EP2433245A1

EP2433245A1 - Apparatus and method for identifying the creator of a work of art

Info

Publication number: EP2433245A1
Application number: EP10728585A
Authority: EP
Inventors: Werner Scholzen
Original assignee: Individual
Current assignee: Individual
Priority date: 2009-05-20
Filing date: 2010-05-17
Publication date: 2012-03-28
Also published as: BRPI1010963A2; JP2012527665A; RU2541917C2; WO2010133204A1; CN102439605A; DE102009023756B4; CA2761382C; DE102009041757A1; DE102009023756A1; CA2761382A1; US20120072454A1; RU2011149157A; DE112010002056A5; KR20120024799A; US8930302B2

Abstract

Method for determining the creator of a painting, wherein the method has at least the following steps: - converting the painting to be examined or parts of the painting to be examined into at least one data record with the aid of a digitizing means, in particular a scanner, - analysing the data record(s) and determining characteristic features or parts of characteristic features, in particular dots or lines, or groups of dots or lines or patterns, which are contained in the data record in digitized form, wherein the characteristic features to be determined are stored in a database, - and wherein the database has an additional associated data record for each of these stored characteristic features.

Description

DEVICE AND METHOD FOR IDENTIFYING THE COUNTER OF AN ART WORK

The present invention relates to apparatus and methods for detecting the origin and authorship of images.

In the art world the origin of a work of art, ie the assignment of the respective work to an artist, is of immensely important importance, both for ascertaining the actual authorship and for assessing the value of a work, as well as from an art historical point of view. As the author of the artwork, the artist has very extensive rights to this work of art, which he does not lose even after the sale of the work of art. Furthermore, the art historians are interested in the true origin of a work of art, since in many cases, especially in older works, the artist of these works can not be determined beyond doubt. This can be due either to a missing or forged signature, or to important masters having a lot of docile students who asked them to write parts or whole works of art, which were then signed by the artists. Rembrandt is such a significant example. In these cases, even a high technical standard based investigation apparatus is used, which is used for checking the works of art. Thus, the X-ray of images is not uncommon and it is often infrared radiation used to detect watermarks of paper mills. UV-lighted materials fluoresce in different colors and thus provide information about the material underlying the work of art. From this conclusions about the respective artist can be deduced. Further methods for the determination of forgeries are the thermoluminescence analysis, elaborate chemical analyzes or the examination of the works of art under a microscope. However, these methods are very complex and must, moreover, be carried out in suitable locations, usually laboratories, that is to say places which are not caused by the radiation to cause any damage to the health of the personnel, or climatically controlled premises which condition the image Take care.

On the other hand, art experts can often determine the originator of a work of art, especially a picture, with their trained eye alone. They have acquired the ability to associate an artist as the author with a particular image in the course of their training and working activities. For example, the application of color, lines and brushstrokes are some of the characteristics that characterize the artist and facilitate the identification of his works. The brushstroke is generally the stroke in painting. He can express the importance of each area of the image for each artist. For example, in an image, the lines on the face can be made finer than the style of the lines that the artist used in his clothes. Hietan one can recognize that the artist has placed particular value on the design of the facial expression or the face itself. Det brushwork is thus the personal signature of an artist. It may include, among other features, the brushwork, the lines, the lightness and the strength of the brush pressure. A work of art, especially an image, can be identified with the help of this personal signature of the artist, ie the author, and then assigned to the correct author. This is currently only possible with the help of a visual inspection of the image by an art expert. Here, the art expert must study the image carefully and carefully examine. Such a process is time consuming and usually very expensive, as the art experts are highly skilled experts.

It is therefore an object of the present invention to provide an apparatus and a method which assigns a predetermined image from a set of also predetermined artists, the artist who has created the image.

This object is achieved according to the invention by a method for determining the originator of an image, the method having at least the following steps:

Transfer of the image to be examined or of parts of the image to be examined by means of a digitizing means, in particular a scanner, into at least one data record, Analysis of the record (s) and determination of characteristic features or parts of characteristic features, in particular points or lines or point or line groups or patterns which are digitized in the data record, the characteristic features to be determined being stored in a database,

and wherein the database has an additional associated data record for each of these stored characteristic features.

In this method according to the invention, expert knowledge is retrieved by the art expert via a database. It depends only on the access speed to this database and the algorithm used to search in the database, as well as the size of the records to be examined and of course the amount of relevant records in the database, how quickly a result can be presented here. Determination of characteristic features in the data record in the context of this application is to be understood as recognition of characteristic features in the data record (and then, of course, finally in the image to be examined) or as identification of characteristic features in the data record.

The transfer of the image or parts of the image into at least one data set by means of a digitizing means in most cases means scanning the image with a scanner. But it is also possible to use a corresponding digital camera. This has the advantage that the digitization with the help of the camera can be made while traveling. This provides a simple and quick way to gain digitized data while ensuring great mobility, especially without transporting the artwork, ie directly at the exhibition site. A scanner also constitutes such a data acquisition device which scans or measures an object, in particular an image, in a systematic and regular manner. In this case, a large number of individual measurements are used to generate an overall image of the object to be scanned or of parts thereof. For 3D artwork, a 3D scanner can be used. With the so-called Cruse scanners, in which the image original is illuminated with synchronized light, and wherein the then reflected light is focused on a lens, even the absolute color fidelity and accuracy of the scanned image can be ensured with the original image. For simple drawings, this may even be one Commercially available scanner, which advantageously has a very good recording quality (the term "very good recording quality" changes over time as the current new generation of scanners can have ever better recording or resolution) Depending on the resolution set, data sets may require a large storage capacity even for small scans, but this requirement, namely the provision of large storage media, can certainly be realized cost-effectively with today's storage options.

The analysis to be carried out according to the inventive method, the data sets obtained by scanning, and the determination of the characteristic features, which naturally exist in their digital form in the data record, are carried out with the help of the templates of the characteristic features stored in the database. These patterns are compared with the image to be examined, whether these patterns are found, either in identical form or with a great similarity in the image to be examined.

The determination or recognition of these characteristic features, in particular points or lines or dot or line groups or patterns, in the image to be inspected requires the ability to recognize regularities, repetitions, similarities or laws in a set of data, recognizing similarities has the greatest chance of success. Such a capability can be provided by the methods of pattern recognition. Other characteristic features can be edges, color edges, the course of colors, ie the color gradient and the brushstroke itself.

Basically, the scanned and inspected image is subjected to image analysis, which is described below including its preparatory steps. This analysis greatly facilitates the subsequent comparison with the patterns available in the database.

After digitization, the image to be inspected is subjected to a preparation step. In this step, among other things, the image is normalized so that it can later be compared with the patterns to be compared. If necessary, interfering impurities can also be suppressed by the use of filters. It is also possible to convert the colors of the image to shades of gray, if this is helpful in the following steps. Hietnach the image can be divided into segments, so homogeneous areas, ie areas with, for example, the same texture or the same color, can be combined to save space.

Thereafter, the feature extraction is performed. Basically, the methods for obtaining characteristics are mostly methods that are obtained at best by intuition or based on many years of experience of a specialist. Important features can be processed into feature vectors. Features in the images may be straight lines, circular arcs, circles, ellipses or other geometrically writable groups. From several of these features or even from a feature alone, the characteristic feature, which can be assigned to the author exist.

The extracted features and patterns are classified in a further step, so subjected to a classification process. The classification method, also called the classifier, divides the extracted features and patterns into classes. Classification procedures are already known in the art. There, manual, automatic, numerical and non-numerical statistical and distribution-free or even dimensioned and learning methods are known. There are also other methods known, so that the list given above is not intended to be limiting. Here, the classification can be made, for example, with the aid of a Bayes classifier. This classifier assigns each feature exactly to the class to which it is most likely to belong.

On the basis of this classification, the actual image analysis can now take place, in which now the image recognition or the image interpretation takes place. Here, in a first step, only what is relevant can be seen; in a second step, the relationship of the features is then additionally weighted among one another.

The features thus identified and analyzed are compared with the characteristic features present in the database. If one or even more matches are achieved here, then in addition to the name of the author, an assigned data record with further information about the author of the picture can also be output. In yet another embodiment of the invention, this is characterized in that the resolution of the digitizing means, preferably a scanner, can be freely adjusted. Through this possibility, both the respective artwork, as well as the final amount of data to be processed, which is created by the digitization process, be taken into account.

If an artist prefers a delicate technique in the creation of his works, which is characterized by thin and possibly still a very light lines, it is necessary to perform a high resolution mode to perform the image analysis, so that even the finest and smallest details digitized and then examined can be. In other cases, such as a work, which z. If, for example, a number of monochrome squares consists of a large canvas, the significance of the work of art is very high, but the information obtained through the digitization of the work is manageable and can be completely captured with only a small amount of data.

In another preferred embodiment of the invention, the Hough method is also used to analyze and determine the characteristic features in the image to be tested or in parts of the image to be tested. The Hough method is a method which is based on the transformation of the same name, namely the Hough transformation. This is suitable for detecting straight lines, ellipses and other geometric objects. From these objects, the characteristic features that are to be found or composed are composed. In addition, the Hough method is still a very robust method with which the structures of lines can still be identified in a noisy picture, ie a picture in which the geometric objects can not be clearly identified. So not only complete lines, but also line segments or other segments and parts of the geometric objects can be determined.

In a further very particular preferred embodiment of the invention, this is characterized in that the method in a further step determines reference features of the characteristic features or of parts of the characteristic features contained in the dataset, the reference features of the characteristic features either already in the database are stored or while running Procedure are generated. The reference feature to a characteristic feature in the sense of this application is generated by a change in the characteristic feature of just this. For the human eye, only slight differences between the reference feature and the characteristic feature can be seen. By determining the reference features in the image to be tested, minor and minor deviations in the characteristic features, which can always occur, can now be identified and errors resulting therefrom in the assignment of the work to the author can be avoided. An artist, of course, has his personal signature characterizing him, he always draws or paints certain forms, or even almost identical at best. But there are always differences in the drawn or painted forms. It is important to recognize these deviations and, if necessary, to determine them as characteristic features in the image to be tested. The generation of the reference features are generated by means of an image processing module. Such image processing modules are known in the art. Thus, for example and not by way of limitation, individual lines, curves, lines, angles or even entire geometrically connected figures are manipulated, with the characteristic features or parts of the characteristic features forming the starting point for these manipulations.

In a very simple case, the characteristic features or parts of the characteristic features are simply enlarged or reduced, and it may well be the case that more than one characteristic feature or part thereof is processed. It is even possible, on the one hand, to enlarge one part of the characteristic feature, but on the other to reduce the size of another part.

In a further preferred embodiment of the present invention, the reference features are generated by stretching or compressing the respective characteristic feature or at least part of the characteristic feature. In this embodiment, unlike the above-mentioned last embodiment of the invention, the characteristic feature is more distorted. However, such a manipulation can certainly lead to a positive result, if the characteristic feature had forced a compression or extension in the creation of the work due to the present image geometry of this image to be tested. Furthermore, in these manipulations of Characteristic features or parts of these that one part can be compressed, another part but stretched.

In a further preferred embodiment of the present invention, at least one reference feature is generated by changing a line curvature.

In another embodiment, the reference feature is generated by changing an angle between at least two lines of the respective characteristic feature.

All these manipulations, which are made, must not change the actual character of the characteristic features so that they can no longer serve their original purpose, namely the assignment of an author to a work of art. However, to the enumerated changes of the characteristic features still further other manipulations are conceivable, in order to produce possible reference characteristics.

It is also out of the question that the present inventive method can also be applied only to parts of images, so image sections. In this case, only parts of these images that need to be checked are digitized. This may be the case when a particular "characteristic feature" catches the eye, but the viewer is not sure whether this particular "characteristic feature" is an original or a forgery to deceive the viewer.

Such a method can, for example, run on a device which has at least one digitizing means, preferably a scanner, a normalization module, a segmentation module, a classifier module and a database module. Furthermore, at least one storage medium and at least one data processing unit for such a device are conceivable.

Fig. 1 schematic sequence of the method

2 shows an exemplary explanation of the Hough method with reference to a straight line (straight line)

Fig. 3 is a schematic representation of a color edge

Fig. 4 schematic representation of a color gradient Fig. 5 Black / white copy of an original with five marked image areas

5a to 5e show the marked image areas of FIG. 5

Fig. 6 Black / white copy of a suspected imitation with five marked image areas

FIGS. 6a to 6e show the marked image areas of FIG. 6

Fig. 7 Copy of a work by the artist Max Clarenbach

7a detail of FIG. 7th

Fig. 8 Copy of a presumed imitation of the artist M. Clarenbach

9 is a schematic representation of the color edge, color gradient and brush stroke

Fig. 1 shows the schematic sequence of the method according to the invention, as well as the use of some of the necessary modules. In a first step, the image to be examined is digitized by means of a digitizing means 1, in this case a high-resolution scanner. The scanned image is probably a miniature artwork by the very important artist XY, who painted this image on the back of an antique matchbox. Over time, the work has suffered considerably and the fact that the matchbox was previously used as intended was not conducive to the quality of the back. For this reason we worked with a very high resolution. As a result, the individual, barely recognizable to the naked eye image objects were digitized so that as little details are lost. The high resolution is also used to better detect and suppress possible ambiguities, such as contamination of the pad, in the subsequent process. Although the high resolution generates a large amount of data, the small size of the image compensates for this disadvantage. In addition, today's technology is quite capable of efficiently storing and managing very large amounts of data. Subsequently, the image is normalized with the aid of a normalization module 2. This serves for better comparability of the respective found characteristic features with the patterns present in the database 5. Since the patterns present in the database 5 are also normalized, so at least the size ratios of the individual characteristic features are similar. An artist will always carry out his characterizing movements, which manifest in the brushstroke, in roughly the same manner, so that in most cases the structure and magnitude of the result of these movements will be similar.

In this example, the image is divided into segments in the following step. This is useful in the present case, since the image consists mainly of four features, namely stroke, circle, heart and sun. By a clever choice of segmentation, which does not necessarily have to be carried out in this way, exactly one feature is found in each image segment. The module which performs these steps is called segmentation module 3.

The features found are now divided into the classes in which they are most likely to belong by a classification method carried out in the classifier module 4 suitable therefor. Thus, the stroke is assigned to the class of {strokes}, the circle of the class of {circles}. However, the heart and the sun may not be stored as classes in the database and thus can not be found. There is now a risk that the characteristics are assigned to a wrong class. Thus, the heart can be assigned to the class of {triangles} or the class of {deformed triangles}. But if the object heart is a characteristic feature of the artist to be found, it is also deposited in its entirety as a class of {hearts} in the database and then, in all likelihood, in the same, albeit false class, namely the class {deformed triangles}. For basically the characteristic features present in the database were also classified by the classifier 4, so that in the case of identity or great similarity the classifier 4 as a rule carries out the same classification. If only parts of the object heart are to be identified as a feature, then these are classified into the respective classes. A heart can be subdivided into two ellipsoidal segments with an adjoining straight line. As can be easily seen, the sun can be divided into a circle and a number of triangles adjoining it. These subdivisions are divided into the respective classes. Now, if the triangles are drawn in a certain characteristic way, there is a pattern of this in the database and will inevitably be found. The database 5 with the present patterns of the characteristic features may be any commercially available database. Only the connection to the respective modules, which provide the features to be compared, is important. The database 5 present in this example includes, among other things, the following four characteristic features, namely stroke, star, double arrow, and heart. The heart from the database is almost identical to the present heart, which can be found in the image to be examined. The sun of the image to be examined also shows great agreement with the sun, which was found in the database. However, since there is no identity in this case, the object sun (not shown here) is further broken down into its details in an intermediate step and then compared with the characteristic features of the database 5. Thus, the sun's association of the database with the sun can be verified on the image under test. There is no equivalent in the database 5 for the object circle. Since all found characteristic features of the database 5 indicate the same originator, the result was unique in this present example case. It was indeed the artist X Y. The result output 6 is supplemented by further, also in the database 5 information on the respective artist, so that the request not only determines the artist, but also more information about the artist's work and effect and outputs via an output module, such as a screen or printed on paper outputs.

2 shows an exemplary explanation of the Hough method with reference to a straight line (straight line). The empty boxes (pixels) are kept in one color, for example white, the boxes (pixels) marked with the X are in a different color, for example black. The human eye now easily recognizes a catch in the present section, even though the pixels are not all the same size. A hook can consist of two touching straight lines. A straight line can be given mathematically by its perpendicular distance r to the origin of the coordinates as well as by the angle φ between the corresponding connecting path and a coordinate axis. The division of the image into pixels now represents a suitable coordinate system for this purpose, the coordinate jump should lie in the lower left corner, the horizontal with the values x (i) and the vertical with the values y (i) is called (with i as a passing natural number). The section shown in Figure 2 has 13 pixels in the horizontal direction, thus i runs from 1 to 13, that is, x (1), x (2), ..., x (13). In the vertical, 12 pixels are visible, so i runs from 1 to 12, so y (l), y (2), ..., y (12). The easy for the human eye identifiable straight line passes over the pixels with the value pairs (x (2), y (10)}, {x (3), y (9)}, {x (4), y (8)}, (x (5) , y (7)}, (x (6), y (6)}, {x (7), y (5)}, {x (8), y (4)}, {x (9), y (3)}, then the line turns in a different direction In the clipping additional pixels are colored, thus marked with an X. This line can also be represented by a series of value pairs (r, φ) For all value combinations (r, φ), whether the pixels lying there all correspond to one color or not.If this color is different from the color of the environment of these pixels, the line is visible and represents a straight line in the image for the viewer. All other geometrical shapes obey other mathematical formulas, but can still be found and determined using the same method adjacent pixels are also recognized as a straight line or part of a straight line. For two directly adjoining and touching straight lines can be recognized by a viewer as a wider straight line. The more of these straight lines lie next to each other without a gap, the thicker the line is perceived as a line in the picture by the viewer.

Fig. 3 shows a schematic representation of a color edge, wherein the shades of white and black each fill a range. The black-colored color range starts at the value xl and ends at x2, the color edge. The white colored area begins at x2 and ends at x3. Underneath, the value curve is also shown schematically in the RGB space, whereby only the R value (marked in bold in the value triple (R ₃ B ₅ G)) is plotted against the length of the color areas, which are marked as x values. In the RGB space, the white area is assigned the RGB code (255,255,255) and the black area is the RGB code (0,0,0). The curve is relatively simple. The black color range has the RGB value (0,0,0) over its entire length, here in this range the curve runs constantly and steadily. At the color edge at the length value x2, there is a point of discontinuity. Throughout the white area, the curve has the RGB value (255,255,255), so it's constant and steady again. Such a curve can be stored as a characteristic feature, thus stored as a record in the database and stored. Of course, fault tolerances can be included as additional information in the data set, so that smaller deviations of the features in the image to be tested are recognized by the characteristic features stored in the database. Fig. 4 shows a schematic representation of a color gradient. The color gradient is shown here from left to right, from white to black. He goes through the grays continuously. Underneath, the value curve in the RGB space is also shown schematically, whereby only the R value (marked in bold in the value triple (R, G, B)) is plotted against the length x. Again, the RGB value goes through the values from (0,0,0) to (255,255,255). The curve in this case is a straight line with a linear slope, according to the formula:

R = a * x, where a is the slope of the curve.

Such a curve can also be stored as a characteristic feature, and thus as a record in the database and stored. Of course, error tolerances can also be included as additional information in the data record in this case, so that smaller deviations of the features in the image to be checked from the characteristic features stored in the database are detected.

FIGS. 5 and 6 show a selection of a plurality of sections from three images, wherein the difference between original and presumed imitation is to be made clear by means of an exemplary explanation.

The original picture steel tube by Joh. Georg Müller from the year 1963 and a presumed imitations, which deal with the same topic, are to explain the erfϊndungsgemäße procedure on a practical example. For this purpose, the original (marked 2009 03 17 -3) and certain parts of the original characteristic of the artist were digitized. In the overall view, the five characteristic areas are characterized by square white borders. In addition, they have each been designated with a number sequence to rule out possible confusion. The numbers are:

• 2009 03 17 -3-001,

• 2009 03 17 -3-002,

• 2009 03 17 -3-003,

• 2009 03 17 -3-004 and

• 2009 03 17 -3-005.

The same has been done with the two alleged counterfeits. In the image with the designation 2009 03 17 -2 as well as in the image with the designation 2009 03 17 -3-1, five image areas are thus also highlighted by square borders Service. However, these image areas have been selected to be comparable to the characteristic locations of the original image. The attached labels are now:

• 2009 03 17 -2-001,

• 2009 03 17 -2-002,

• 2009 03 17 -2-003,

• 2009 03 17 -2-004 and

• 2009 03 17 -2-005

In FIG. 5, the original image, the following five image sections have been identified by the white border:

1. Fig. 5a is a picture section (marking: 2009 0317-3-001) in the lower right quadrant of the picture, which in the colors orange, white, gray, HIa, black and ocher, as well as individual intermediate stages of these shades, a stylized tube shows. In this image section, the color transition from the dark area to the bright area ending at the black horizontal line is particularly characteristic of the artist's technique and thus represents a characteristic feature that is incorporated into the database. The color gradient, seen from top to bottom, changes from a dark hue (black) continuously into a light hue (yellow-white). Designing a color gradient just as carefully requires not only a special craft and artistic talent, but also a sophisticated technique. Such a continuous transition can be accurately represented, for example, by an equally continuous function in the RGB color space and is thus easy to digitize.

2. Fig. 5b shows a completely black image section, which is also located in the lower right quadrant, but to the left of the first image section and towards the center. Here, the brushstroke, which is so characteristic of the artist, becomes visible, which is reflected here in the uniform color density and color. This too can be mathematically described by a function.

3. Fig. 5c of this image section again illustrates an example of how the artist designed a gradient. The image section is located in the lower left quadrant of the image and shows the shadow play on the surface of a pipe. Again, this is true for IU

the image area shown in Fig. 5a, namely that a description of the color gradient mathematically by means of a function, is easily possible. This function represents an approximately linear curve in the three-dimensional RGB space, since the respective RGB values change continuously and continuously.

4. Fig. 5d, the fourth image section is located in the left upper quadrant of the image and shows the artistic treatment of the color edges that separate different color areas. The respective color areas are subdivided by a color edge which discretely changes to the hue of the respective color area. The color areas themselves have the same color throughout, as can already be seen in FIG. 5b. These features, ie the same color image areas separated by color edges whose hue corresponds to the respective following image area, is another characteristic feature of the painter's art of the artist.

The RGB value of the first color range and the RGB value of the second color range are approximately constant within their color ranges. At the color edge, however, the RGB bet changes abruptly, the mathematical function described here has a discontinuity at this point.

5. Fig. 5e The fifth picture detail from the upper right quadrant of the picture again shows the shading on a tube and clearly shows the way in which the artist creates a shadow with the aid of a gradient from the light into the dark.

These five image sections are part of the datasets stored in the database, which contain some characteristic features of the artist J. G. Müller, namely the design of a color edge, the design of the color run and the brush stroke, which ensures a consistently uniform color within a color range. Thus, in this example, the five above-mentioned image sections are stored as an example of the characteristic features in the database.

Fig. 6 now shows a copy of a presumed imitation of a work by the artist JG Müller, which is to be checked for authenticity. For this purpose, for example, also scanned five image areas and then compared using the inventive method with the deposited in the database characteristics of the artist. It is natural It is also quite possible to automate this process by treating the image either as a whole or subdividing it into arbitrarily selected segments using a program. These segments are then individually examined for possible characteristic features using the invention.

Fig. 6a shows the first marked image area (2009 03 17 -2-001) of the alleged imitation. When viewing the overall picture, the viewer perceives a color edge in the image area, which separates a yellow color area from a yellow-green color area. It turns out, however, that the color edge does not separate the two color ranges from each other, as the artist J. G. Müller realized in his pictures. Thus, the edge itself has a darker hue than the subsequent color range, wherein the subsequent green color strip in itself also has no uniform color. A comparison with the data records stored in the database does not give a positive agreement with the data records available there. In this case, since the color edge, as the characteristic feature, separates two color areas which have the same color tone throughout. This can not be described by a function having the selection criteria described above.

Fig. 6b shows the shadow of a tube. This section is very similar to Figure 5c, but the shades do not change slowly, but their transitions are more abrupt. This can then by no means be described as an approximately linear curve in the RGB color space. Thus, no characteristic feature can be found in the database which can be assigned to this image detail.

Fig. 6c and Fig. 6d shows the representation of color edges. If one compares this presentation shown here with the representation in the image detail Fig. 5a, it is easy to see that, even in this case, the feature of color edge matching does not correspond to the feature stored in the database.

Fig. 6e shows the representation of a color gradient. Again, the above applies.

The image excerpts shown above and the comparison of the original with the presumed imitation now shows emphatically how the author of an image can be determined or disputed with the help of the method according to the invention. Figures 7 and 7a show a copy of a work by the artist Max Clarenbach, born in 1880 in Neuss and died in 1952 in Wittlaer. Max Clarenbach was a German painter and co-founder of the Sonderbund in Dusseldorf. His nuance-rich style of painting was determined mainly by the Impressionists. This copy and the accompanying detail is a landscape image showing a snowy landscape along a river. This picture clearly illustrates how an artist can be identified by his brushwork and brushstroke. Here you can clearly see the constantly repeating semi-circular, sweeping brushwork, which probably runs from left to right and therefore ends with a color patch on the left side.

FIG. 8 shows a presumed imitation of the painting style of the artist M. Clarenbach. Looking more closely at the sky here, one sees beyond any doubt that the creator of this painting has guided the brush differently, not always from left to right, but also from top to bottom. Also, the individual brushstrokes do not have the characteristic arcs and the color cast produced thereby.

FIG. 9 again shows a possible comparison between the possible characteristic features color edge, color gradient and brush stroke.

Claims

claims

A method of determining the originator of an image, the method comprising at least the following steps:

Transfer of the image to be tested or of parts of the image to be examined by means of a digitizing means, in particular a scanner, into at least one data record,

Analysis of the record (s) and determination of characteristic features or parts of characteristic features, in particular points or lines or point or line groups or patterns which are digitized in the data record, the characteristic features to be determined being stored in a database,

2. The method according to claim 1, characterized in that the resolution of the digitizing means is freely adjustable.

3. The method according to any one of the preceding claims, characterized in that the Hough method is used to analyze and determine the characteristic features in the image to be tested or in the parts of the image to be tested.

4. The method according to any one of the preceding claims, characterized in that the method in a further step, reference features of at least one of the characteristic features or parts of the characteristic features contained in the record are determined, wherein the reference features of the characteristic features are either already stored in the database, or generated during the current process.

5. The method according to any one of the preceding claims, characterized in that the reference features are generated by manipulations of at least one of the characteristic features or a part of the characteristic features.

6. The method according to any one of the preceding claims, characterized in that at least one reference feature of the characteristic features by increasing or decreasing the respective characteristic feature or at least a portion of the characteristic feature is generated.

7. The method according to any one of the preceding claims, characterized in that at least one reference feature is generated from the characteristic features by stretching or compressing the respective characteristic feature or at least a part of the characteristic feature.

8. The method according to any one of the preceding claims, characterized in that at least one reference feature of the characteristic features by changing a curvature of a line of the respective characteristic feature or at least a portion of a line of the characteristic feature is generated.

9. The method according to any one of the preceding claims, characterized in that at least one reference feature is generated from the characteristic features by changing an angle between at least two lines of the respective characteristic feature.

10. A device for determining the originator of an image, which operates according to one of claims 1 to 9, characterized in that said device at least one digitizing means (1), preferably a scanner, a normalization module (2), a Segmentiermodul (3), a Classifier module (4) and a database module (5).

11. Computer system having at least one data processing unit, at least one memory, at least one digitizing means (1), preferably a scanner, at least one standardization module (2), at least one segmentation module (3), at least one Classifier module (4) and a database module (5), characterized in that the data processing unit is programmatically arranged to operate according to the method of any one of claims 1 to 9.

12. Computer program having instructions that are adapted to carry out the method according to one of claims 1 to 9.

13. Computer program product which has a computer-readable medium with computer program code means in which, in each case after loading the computer program, a computer is caused by the program for carrying out the method according to one of claims 1 to 9.

14. Computer program product, which has a computer program on an electronic carrier signal, in each case after loading of the computer program, a computer is caused by the program for carrying out the method according to one of claims 1 to 9.