Methods for camera-based optical measurements

Abstract

The invention relates to a method for a camera-based optical measurement in which a reference screen and a digital camera are coupled in an optical configuration, wherein reference patterns are displayed on the reference screen and recorded by the digital camera as a camera image and the recorded camera image is decoded and evaluated as a data set by a computer, wherein at least one visual marking is integrated in the reference pattern.

Description

The invention relates to a method for camera-based optical measurement according to the preamble of the appended main claim.Such optical measurement technology is used, for example, in industrial optical quality testing or also in computer-based vision (for short: computer vision or computer vision) or machine vision (machine vision), and also associated therewith in robotics and increasingly also in autonomous driving.For camera-based optical measurements, the method of moving targets ("active targets" =ATs) is frequently used. Such camera-based optical measurements are used, for example, to calibrate high-quality digital cameras precisely or for deflectometric assessment of specular surfaces or also for optical deflectometric tomography. Data sets are generated in this case by high-quality images in the form of reference patterns being generated on flat or also curved LCD, OLED or comparable screens or computer monitors. Screens of this type are collectively referred to below as reference screens. The reference patterns mentioned are recorded with digital cameras. The reference patterns generated with the reference screens mentioned are considered here to be precise reference objects for the further processing.In the previously known methods, a control PC displays a sequence of specific, in particular full-area reference patterns on a connected reference screen for a specific predefined duration. Once the reference pattern appears on the reference screen, the control PC causes a connected digital camera to take a photograph.The reference patterns generated or displayed by the control PC are designed so that the sequence of gray level values displayed on a particular screen pixel uniquely identifies the 2D position of that pixel on the screen. For such applications, there are known different reference pattern sequences: gray codes, cosine phase shifted codes, stochastic position codes, etc. The number of reference patterns in practical reference pattern sequences typically varies between a few dozens and a few hundreds.In the methods and devices known up to now, the reference screen on which the reference pattern is shown and the digital camera which recognizes the reference pattern are coupled in an optical configuration. - In practice, the digital camera would be formulated to watch the screen.There are several typical optical configurations that use the moving target method: one is a digital camera directed to the screen. This is used, for example, for calibrating a digital camera. There is also, however, the deflectometry, in which the digital camera monitors a reflection of the reference screen, which occurs in a specular surface examined in this case. Furthermore, there is also the variant of optical deflectometric tomography in which the digital camera monitors the reference screen as it appears through a transparent object examined in this case.In all these cases, a single pixel on the light-sensitive image sensor of the digital camera is intended to record an image of at most a single pixel of the reference screen. An effort is being directed under this aspect to avoid multiple reflections, overlapping / duplex images, and other types of ambiguous signals.In data acquisition, the recorded camera images are decoded to establish correspondences between the position of the pixel on the image sensor of the digital camera and the pixels on the reference screen. In the simplest case, the resulting data record contains, for each 2D point (pixel) at the position on the image sensor of the digital camera, the 2D coordinates of the respective emitting pixel on the reference screen.Based on this information, the quantities of interest, such as the specific distortions occurring with respect to each pixel on the image sensor in the digital camera, can be determined or the shape of an examined reflecting surface or a refractive object, etc. can be derived.Since the control PC controls both the digital camera and the reference screen, the association between the recorded images and the displayed reference patterns is easy to establish: As soon as a reference pattern is displayed, the digital camera is initiated, for example, by means of an external, in particular electronically controllable, trigger for creating an image. When the exposure on the digital camera side is completed, the reference pattern on the reference screen side can be changed.It should be taken into account that if the control PC instructs the operating system of the reference screen to change the displayed reference pattern, it takes a certain time interval until the screen pixels to be individually controlled for this purpose (again) reach their necessary target intensities.For example, when a display provided with the in-plane switching technology (a so-called "IPS display") is used as the reference screen, the duration of the time interval is about 0.1 s.It is therefore expedient to plan a time interval after the change of a reference pattern which is at least as long as the mentioned duration of the time interval before the digital camera is triggered.As will be apparent from these explanations, the implementation of a corresponding system requires precise synchronization between the reference screen and the digital camera in order to correctly associate the images recorded by the latter with the images displayed by the reference screen with reference patterns present therein.In practice, appropriate fine tuning of device drivers and / or programming interfaces ("application programming interfaces"= APIs), which enables independent applications to communicate with one another and to exchange data, may be difficult.While controlling commercially available reference screens in the form of the above-mentioned. LCD, OLED screens or computer monitors with the common tools and APIs, which are provided for computer games, for example, the control of a digital camera can still represent a challenge in the method described here.Digital cameras are available that provide full control over dozens of capture parameters. However, these are generally expensive industrial cameras.It is therefore the object of the present invention to simplify the implementation of a corresponding system and in particular also to make it less expensive.The invention is based on the finding that, instead of expensive industrial cameras, it is also possible in principle to use cheaper (digital) consumer cameras which, on account of their specifications, are in principle also completely sufficient for the object to be fulfilled by them within the scope of the method described above. However, these consumer cameras do not provide full control over individual recording parameters of the digital camera. For example, they often lack an external, in particular electronically controllable, trigger.In addition, many modern consumer cameras have integrated automatic functions by default, so that, for example, the exposure of the digital camera is automatically adapted or histogram-based level corrections are carried out by itself. These automatic functions cannot be easily disabled or likewise disabled. Although these automatic functions are advantageous for the usual applications of these consumer cameras, they can make the trouble-free use of such consumer cameras difficult or even prevent for the methods in the foreground here.The above-mentioned procedure is described. Taking these aspects into account, the object is achieved according to the invention by a method according to the appended main claim.In this case, at least one visual marking is integrated into each reference pattern represented, said visual marking preferably containing only a few, in particular the same and highly contrasting colors: For example, ArUCO markers can be used which consist only of black and white pixels, such ArUCO markers having a square geometry with a black edge and an inner grid pattern of black and white boxes which serves to reproduce a numerical identifier in binary code. However, other markers such as QR codes may also be used.The invention has the advantage that it no longer has to be used to assign a displayed reference pattern to a data record derived from an image recorded by the reference pattern using a digital camera via a complex apparatus coupling and synchronization of the reference screen displaying the reference pattern and the mentioned digital camera, but rather can be achieved numerically independently of such a coupling.The problem with the use of corresponding markings is that the pixels used for markings in the reference patterns leave a recess in the reference pattern, which likewise leaves a corresponding hole in the image captured by a digital camera from this reference pattern and in the data record derived therefrom. In order to prevent such holes from disturbing evaluation algorithms that use these data sets with the holes present therein, it is proposed for a preferred development of the invention to vary the position of a marking within a reference pattern during the duration of its display, so that such holes in different data sets that result from reference patterns displayed in a time-offset manner with markings used at different positions relate to quasi different locations.Thus, the effect of the presence of corresponding markers on the final result can be minimized.In principle, it is of course also possible to simultaneously embed either a plurality of markings or a plurality of copies of the same marking at different points in the reference patterns. This makes it possible to achieve a more robust, i.e. less trouble-prone, creation of data sets.In a further preferred embodiment of the invention, a transformation of all pixel values to predefined values is carried out on the basis of the relative gray-scale pixel values of the markings in each data record. As a result, the data sets of recorded reference patterns can be normalized and adjustments to recording parameters of the digital camera used must thus be compensated automatically or manually during the recording or data acquisition.Thus, inexpensive consumer cameras can also be used despite their possibly integrated automatic functions, such as exposure adaptation, level corrections, etc. The effects brought about by these automatic functions can be eliminated, as explained.It is possible that an observation constellation results in distortions in the image captured by a digital camera. These distortions can go beyond a tolerance of the recognition algorithm for the marking. If this distortion can be completely or partially predictable, it is possible to pre-distort a marking before it is embedded in a reference pattern such that the marking appears undistorted in the image.As an example of such predistortion, the rotation or mirroring of a marking can also be mentioned. These pre-distortions may be advantageous if the optical configuration includes a reflection from a surface.Further advantages and features of the invention are evident from the following description of exemplary embodiments. This shows FIG. 1 shows a schematic diagram of a device for camera-based optical measurements, FIG. 2 illustrates typical screen reference digital camera configurations, FIG. 3 shows a schematic diagram of a temporal sequence of displayed reference patterns and images recorded in a synchronized manner by these according to the prior art, FIG. 4 shows a schematic diagram of an improved device for camera-based optical measurements, FIG. 5 shows a schematic diagram of a temporal sequence of reference patterns shown and of images recorded in an unsynchronized manner by these reference patterns, FIG. 6 shows reference patterns with markings contained therein, FIG. 7 shows a recorded image of a reference pattern with a marking contained therein, FIG. 8 shows histogram of gray values of a marking, FIG. 9 shows a schematic diagram of a temporal sequence of reference patterns shown, markings shown therein and images recorded in an unsynchronized manner by these.FIG. 1 shows a device for camera-based optical measurements as a schematic diagram.A control PC 1 sends a sequence of special reference patterns over a line 2 to a connected reference screen 3, which is for example a flat or curved computer monitor. The reference pattern is displayed for a certain predefined period of time. Once the reference pattern appears on the reference screen 3, the control PC 1 causes a digital camera 5 connected to it via a control line 4 to take a photograph. The reference patterns are designed such that a sequence of the grayscale values displayed on a particular screen pixel 6 uniquely identifies its 2D position on the reference screen 3As shown in FIG. 1, the reference screen 3 and the digital camera 5 are optically coupled via an "optical system" 7 (in other words, the digital camera 5 views the reference screen 3).Ideally, a single pixel 8 on the light-sensitive image sensor 9 of the digital camera 5 is intended to record an image of at most a single pixel 6 of the reference screen 3.In data acquisition preferably in the control PC 1, the camera images resulting from the images, which are transmitted again from the digital camera 3 to the control PC 1 via a data line, are decoded in order to establish correspondences between the position of the pixel 8 on the image sensor 9 of the digital camera 3 and the screen pixels 6 on the reference screen 3.Based on this information, the quantities of interest, such as the specific distortions occurring with respect to each pixel 8 on the image sensor 9 in the digital camera 3, can be determined, which can be generated, for example, by optical errors in a camera lens 11. Or, the shape of a reflecting surface 12 under examination or a refractive object 13, etc. is derived. Corresponding optical configurations are shown in FIG. 2: on the left, a digital camera 5 is directly directed onto a reference screen 3. In the middle, the deflectometry is shown, in which a digital camera 5 monitors a reflection of a reference screen 3 that occurs in a specular surface 12 examined in the process. Furthermore, there is also the variant of optical deflectometric tomography shown on the right in FIG. 2, in which a digital camera 5 monitors the reference screen 3 as it appears through a transparent object 13 examined in this case.Since the control PC in FIG. 1 controls both the digital camera 5 and the reference screen 3, the association between the recorded images and the displayed reference patterns is easy to establish, as shown in FIG. 3: FIG. 3 shows the times 14, 15, 16 in which different reference patterns are displayed on the reference screen 3. Once a reference pattern is displayed, the camera 5 is caused to take a shot. The times 17, 18, 19 required for exposure of the image are likewise shown in FIG. 3. Each time an exposure 17, 18, 19 is completed, the reference pattern 14, 15, 16 can be changed. The darker blocks in Figure 3 in the pattern of reference pattern display times represent transition periods 20, 21, 22: When the program instructs the operating system to change a reference pattern on the reference screen 3, it takes some time for its LCD pixels to reach their target intensities. It is therefore expedient, after the changing of a reference pattern, to plan a certain period of time for the corresponding transition period 20, 21, 22 before the digital camera 5 is triggered and takes a recording during the exposure times 17, 18, 19.As can be seen from FIG. 3, since the control PC 1 controls both the digital camera 5 and the reference screen 3, the correspondence between the captured images and the displayed reference patterns is easy to establish: Once a reference pattern is displayed (times 14, 15, 16), the digital camera 5 is caused to establish a capture or an image. When the exposure (exposure times 17, 18, 19) on the digital camera 5 side is completed, the reference pattern on the reference screen 3 side can be changed.In order to simplify the use of corresponding measurement systems, a further developed device for camera-based optical measurements is proposed according to FIG. 4, in which the structure shown in FIG. 1 is reconfigured. Incidentally, in FIG. 4, the same parts as in FIG. 1 are denoted by the same reference numerals.The device according to FIG. 4 comprises a plurality of components, in particularan "encoder" 23 is a computer system which generates reference patterns sent to it as data as image files 24;a "pattern display" 25 is a device which displays a sequence of reference patterns on a reference screen 3 which is fed to it as an image file 24;a "recorder" 26 - a device which records video / time-lapse sequences ("raw data") as raw data files 27 with the aid of a digital camera 5;a "decoder" 28 - a computer system which receives raw data files 27 and creates the final set of geometric correspondences between pixels 6 on the reference screen 3 and pixels 8 on the image sensor 9.Each of these components can be implemented in many ways, but can generally be more simply constructed than in the original monolithic system. For example, the pattern display 25 may be realized as a flat panel connected PC, a tablet, or a smart TV.The task of the pattern display 25 is to display reference patterns one after the other with (approximately) constant delays over short periods 14, 15, as shown in FIG. 5. A strict control of latencies is not provided in this case. Therefore, in this embodiment, with respect to this apparatus, there is no need to ensure special display control.Moreover, each stand-alone digital camera may function as the recorder 26. No external triggering is required, but rather only the possibility of storing files in standard formats as raw data 27 on removable media or in a network.Finally, the encoder 23 and the decoder 28 can be implemented both as separate systems and can be executed on the same computer. They take over extensive data processing, but do not require any additional preconditions apart from access to the image files 24 of the reference patterns and the raw data files 27, for example also via a network.The corresponding time diagram is shown in FIG. 5. The files at the exposure times 29 to 35 cannot in this case simply be related to the periods 14, 15 of the representation of reference patterns: some of the images recorded may not capture reference patterns (e.g. at times 29 or 35), or several images capture the same reference pattern (cf. times 30 and 31), and some images are recorded during the transition between two reference patterns (time 32) and thus contain undefined pixel values not suitable for decoding.According to the invention, the explicit synchronization explained above between the reference patterns and the files derived therefrom is now replaced by a virtual / numerical synchronization. For this purpose, the reference patterns are provided with an optical marking and the generation of the raw data is supplemented by a corresponding identification algorithm for the markings.The method now proceeds as follows:First, the encoder 23 generates a sequence of reference patterns as image files 24. Each visual marker 37, 38 uniquely identifies a displayed reference pattern 35 or 36 within the sequence of reference patterns. The dimensions of the marking 37, 38 are sufficiently large in this case to ensure trouble-free detection, but sufficiently small in order to conceal only an insignificant part of the reference pattern 35 or 36 to be detected by the digital camera. The type and type of markers is selected so that their recognition in the recorded images is not affected by typical distortions caused by the optical system between the reference screen 3 and the digital camera 5 in the respective application.The pattern display 25 then represents the sequence of marked reference patterns continuously in a loop. The display time 14, 15 for each reference pattern is chosen arbitrarily, but generally it should exceed the combined gray-to-gray transition time of the reference screen 3 and the frame exposure in the downstream recorder 26.The recorder 26 stores a video stream or sequence of time-lapse images of the digital camera 5 as raw data 27 without explicit synchronization with the reference pattern display. Depending on requirements and constraints, the collected raw data files 27 may include one or more repetitions of the displayed sequence of reference patterns. It is also assumed that the field of view of the digital camera 5 always captures a visual mark 37 or 38 embedded in the reference patterns 35, 36. An example of an image taken by a digital camera is shown in FIG. 7.Finally, the decoder 28 receives the raw data files 27 and processes them as follows:First, the decoder 28 checks each recorded raw or video image successively or in parallel and tries to recognize markings 37, 38 in the corresponding raw data files 27. If no mark 37 or 38 is found, the corresponding data set is discarded. A successfully recognized marking 37 or 38 determines the associated reference pattern which is then to be associated with the data record. All data sets that can be assigned to a reference pattern are collected in a common buffer memory.At the same time, accumulation is made by adding the collected files to a single integrated image. The resulting pixel values in the buffer are obtained as the sum of the pixel values in all original files.As soon as all raw images have been processed and all reference patterns have been recognized in a sequence, the images or the files belonging to them are normalized, i.e. reduced to the same brightness and contrast levels.If a digital camera 5 has fixed exposure and sensitivity settings during acquisition, it is sufficient to simply divide the pixel values of the integral image by the number of images accumulated. Thus, the normalized image is the average of the respective raw images for a particular reference pattern.The normalized images or files thereof are then treated as camera images of a digital camera with default settings and in a regular synchronous setup with a known relationship between images and reference patterns. They are then decoded using the usual algorithms corresponding to the selected reference pattern sequence type.As explained, automatic or manual adjustments to camera parameters during data acquisition can interfere with a decoding algorithm and result in an erroneous result. It should be appreciated that, for example, digital cameras in modern smartphones may use rather complex linear and non-linear "image enhancement techniques", which are not always fully documented.In the following, only a basic image transformation is considered, namely a brightness / contrast adaptation.Mathematically, given the "original" incident intensity g o on a pixel on the image sensor, the digital camera records and stores a value g f= A g o+ B, where the coefficients A and B are uniform and constant across all pixels, but may vary from one image to another. Each change in parameters A and B changes the recorded black and white levels.This transformation can serve, for example, to avoid excessive over- and underexposure effects and makes better use of the dynamic range of an image sensor. Many consumer cameras implement appropriate automatic exposure and / or gain stage controls. However, the methods focused here require that the brightness and contrast remain constant throughout the sequence in all images, or in other words that the "black" and "white" values displayed on the reference screen each lead to the same results in the camera sensor, regardless of the reference pattern displayed.To undo such automatic or manual level adjustment and reduce all files to the same black and white values, the embedded markers used in the method of the invention are used.The prerequisite is that the marking contain only a few solid and very different colors: for example, ArUCO markers consist only of black and white pixels, which are optionally also respectively combined in "boxes".In principle, all images generated by the digital camera are first recorded as raw data.When first reviewing this raw data, a detector then not only identifies the markings, but also finds their corners in the image. This information makes it possible to restrict a substantially square area in the image in which only white and black pixels can be expected (see FIG. 7 and the mark area at the center of the frame).Thereafter, a histogram of the relative gray level pixel values will be prepared only for this square region, as shown in Fig. 8. The histogram is expected to contain two distinct peaks 39, 40 - one for the pixels which are black in the displayed image and the other for those which are white in the displayed image.As explained with reference to FIG. 6, a sum of two Gaussian peaks can then be adapted to this histogram (possibly with a linear base) and thus the most likely "black" and "white" levels in the image are recognizedThe adaptation can use various non-linear optimization techniques such as the Levenberg-Marquardt algorithm, but is usually quite simple and quickly completed. It is trivial to derive the coefficients of a linear transform that rescales these two levels and shifts them to some predefined points (e.g., 0.1 for "black" and 0.9 for "white" on the relative scale).The quality of the data sets is thus improved considerably and counteracts the influences which many digital cameras perform in practice.The raw data can then be stored on the basis of assignment lists which assign recorded images to, for example, specific camera positions and positions and reference patterns detected in the images.A further possibility for improving the quality of the data evaluation is described with reference to FIG. 9.One problem could be that individual images, such as the image with exposure time 32 in Figure 5, have "scrap" images produced since they do not represent a particular reference pattern and are therefore useless to the decoder. When each reference pattern is acquired in many images, the above-described accumulation and normalization steps significantly suppress the effects of such "scrap" images on data quality by disregarding the images listed in the assignment lists that cannot be associated with a particular reference pattern in further data processing.The explained storage of raw data, the evaluation thereof as described and the partial disregard thereof is referred to within the scope of this invention as "book-keeping", wherein this relates both to the individual steps and to the totality thereof.In an evaluation as explained above, it must also be taken into account that the marking embedded in a reference pattern is also influenced by the transition effects that occur on the reference screen during its fade-in or fade-out. In data processing, images in which the marking is not recognized are discarded as part of book-keeping and then no longer have any influence on the processing of "good" images.With the following steps, the effect of corresponding "waste" images can be further reduced, in particular within the scope of book-keeping.As shown in FIG. 9, the reference patterns are displayed at times 14, 15, and 16, etc. as shown in FIG. 5. However, the embedded markers are only displayed for shorter marker periods 41, 42, 43 that do not overlap with transition periods 20, 21, 22.As already explained and as can be seen in FIG. 9, the fade-in and fade-out of markings is also associated with a few marking transition periods 44. However, these have an effect only on the pixels in the marking itself, while the pixels of the reference pattern surrounding the marking retain their values. Therefore, "normal" images recorded at exposure times 47 or 50 are decoded as usual, "waste" images recorded at exposure times 45 or 52 are discarded due to missing markings, in particular during book-keeping. Images recorded at exposure times 46, 48, 49 or 51 during the fade-in or fade-out of a mark can either be discarded if the mark is too weakly recognizable in the image, or they contribute to the above-explained accumulator with the remaining pixel values relevant to the reference pattern. In either case, there will be no adverse effects on the subsequent decoding operation.It should be mentioned that the device provided for the method described above can also be realized in the "cloud". The communication connections between the components can then be selected depending on the application. For example, the reference pattern display may be implemented as a web application executing on a web browser, where the reference patterns are received from a server via the HTTP protocol and displayed on the reference screen using standard browser functionality. The same application may allow camera files to be uploaded back to the server via the same HTTP protocol. In this way, additional software need not be installed locally in addition to a web browser, and the compute-intensive components can be performed on the server with the most efficient hardware and software resources.The proposed architecture and the described algorithms can be modified as follows:It is possible to embed several marks or several copies of the same mark at different locations in the reference patterns to achieve a recognition that operates more reliably.In addition, it is possible to change the position of the embedded mark in each displayed reference pattern. As long as the marker remains visible, a detector will find it, but such "jumping" markers and the correspondingly altered accumulation logic can reduce the number of pixels ("holes") "lost" by markers in the resulting data sets.If the selected observation constellation leads to a distortion of the camera image which goes beyond the tolerance of a mark detector, but is partially predictable, it is also possible to pre-distort the mark before embedding it in a reference pattern such that it appears undistorted in the camera image. This is particularly advantageous if, for example, the optical scheme comprises a reflection from a surface.In the above discussion, it is assumed that the optical system between the screen and the digital camera remains static and that the data acquisition aims to generate a single data record. However, if the camera position or some optical parameters change, it is possible to capture multiple data sets in a single session within the scope of book-keeping. The change in the parameters (i.e. the need to process a new data record) can then be detected on the basis of the changes in the detected marking positions, in particular via their corner coordinates, which exceed a specific tolerance level.Finally, it should be noted that instead of accumulating an integral image, the aggregation of synthetic images may follow a different logic. For example, if the noise in the data is negligible, averaging based on an integral image does not provide an advantage. Instead, it is possible to select a single representative image and take it into account for the subsequent normalization.List of reference characters1 Control PC 2 Line 3 Reference screen 4 Control line 5 Digital camera 6 Screen pixel 7 Optical system 8 Sensor pixel 9 Image sensor 10 Data line 11 Camera lens 12 Reflective surface 13 Transparent object 14 Time for reference pattern 15 Time for reference pattern 16 Time for reference pattern 17 Exposure time 18 Exposure time 19 Exposure time 20 Transition period 21 Transition period 22 Transition period 23 Encoder 24 Pattern display 25 Image files 26 Recorder 27 Raw data files 28 Decoder 29 Exposure time 30 Exposure time 31 Exposure time 32 Exposure time 33 Exposure time 34 Exposure time 35 Reference pattern 36 Reference pattern 37 Visual mark 38 Visual mark 39 Peak value 40 Peak value 41 Mark period 42 Mark period 43 Mark period 44 Mark transition period 45 Exposure time 46 Exposure time 47 Exposure time 48 Exposure time 49 Exposure time 50 Exposure time 51 Exposure time 52 Exposure time 53 Exposure time

Claims

Method for a camera-based optical measurement, in which a reference screen (3) and a digital camera (5) are coupled in an optical configuration, wherein reference patterns (35, 36) are displayed on the reference screen (3) and recorded by the digital camera (5) as a camera image and the recorded camera image is decoded and evaluated as a data set (27) by a computer (28), characterized in that at least one visual marking (37, 38) is integrated in the reference pattern (35, 36).Method according to claim 1, characterised in that the marking (37, 38) contains only a few, in particular the same and strongly contrasting colours.Method according to one or more of the preceding claims, characterized in that the position of a marking within a reference pattern (35, 36) is varied during the duration of its display (14, 15, 16).Method according to one or more of the preceding claims, characterized in that several marks or several copies of the same mark are displayed at different locations in the reference pattern.Method according to one or more of the preceding claims, characterized in that the at least one mark is pre-distorted before being displayed in the reference pattern.Method according to claim 5, characterised in that the pre-distortion is a rotation or a mirroring of the marking.Method according to one or more of the preceding claims, characterized in that in a data record (27), on the basis of the relative grey level pixel values (39, 40) of the marking (37, 38), a transformation of all pixel values to predefined values is carried out.

Images