs 31 a, 31 b, ... are directed. The individual light sources 31 a, 31 b,... Radiate onto the DOVID region 11 of the object 10, each with a different angle of incidence φ, θ. How out Fig. 1b As can be seen, the angle φ denotes the angle of the illuminant in the plane, while the angle θ represents the angle at which the light source is above the plane of the object 10. In the present embodiment of the invention, a digital image 32a, 32b, ... is created by the object 10 with the image recording unit 30, wherein in each of the created digital images 32a, 32b, ... each other bulbs 31, 31 b, ... is used to illuminate the article 10 and the DOVID area 11 of the object 10. Thus, digital images 32a, 32b ( Fig. 5 ) each created with the same size. The entirety of the digital images 32a, 32b,... Created in this way is referred to as image stack 39 or combined into an image stack 39.

Starting from the determined image stack 39, a feature image 33 is created, which has the same size as the created digital images 32a, 32b,... Of the image stack 39. In this process, the intensity values of the individual digital images 32a, 32b, accumulates several characteristic features 37x, 37y. From the intensity values of the individual pixels 34x, 34y at respectively predetermined positions corresponding to each other in the digital images 32a, 32b, ..., an illumination distribution function 36x, 36y is separately determined, which determines the distribution of the brightness of the image pixel 34x, 34y Indicates the range over the angle of incidence φ, θ. For each pixel 34x, 34y, one or more characteristic features 37x, 37y are respectively extracted from the illumination distribution function 36x, 36y, which are assigned to the respective pixel 34x, 34y of the feature image 33.

One possible illumination distribution function 36x, 36y is in Fig. 2 shown in more detail. This illumination distribution function 36x, 36y indicates how much the part of the DOVID region 11 imaged in the pixel 34x, 34y reflects light when illuminated at the respective angle of incidence φ, θ. For this purpose, the respective intensity value i is separately used for the individual pixels 34x, 34y of the image stack 39 and is entered at the relevant location in the illumination distribution function 36x, 36y created for the respective pixel 34x, 34y. Owing to the known angles φ, θ of the individual illumination units 31 a, 31 b,..., It is possible to obtain the interpolation points 35 a, 35 b,... By stereographic projection onto a plane with the relevant coordinates x, y.

Each of the existing interpolation points 35a, 35b,... Is assigned an intensity i in accordance with the determined intensity i of the relevant pixel 34x, 34y. By interpolation, it is possible to determine a lighting distribution function 36 on the basis of the support points 35a, 35b,... As well as the associated intensities i.

• In einem ersten Schritt wird die Summe M₀₀ als Summe sämtlicher Intensitätswerte ermittelt. Die einzelnen x- und y-Koordinaten der Stützpunkte 35a, 35b, 35c, 35d werden zu Vektoren X, Y zusammengefügt. Die jeweils diesen Stützpunkten 35a, 35b, 35c, 35d zugeordneten Intensitätswerte werden zu einem Intensitätsvektor I zusammengefügt.

As a feature 37x, 37y for the creation of the feature image 33 in this embodiment of the invention, the eccentricity e of the respective Lighting distribution function 36x, 36y used. The eccentricity e can be simplified and determined without the need to actually determine a lighting distribution function as a continuous function in two variables x, y by discrete methods as follows:

• In a first step, the sum M _{00 is determined} as the sum of all the intensity values. The individual x and y coordinates of the vertices 35a, 35b, 35c, 35d are merged into vectors X, Y. The respective intensity values associated with these interpolation points 35a, 35b, 35c, 35d are combined to form an intensity vector I.

$$\mathrm{y\_}=\left(\mathrm{l},\mathrm{Y}\right)/{\mathrm{M}}_{\mathrm{00}}\mathrm{.}$$

Subsequently, a weighted center of gravity of the specified intensity values is determined. The x-coordinate x_ of the weighted centroid is obtained as the scalar product of the vector X of the x-coordinates of the vertices and of the intensity vector I of the intensity values divided by the sum M ₀₀ . Accordingly, the y-coordinate y_ of the weighted center of gravity results as the scalar product of the vector Y of the y-coordinates of the vertices and of the intensity values I divided by the sum M ₀₀ . $$\mathrm{x\_}=\left(\mathrm{l},\mathrm{X}\right)/{\mathrm{M}}_{\mathrm{00}}\mathrm{,}$$

$$\mathrm{y\_}=\left(\mathrm{l},\mathrm{Y}\right)/{\mathrm{M}}_{\mathrm{00}}\mathrm{,}$$

$${\mathrm{\mu }}_{11}=\left(\mathrm{l},\left(\mathrm{X}-\mathrm{x\_}\right)*\left(\mathrm{Y}-\mathrm{y\_}\right)\right)/{\mathrm{M}}_{\mathrm{00}}$$

$${\mathrm{\mu }}_{\mathrm{02}}=\left(\mathrm{l},\left(\mathrm{Y}-\mathrm{y\_}\right)*\left(\mathrm{Y}-\mathrm{y\_}\right)\right)/{\mathrm{M}}_{\mathrm{00}},$$

wobei "Vektor minus Skalar" (X-x_) bzw. (Y-y_) bedeutet, dass der Skalar von jedem Element des Vektors subtrahiert wird, und die binäre Operation "*" für zwei Vektor-Operanden die elementweise Multiplikation der beiden Operanden bedeutet. Das Resultat ist dann wieder ein Vektor.In the following, second order moments μ ₂₀ , μ ₁₁ and μ _{02 are determined} according to the following rules: $${\mathrm{<figure-callout\; id="20"\; label="\mu "\; filenames="imgf0001.png"\; state="\{\{state\}\}">\mu </figure-callout>}}_{\mathrm{20}}=\left(\mathrm{l}.\left(\mathrm{X}-\mathrm{x\_}\right)*\left(\mathrm{X}-\mathrm{x\_}\right)\right)/{\mathrm{M}}_{\mathrm{00}}$$

$${\mathrm{\mu }}_{11}=\left(\mathrm{l}.\left(\mathrm{X}-\mathrm{x\_}\right)*\left(\mathrm{Y}-\mathrm{y\_}\right)\right)/{\mathrm{M}}_{\mathrm{00}}$$

$${\mathrm{\mu }}_{\mathrm{02}}=\left(\mathrm{l}.\left(\mathrm{Y}-\mathrm{y\_}\right)*\left(\mathrm{Y}-\mathrm{y\_}\right)\right)/{\mathrm{M}}_{\mathrm{00}}.$$

where "vector minus scalar" (X-x_) or (Y-y_) means that the scalar is subtracted from each element of the vector and the binary operation "*" for two vector operands means the element-wise multiplication of the two operands , The result is then a vector again.

Subsequently, the eccentricity e is determined according to the following formula: $$\mathrm{e}=\left({\left({\mu }_{20}-{\mu }_{02}\right)}^2-{{\mu }_{11}}^2\right)/{\left({\mathrm{<figure-callout\; id="20"\; label="\mu "\; filenames="imgf0001.png"\; state="\{\{state\}\}">\mu </figure-callout>}}_{20}+{\mu }_{02}\right)}^2$$

This procedure can be carried out separately for each individual pixel 34x, 34y,... Of the feature image 33 to be created. As a characteristic feature 37x, 37y, ... of the respective pixel 34x, 34y, ... the respectively determined eccentricity e is assigned.

With the same procedure, a reference feature image 23 of the reference image 20 or the DOVID region 21 of the reference feature image 20 is created. The feature image 33 and the reference feature image 23 are then compared with each other. Here, starting from the feature image 33, a histogram 38 is created, which indicates the frequency of occurrence of a feature value within a certain range of values. For the reference feature image 23, a reference histogram 28 is created according to the same criteria. Fig. 3 shows such a histogram. Subsequently, the histogram 38 of the feature image 33 and the reference histogram 28 of the reference feature image 23 are checked for coincidence. A large coincidence of the histograms 28, 38 suggests that the DOVID areas 21, 11 of the objects 20, 10 are in great agreement.

In order to check the coincidence of two histograms 28, 38, the histograms are first normalized. Subsequently, the individual frequencies assigned to the areas can be multiplied element by element and the resulting products summed together. When the histograms 28, 38 are represented as frequency vectors, a match score can be found by forming the scalar product of the normalized frequency vectors of the histograms. For the standardization, in particular the L ₂ standard can be used.

In a second embodiment of the invention, which otherwise corresponds to the first embodiment of the invention, it is possible not to compare the feature image 33 and the reference feature image 23 by histogram comparison, but to perform an immediate examination of the two feature images 23, 33 by image comparison.

A third advantageous embodiment of the invention, which otherwise corresponds to the first embodiment of the invention, uses the orientation angle ω as a characteristic 37 as characteristic features instead of the eccentricity e. This orientation angle ω indicates in which tilting direction the DOVID area 11 has the strongest reflectivity , In the same direction, the dominant color change of the DOVID area takes place.

Based on the moments μ ₂₀ , μ ₁₁ and μ ₀₂ already described in connection with the eccentricity e, the orientation _angle can be described as follows: $$\mathrm{\omega }=\mathrm{0.5}\cdot \arctan \left(\mathrm{2}\cdot {\mathrm{\mu }}_{\mathrm{11}}/\left({\mathrm{\mu }}_{\mathrm{20}}-{\mathrm{\mu }}_{\mathrm{02}}\right)\right)$$

When the orientation angle ω is used as a feature, it is advantageous that before comparing the histogram 38 with the reference histogram 28, both histograms are Fourier transformed and the Fourier transforms of the two histograms are compared with each other. A histogram created using the orientation angle ω is in Fig. 4 and has a periodic profile with respect to its argument, which is shown only between -π / 2 and π / 2. In order to be able to neglect displacements of these histograms, which arise, in particular, from the fact that the object in question was not drawn in exactly at the same angle as the reference object, the relevant histogram is Fourier-transformed in each case. The respective Fourier transforms are invariant against rotations of the object 10 or reference object 20 during recording, so that a particularly simple comparison is possible in this case.

In a fourth preferred embodiment of the invention, the object 10 is received by an image pickup unit 30 in each case under illumination of a light source 31 a, 31 b, ..., each with a different angle of incidence φ, θ, wherein in each case one multicolor digital image 32 a, 32 b,. .. is recorded. These multicolor digital images 32a, 32b,..., Each have the same size and show the same subject area of the article 10. In the present embodiment, each of these multicolor digital images 32a, 32b has three image channels, one image channel for the red, one image channel for the green and an image channel for the blue color component of the image in question. In a first step, a separate image stack 39R, 39G, 39B is created for each image channel, and then a feature image 33R, 33G, 33B is determined separately for each image stack. This is subsequently assigned to the respective color channel as a color feature image 33R, 33G, 33B.

In principle, it is possible to compare the color feature images 33R, 33G, 33B thus obtained with color feature images of the reference object obtained in the same way and thus to verify the match. In the present exemplary embodiment, however, a feature image 33 is created from the color feature images 33R, 33G, 33B before the comparison with the reference image, wherein the feature values 37x, 37y are created pixel by pixel by linking the feature values to one another in their position corresponding pixels 34x, 34y of the color channel feature images. As the feature 37x, 37y, for example, the mean value, the sum or the maximum of the respective features in the color feature images 33R, 33G, 33B at the position concerned can be used.

In a fifth embodiment of the invention, each pixel 34x, 34y of each color feature image 33R, 33G, 33B is assigned both the eccentricity e and the orientation angle ω as characteristic features. When creating the feature image 33, pixel by pixel is examined for each pixel 34x, 34y for which image channel the greatest eccentricity is present. The orientation angle w of that color channel in the respective pixel 34x, 34y is used as a feature 37x, 37y whose eccentricity is greatest for this pixel 34x, 34y.

Claims (10)

Method for comparing the surface of an object (10), in particular a banknote, a chip card or a badge, comprising a DOVID area (11) with a reference object (20), also comprising a DOVID area (21), a) wherein the object (10) is taken by an image recording unit (30) and digital images (32a, 32b, ...) of the same size are created, each digital image (32a, 32b, ...) each under a lighting of a Illuminant (31 a, 31 b, ...) is created with each different angle of incidence (φ, θ), and such a picture stack (39) comprising the individual digital images (32a, 32b, ...) is created, b) wherein a feature image (33) is created, which has the same size as the created digital images (32a, 32b, ...), wherein for each pixel (34x, 34y, ...) of the feature image to be created (33) in each case an illumination distribution function (36x, 36y,...) is determined which indicates how strongly the part of the DOVID area (11) imaged in the pixel (34x, 34y,...) illuminates at the respective angle of incidence (φ, θ) reflects light, and one or more characteristic features (37x, 37y, ...) are extracted from the illumination distribution function (36x, 36y, ...) and assigned to the respective pixel (34x, 34y, ...) of the feature image (33), and c) wherein the feature image (33) is compared with a reference feature image (23) created by a reference object (20) according to the same criteria. Method according to claim 1, characterized in that from the illumination distribution function (36x, 36y, ...) the eccentricity (e) is extracted as a characteristic feature (37x, 37y, ...). Method according to claim 1 or 2, characterized in that from the illumination distribution function (36x, 36y, ...) the orientation angle (ω) is extracted as a characteristic feature (37) indicating after which tilting direction the DOVID region (11) its reflectivity changes most or least or indicates the dominant direction of the reflection distribution in the illumination distribution function (36x, 36y, ...). Method according to one of the preceding claims, characterized - That the feature image (33) and the reference feature image (23) are compared with each other by
on the basis of the feature image (33) a histogram (38) is created which indicates the frequency of occurrence of a feature value within a certain value range, - That for the reference feature image (23) according to the same criteria, a reference histogram (28) is created, and in that the histogram (38) of the feature image (33) is compared with the reference histogram (28) of the reference feature image (23) for agreement, and a large match of the histograms (28, 38) of a large match of the DOVID ranges of Objects (20, 30) is kept the same. Method according to claim 4, characterized in that - that for the creation of the feature image (33) from the illumination distribution function (36x, 36y, ...) of the orientation angle (ω) is extracted as a feature (37), and - be that prior to the comparison of the histogram (38) with the reference histogram (28) both histograms (28, 38) are Fourier transformed and the Fourier transforms of the two histograms (28, 38) are compared. Method according to one of the preceding claims, characterized - That each of the object (10) of an image pickup unit (30) under a lighting of a luminous means (31 a, 31 b, ...), each having a different angle of incidence (φ, θ) recorded a multi-color digital image (32a, 32b) and these multicolor digital images (32a, 32b, ...) are each of equal size, - that the individual image channels of the multicolor digital images (32a, 32b, ...) are treated separately and a separate image stack (39R, 39G, 39B) is created for each image channel, Is and that after one of the preceding claims, respectively for each of the image stack (39R, 39G, 39B) separately a feature image (33R, 33G, 33B) is determined and assigned as the color channel feature image (33R, 33G, 33B) of each color channel of the color channel - Feature images (33R, 33G, 33B) are used for comparison with reference color channel feature images (23R, 23G, 23B) created in the same way. A method according to claim 6, characterized in that from the color channel feature images (33R, 33G, 33B) a feature image (33) is created by the respective feature value by pixel-wise linking the feature values of the color channel feature images, in particular by forming the average, the Sum or a maximum is formed. A method according to claim 6, characterized in that each pixel (34x, 34y) of each color channel feature image (33R, 33G, 33B) is assigned both eccentricity (e) and orientation angle (ω) and a feature image (33) by pixel-by-pixel of the orientation angle (w) of that color channel in the respective pixel (34x, 34y) is formed as a feature (37x, 37y) whose eccentricity (e) is greatest for that pixel (34x, 34y). Method according to one of the preceding claims, characterized in that the reference feature image (23) on the basis of the measurement of a reference object (20) with a DOVID area (21) according to one of the preceding claims is made analogously to the creation of the feature image. Data carrier on which a computer program for carrying out a method according to one of the preceding claims is stored.

Images