METHOD FOR DETERMINING THE POSITION OF A FIRST IMAGE AREA IN A CORRESPONDENCE IMAGE, SOC AND CONTROL DEVICE AND SYSTEM FOR EXECUTING THE METHOD, AS WELL AS COMPUTER PROGRAM

Description

[0001] The invention relates to a method for determining a first image region in a correspondence image. Furthermore, the invention relates to a system-on-a-chip (SoC) and a control device for carrying out the method. The invention also relates to a system, a computer-readable storage medium, and a computer program, each configured to execute the proposed methods. State of the art

[0002] Several methods are known in the art for determining the position of a first image region in a corresponding image, for example, so-called "block matching." In this method, the first image region is compared with all potentially matching second image regions of the same size from the corresponding image. In each of these comparisons, all pixel pairs are compared. Each comparison evaluates the similarity or difference of the pixels under consideration, for example, an absolute difference in grayscale values. By summing the absolute differences, a score is obtained for each potential second image region, with the sum being higher the greater the difference between the image regions. Block matching is computationally very intensive, as the computational effort increases with both the dimensions of the first image region and the dimensions of the corresponding image.Furthermore, block matching is not very reliable, as the result is not very discriminative, i.e., there is no clear extremum and therefore no unambiguous position can be determined. Geänderte Beschreibungsseite 2 - Reinstatement

[0003] Another possibility is the so-called "feature comparison," i.e., comparing at least one feature extracted from the first image region with features extracted from the corresponding image. This generates sequences of binary values, called signatures, that characterize the corresponding image regions. The approach is based on the assumption that a similar signature of image regions indicates similar image regions. The signatures are then compared by bitwise comparing the first signature with the second signature (XOR comparison) and counting the bits that differ. The smaller the difference, the higher the correspondence between the image regions. However, comparing signatures can also be computationally intensive, as a comparison of signatures is required for each possible position in the corresponding image, with two signatures (for example, of 256 bits in length) having to be compared bitwise at a time.Furthermore, the problem here too may be that the result of the feature comparison is not sufficiently discriminatory.

[0004] German patent application DE 103 51 778 A1 discloses a method for processing image data of moving scenarios. For this purpose, corresponding pixels or image areas are identified in individual, temporally successive image data sets.

[0005] WO 2005 / 048195 A2 describes the processing of image data from moving scenes, in particular for the recognition and tracking of objects within them. This requires identifying corresponding pixels or image areas within the individual, sequentially generated image data sets.

[0006] DE 10 2010 006522 A1 describes a method for analyzing corresponding image points (B1, B2) in at least two image data sets, in which signature information is determined for each image point (B1, B2) of the image data sets, whereby a correspondence hypothesis is created to determine corresponding image points (B1, B2) for signature information that matches in both image data sets. Geänderte Beschreibungsseite 2a - Reinfassung

[0007] In DE 103 56 090 B3, a method for the computer-aided determination of image areas in a digitized image is disclosed, in which a first image area (B1) in a reference image (RB) is compared with several second image areas (B2) in a search area (S) of a search image (SB) and as a result a second image area (B2) is determined which has the highest similarity with the first image area (B1), and in which the determined second image area (B2) is marked if the determined second image area (B2) is located at the edge of the search area (S). Disclosure of the invention

[0008] The invention relates in a first aspect to a method for determining a position of a first image area in a correspondence image according to claim 1.

[0009] The invention solves the problem of locating a position of a first image area in a corresponding image. This problem frequently arises in the fields of image processing, image analysis, machine vision, computer vision, etc. For example, the method according to the invention can be used for the following applications: Indirect distance measurement, wherein distances to at least two distant points are successively measured using a handheld laser distance meter. The laser distance meter indirectly determines the distance between the two points from the two distances and synchronously recorded camera images, and outputs this distance. According to the inventive method, it is possible to locate the first image area captured when aiming at the first measuring point in the corresponding image captured when aiming at the second measuring point and to determine its position. This makes it possible to compensate for errors caused by translation and / or rotation of the handheld laser distance meter, thus eliminating the need for a tripod. DE 10 2018 201 692 A1 is known to those skilled in the art regarding this measuring principle. Tracking a stationary object while a camera moves past the object.For example, a traffic sign can be tracked in this way as a vehicle passes it. This method is used to obtain multiple observations of the same traffic sign, which have different sizes or resolutions, thus enabling more reliable reading of the traffic sign. Tracking a moving object with a camera, for example, a pedestrian, in a sequence of camera images from a moving camera in a vehicle or as part of surveillance cameras. Successive evaluation of image shifts during the navigation of autonomous robots, which, as they move relative to a surface (e.g., concrete or grass), take pictures of this surface with a ground-facing camera, thereby enabling precise localization of the autonomous robot.

[0010] In one embodiment, the method is implemented exclusively in hardware, in particular in the form of a system-on-a-chip (SoC).

[0011] In an alternative embodiment, the method is also implemented in the form of software or in a mixture of software and hardware. In particular, the method is a computer-implemented method that is carried out by a control device, for example, a processor device, of a computer device. For carrying out the method, the control device further comprises at least one storage device in which the method is stored as a machine-readable computer program.

[0012] The term "provide" means, in particular, that the relevant information—here, at least a first image area and a corresponding image—is transmitted to a device performing the method, especially a control device or a system-on-a-chip (SoC), for further processing. In one embodiment of the method, the relevant information can be provided by reading corresponding data from a file, for example, by reading data stored on a data server. In a further alternative or additional embodiment of the method, a first image area and / or a corresponding image can be acquired and provided using at least one imaging sensor device. In particular, a

[0013] Provision of the image using an optical camera, for example a visual camera and / or a thermal imaging camera, or using another imaging sensor – for example, a radar sensor, NMR sensor, 3D time-of-flight sensor, or the like – is conceivable. Furthermore, it is conceivable that a first image area is provided using a first imaging sensor device, in particular a first camera, and a corresponding image is provided using a second imaging sensor device, in particular a second camera. In particular, the first imaging sensor device and the second imaging sensor device are of the same type (for example, visual cameras). A defined (in particular, fixed) geometric relationship can exist between the two sensor devices.Alternatively or additionally, it is conceivable that both the first image area and the corresponding image are provided using the same sensor device. For example, acquisition and provision by the same sensor device at different times and / or from different perspectives is conceivable.

[0014] The invention relates to the efficient location of a first image area, in particular a first image area from a source image, in a corresponding image. The corresponding image can be another image or the source image itself. If the first image area corresponds to at least one other image area in the corresponding image, the first image area can be located. Locating the first image area includes determining its position within the corresponding image. "Determining a position" means that the first image area can be uniquely located in the corresponding image by specifying a position—for example, in the form of pixel coordinates. This method is also referred to as "correspondence finding." The first image area can be significantly smaller than the corresponding image.It is conceivable, for example, to determine the position of a first image area comprising 41×51 pixels in a corresponding image comprising 2560×1440 pixels. In particular, it is also possible to determine the position of a single pixel in the corresponding image, whereby a first image area is still used for carrying out the method according to the invention, wherein the first image area includes a, in particular small, surrounding area of ​​the pixel.

[0015] The first image area and the corresponding image comprise a plurality of image positions, which can be designated by (xi, yi) in the first image area and by (uj, vj) in the corresponding image. These image positions can be defined, in particular, by the pixel resolution of the first image area or the corresponding image. Alternatively, other signal-related subdivisions of the first image area or the corresponding image into distinguishable image positions are also conceivable. In particular, it is possible that a transformation (for example, scaling, rotation, so-called warping (especially image rectification for lens correction), or the like) is applied to a signal-related subdivision of the first image area or the corresponding image. Furthermore, a signal-related subdivision can already result from the sensor device itself, for example, in the case of a detection array such as a phased array radar.

[0016] The first image region comprises at least one reference position (for example, denoted as (xr, yr)) to which the determination of the first image region's position in the corresponding image refers. For example, if it is determined that the first image region is located at position (x0, y0) in the reference image, then the reference position of the first image region is indeed located at this point. The reference position can, in principle, be chosen arbitrarily. In particular, the reference position can be chosen as a pixel of the first image region. For example, the reference position can be chosen as the geometric center of the first image region. Alternatively, the reference position can also be chosen as a corner point or another distinctive point of the first image region—for example, depending on the geometry of the first image region.It is also conceivable that the reference position is not located at an integer pixel position, since it can, in principle, be located at any (for example, calculated) position between the pixels. It should be noted that any image position (xi, yi) in the first image region with respect to the reference position can be described by a relative offset (Δx i, Δy i). The expression (Δx i, Δy i) explicitly includes the description of the offset in different coordinate systems, for example, in Cartesian and non-Cartesian (e.g., polar coordinates) coordinate systems. Furthermore, the formation of a signature of an image position in the first image region can also refer to the reference position (see explanation below).

[0017] In one embodiment of the method, it may be provided that the first image area and / or the corresponding image are first adjusted after they have been provided, in particular by filtering, smoothing, transforming (linearly or non-linearly) into grayscale images, scaling, rotating, distorting, changing the resolution or the like.

[0018] According to the invention, in a process step, a plurality of signature values ​​si are determined for different image positions (xi, yi) in the first image area. A signature value (also known as a signature) refers in particular to a feature determined from data of a provided image (or image area) for a (selected or defined) image position (xi, yi), which characterizes an environment of the associated image position (xi, yi). Thus, signature values ​​determined for image positions have an assignment, in particular a fixed assignment, to the corresponding image position, especially to the coordinates of this image position. The signature values ​​characterize the respective assigned image positions by a computational description of the environment of the assigned image position. Preferably, the signature values ​​are each determined by one or more simple arithmetic operations from the adjacent data content (in particular image information or...The image data of the assigned image position is determined, for example, by multiple, particularly weighted, differentiation and / or summation of several gray values ​​of immediately adjacent or nearby image positions. In weighted differentiation, the weights can depend on the relative offset. The creation of such signatures, in particular different methods for generating or determining such signature values, is sufficiently known to those skilled in the art, for example under abbreviations such as "Scale-invariant feature transform (SIFT)", "Speeded Up Robust Features (SURF)", "Binary Robust Independent Elementary Features (BRIEF)", "Oriented FAST and Rotated BRIEF (ORB)" or the like.In one embodiment, a signature value can be generated from the first image area by checking which of the two gray values ​​is greater for two pixels located in a predefined relative arrangement within the image area, and encoding the result as 0 or 1 in binary. This check is repeated for a number of further pixel pairs, thus generating a sequence of binary values, for example, consisting of 8, 16, or 24 bits. This sequence of binary values ​​characterizes the image area and is called the signature value (or signature). The signature value can also be interpreted as a decimal number. The approach of determining the position of the first image area in the corresponding image using signature values ​​is based on the assumption that a similar signature value in the first image area indicates a similar image area in the corresponding image, and vice versa.Signature values ​​can be compared in particular by bitwise comparison (XOR comparison) and counting the bits that differ (in this case, a dissimilarity is measured).

[0019] According to the invention, a signature value is not only generated for the reference position, but for a plurality of image positions in the first image area. For example, signature values ​​are determined for all image positions defined by pixels in the first image area, in particular for more than 90%, more than 75%, more than 50%, or especially for more than 30% of all image positions defined by pixels in the first image area. The plurality of signature values ​​si for different image positions (xj, yj) in the first image area can be determined in the form of a matrix, a table, or a signature value image. The matrix is ​​referred to as the first signature value matrix. A signature value si is assigned to a corresponding image position (xj, yj). The matrix, table, or signature value image can have the same dimensions as the first image area.

[0020] In one embodiment of the method, ambiguities in the majority of the determined signature values ​​si can be resolved by deleting and / or combining the ambiguous signature values ​​si. Deleting ambiguous signature values ​​si thus excludes them from further processing. Deleting ambiguities enables a particularly simple method, although this results in some loss of information. Combining ambiguities can be achieved, in particular, by weighted averaging relative offsets.

[0021] In one process step, relative offsets (Δx i ,Δy i ) of a plurality of image positions (xi ,yi ) in the first image area are assigned to the respective determined signature values ​​si. In particular, the relative offsets (Δx i ,Δy i ) of a plurality of image positions (xi ,yi ) in the first image area can be assigned to the respective determined signature values ​​si in an assignment rule, especially in a lookup table. It is conceivable to determine each relative offset (Δx i ,Δy i ) of an image position (xi ,yi ) with respect to the reference position from a matrix of relative offsets (Δx i ,Δy i ). The relative offsets can, in particular, be fixed and immutable, or stored.

[0022] In one embodiment of the method, an address of the mapping rule, in particular the lookup table, is determined using a respective signature value si. In this way, a particularly efficient and space-saving mapping can be performed in the mapping rule based on the signature values, whereby a corresponding entry (i.e., at least a relative offset (Δx i ,Δy i )) is located at each signature value. In one exemplary embodiment, the mapping rule, for example the lookup table, has a number of addresses corresponding to the number of unique (in particular possible or representable) signature values. For example, with a signature length of n bits, there are 2n addresses.Since the addresses themselves do not require any storage space - as they only need to be implicitly and fixedly present in ascending order in a memory allocation - a particularly resource-efficient method can be advantageously specified.

[0023] The method according to the invention further comprises a process step in which a plurality of correspondence signature values ​​sj are determined for different image positions (uj, vj) in the correspondence image. The same explanations apply as above regarding signature values ​​and image positions of the first image area. In particular, a correspondence signature value sj characterizes a neighborhood of an image position (uj, vj) in the correspondence image. Specifically, a correspondence signature value sj for an image position (uj, vj) in the correspondence image is formed analogously, and in particular according to the same procedure, as a signature value si for an image position (xi, yi) in the first image area. Furthermore, the plurality of correspondence signature values ​​sj for different image positions (uj, vj) of the correspondence image can be determined in the form of a matrix, a table, or a correspondence signature value image. The matrix is ​​then referred to as the second signature value matrix.This is particularly conceivable in a software implementation. Alternatively, it is also conceivable to forgo (even temporary) storage of individual correspondence signature values ​​sj and to process the correspondence signature values ​​sj immediately after they have been determined. Particularly in a hardware implementation of the method according to the invention – for example, in a SoC – it is therefore possible to specify a particularly memory-space-reduced and thus resource-efficient implementation.

[0024] In one embodiment of the method, a signature value si and a corresponding signature value sj have a predetermined length of 24 bits or less, in particular 16 bits or less, and most especially 8 bits or less. This reduces the number of arithmetic operations and the computation time required to determine the signature values. Furthermore, the size of the mapping rule, particularly the lookup table, can be very small, since its size grows exponentially (i.e., 2n < ) with the number of bits n. In one embodiment, a signature value of length 10 bits is implemented.

[0025] Using the available correspondence signature values ​​sj, relative offsets (Δu j ,Δv j ) to determined correspondence signature values ​​sj can then be calculated using the previously established mapping. For this purpose, the entry in the mapping rule corresponding to the correspondence signature value sj is read out, i.e., by searching for the address based on the correspondence signature value sj.

[0026] Once the offsets have been determined, target positions (Δuj + uj, Δvj + vj) in the corresponding image can be calculated. Each target position determines the image position of the reference position in the corresponding image, based on different correspondence signatures. The calculation is performed by compensating the offset (i.e., if the offset is defined, by addition) of the relative offset (Δuj, Δvj) and the respective image position (uj, vj) belonging to the corresponding correspondence signature value sj. In one embodiment of the method, correspondence signature values ​​are determined for all pixel-defined image positions in the corresponding image. In an alternative embodiment of the method, correspondence signature values ​​are determined for more than 90%, for more than 75%, for more than 50%, or, most importantly, for more than 30% of all pixel-defined image positions in the corresponding image.

[0027] The term "incremental assignment of a weighting value w" refers specifically to the incremental (i.e., additive) assignment of a weighting value w to a calculated target position. If a weighting value is already assigned to the target position, it is increased by the newly assigned weighting value w to form the incremental weighting value w'. The weighting values ​​w can be assigned to target positions in the corresponding image in the form of a matrix or table. The matrix is ​​also referred to as a reconciliation matrix. In a further step, the position (u₀, v₀)* of the first image area in the corresponding image is determined by finding an extremum, particularly a maximum, in the incremental weighting values ​​w'. It may be possible to further process the determined incremental weighting values ​​w', for example, by filtering and / or smoothing, so that the result of the determination becomes more unambiguous.In particular, isolated extrema (e.g., peaks) in the incremental weighting values ​​w' can be suppressed at this stage of the process, while the extremum actually being sought, especially the maximum, stands out more clearly. Furthermore, precision can be improved, as smoothing at least partially suppresses noise in the determined incremental weighting values ​​w'.

[0028] In one embodiment of the method, the assigned weighting value w is selected depending on the relative offset (Δu j ,Δv j ), wherein the weighting value w increases or decreases, in particular, with increasing relative offset. In an alternative embodiment of the method, the assigned weighting value w is selected depending on the relative offset (Δx i ,Δy i ), wherein the weighting value increases or decreases, in particular, with increasing relative offset. Preferably, a weighting value w is higher the closer the image position and the reference position are to each other and / or the more unique the signature value is within the first image area. Thus, the weighting values ​​w can be high at those image positions where there is a high similarity to the target position being sought, and low values ​​at the remaining (dissimilar) image positions. A respective weighting value can be calculated or also read from a matrix or table.In this way, the relative position of the image position, to which a corresponding signature value belongs, is advantageously integrated into the determination of the target position.

[0029] In one embodiment of the method, a position (u 0 ,v 0 )* of the first image area in the corresponding image is determined when the incremental weighting value w' assigned to the position reaches a relative or absolute threshold. This allows for a particularly reliable determination of the position. Especially in combination with the previously mentioned filtering and / or smoothing, this results in a particularly robust method.

[0030] In one embodiment of the method, a relative offset (Δu j ,Δv j ) to determined correspondence signature values ​​sj is calculated using the assignment and a transformation quantity, in particular a scaling quantity and / or a rotation quantity and / or a translation quantity and / or a shear quantity. In one embodiment, the calculation including the transformation quantity can be performed by multiplying the relative offset (Δu j ,Δv j ) (which represents a 2-dimensional vector) by a (2×2) matrix. In particular, scaling can thus be performed, for example, depending on a change in the resolution of the first image area and the correspondence image and / or depending on a change in the object distance in the first image area and in the correspondence image.It is therefore acceptable that an object imaged in the sensor device has increased in size (r > 1), decreased in size (r < 1), or remained the same size (r = 1) by a factor r between the provision of the first image and the corresponding image. Prior knowledge about this relationship (r) can, for example, be predicted from previous observations or determined from other available information (e.g., laser distance measurement data) and incorporated into the method according to the invention. u 0 υ 0 = u j υ j + r ⋅ Δ u j Δ υ j .

[0031] This makes it advantageous to ensure that an extremum, especially a maximum, to be determined for voting purposes is particularly clearly defined.

[0032] Furthermore, a relative rotation of the first image region and its corresponding image relative to each other can be compensated. For this purpose, the scalar factor r can be replaced by a 2×2 matrix. u 0 υ 0 = u j υ j + r 1 , 1 r 1 , 2 r 2 , 1 r 2 , 2 ⋅ Δ u j Δ υ j

[0033] In this way, it is possible to perform rotation, stretching, compression, and / or shearing in addition to scaling. Such transformations are of high practical relevance, especially when the perspective between the first image area and the corresponding image has changed.

[0034] In one embodiment of the method, ambiguities are removed by deleting and / or combining the ambiguous signature values.

[0035] In a further aspect, the invention relates to a system-on-chip (SoC) configured to carry out the method according to the invention. In particular, the SoC is configured to acquire at least a first image area and a corresponding image by means of at least one imaging sensor device and to generate an output signal depending on the determined position (u 0 ,v 0 )* of the first image area in the corresponding image.

[0036] In a further aspect, the invention relates to a control device configured to carry out the method according to the invention. In particular, a control device comprising at least one processor and a storage device is proposed, wherein instructions are stored on the storage device which, when executed by the processor, cause the control device to execute the method according to the invention. A "processor" is understood to be, in particular, a device comprising at least one information input, an information processing unit for processing, and an information output for transmitting the processed and / or evaluated information. In one embodiment, the processor comprises at least one processor.A "storage device" serves to keep a computer program necessary for carrying out one of the described procedures available to the processor device.

[0037] Finally, a further aspect of the invention relates to a system comprising at least one control device and / or a system-on-a-chip (SoC) according to the invention and at least one imaging sensor device, for example a camera, which is configured to capture and provide a first image area and a corresponding image. In particular, the invention relates to a vehicle, a laser distance measuring device, and an autonomous robot with the SoC and / or the control device according to the invention and at least one imaging sensor device.

[0038] In another aspect, the invention relates to a computer program comprising instructions which, when executed by a processor device, cause a control device to execute the method according to the invention. The computer program includes instructions that cause a control device to execute the method according to the invention with all its steps when the computer program is executed.

[0039] Another aspect of the invention relates to a computer-readable storage medium on which the computer program according to the invention is stored, in particular saved. Storage media themselves are known to a person skilled in the art.

[0040] In the following, "intended" and "configured" are specifically to be understood as "programmed," "designed," "conceived," and / or "equipped." The fact that an object is "intended" for a specific function means, in particular, that the object fulfills and / or executes this specific function in at least one application and / or operating state, or is designed to fulfill the function.

[0041] The method according to the invention is a particularly efficient and therefore processor-resource-saving method, in which only very few computational operations are required to determine the position in the correspondence image. Furthermore, the proposed method is particularly discriminative compared to prior art methods and therefore exhibits particularly high reliability. These advantages are achieved in particular by the fact that The first image area is described with multiple signature values, whereby a signature value is determined not only for the reference position but for several image positions in the first image area; an assignment rule (in particular a lookup table) is used for fast access, wherein the address for access is directly based on the signature values, in particular corresponds to the signature values ​​or can be determined from the signature values ​​with simple operations, for example by binary masking, binary XOR operations, concatenation, changes in the bit order or the like; signatures for image positions in the first image area are used to determine the assignment rule once; signature values ​​of the correspondence image are used only for reading from the assignment rule, in particular the lookup table;The assignment rule stores relative offsets that make it possible to deduce a target position from a given image position to a corresponding signature by calculating the relative offset; a weighting value w is assigned at the respective target position, whereby an incremental weighting value w' is determined, and the position where the first image area is located in the corresponding image can be found by determining an extremum in the incremental weighting values ​​w'.

[0042] The process can be made even more efficient by using particularly short signature values ​​(and correspondence signature values), for example with a length of less than 24 bits. Drawings

[0043] The invention is explained in more detail below with reference to exemplary embodiments illustrated in the drawings. The drawing, the description, and the claims contain numerous features in combination. A person skilled in the art will expediently consider the features individually and combine them into meaningful further combinations. Identical reference numerals in the figures denote identical elements.

[0044] They show: Figure 1 shows a schematic view of an embodiment of a system according to the invention in the form of a laser distance measuring device, which is located in an exemplary measuring constellation; Figure 2a-2 shows a schematic illustration (a) of a first image area and (b) of a corresponding image starting from the exemplary measuring constellation. Figure 1 Figure 3 shows a schematic representation of an embodiment of the method according to the invention, Figures 4a-4f show the method being carried out. Figure 3 Matrices used and / or created: (a) matrix of signature values, (b) matrix of relative offsets, (c) matrix of weighting values, (d) allocation rule, (e) matrix to explain the voting process, (f) matrix of incremental weighting values, Figures 5a-5b schematic, pictorial representations of voting matrices, Figure 6 a schematic representation of a system-on-a-chip according to the invention. Description of the exemplary implementations

[0045] Preliminary note: The previously widely used expressions (xi ,yi ) to an image position in the first image area, (uj ,vj ) to an image position in the correspondence image, (Δx i ,Δy i ) or (Δu j ,Δv j ) for a relative offset with respect to the reference position, si for a signature value to an image position in the first image area, sj for a correspondence signature value to an image position in the correspondence image, (u 0 ,v 0 ) = (uj ,vj ) + (Δu j ,Δv j ) = (Δu j +uj ,Δv j +vj ) for calculated target positions in the correspondence image, w for a weighting value, w' for an incremental weighting value and (u 0 ,v 0 )* for the determined position of the first image area in the correspondence image These will not be discussed further below due to the introduction of reference symbols, but can be used consistently in principle.

[0046] The following description refers to Figure 1 and Figure 2a and 2b . Figure 1Figure 1 illustrates an application that demonstrates the object of the invention using a real-world example – in this case, the indirect measurement of distances (also: lengths). A schematic view shows a measurement setup in which a laser distance measuring device 10 is used with an imaging sensor device 12a, here implemented as a camera 12. In the illustrated embodiment, the laser distance measuring device 10 is located in a three-dimensional space. The laser distance measuring device 10 is oriented such that a distance measurement direction 14 is directed towards a target object 16, here represented as a house. The user's objective could be to determine the distance 18 in the illustrated measurement setup by indirect measurement. Using the handheld laser distance measuring device 10, a first distance 20 to a first (distant) point 22 (for example, the corner 26 of a window shown) is measured.Simultaneously with the measurement of the first distance 20, the image 24 – comprising the first image area 50 – of at least one area surrounding the first point 22 is captured by the laser distance measuring device 10 using the camera 12 integrated in the laser distance measuring device 10. The image 24 is in . Figure 2a schematically represented. For this purpose, the camera 12 is integrated on a side of the housing facing the distant object. Subsequently, a second distance 28 to a second (distant) point 30 is measured. Simultaneously with the measurement of the second distance 28 to the second point 30, an image of at least one area surrounding the second point 30 is captured – the correspondence image 52. The correspondence image 52 is in Figure 2b schematically represented.

[0047] The laser distance measuring device 10 indirectly determines the distance 18 between the first point 22 and the second point 30 from the two measured distances and the synchronously recorded images, and outputs this distance to the user using an output device 32. In particular, it is necessary, according to the inventive method 100, to determine the most precise possible position of the first image area 50 in the corresponding image 52 in order to subsequently determine a (possible) translation of the handheld laser distance measuring device 10 between the two distance measurements. A possible rotation of the laser distance measuring device 10 between the two image acquisitions, which is hardly avoidable in handheld operation, is assumed here to be known, since it can be determined using cost-effective inertial sensors (not shown in detail here).This rotation can be incorporated as prior knowledge, so that the degree of freedom of rotation can be disregarded when carrying out the method 100 according to the invention. Alternatively or additionally, this rotation can also be determined according to the method 100 according to the invention (see description above). In particular, a possible rotation in the image plane has already been applied to the first image area 50, so that gray or color values ​​are read out from the rotated rectangular first image area 50. Interpolation of the values ​​can be advantageous in this case. Since the possible rotation has already been compensated in the first image area 50, no further compensation of the rotation is required in the corresponding image 52. Alternatively, the rotation can also be compensated only in the corresponding image 52. Furthermore, as an alternative, the compensation of the rotation can be distributed proportionally between image area 50 and the corresponding image 52.

[0048] In the corresponding image 52, the second image area 56 is sought whose content (i.e., image content or image data content) corresponds to, or is similar to, the content of the first image area 50. The first image area 50 is found in the corresponding image 52 when the position 58 of the first image area 50 (here marked by the center of image area 56) in the corresponding image 52 is determined – for example, specifyable in pixel coordinates relative to the corresponding image 52 (see coordinate specifications in Figures 5a or 5b).

[0049] The result of this search is in Figure 2bThis has already been anticipated and illustrated in the corresponding image 52 by a dashed rectangle that characterizes the second image area 56. It should be noted that, unlike in this embodiment, the first image area 50 and / or the corresponding image 52 may also have a different shape, for example a rectangular, round, oval, elliptical, crescent-shaped, cross-shaped, polygonal, isogonal shape, or other shape known to those skilled in the art.

[0050] The laser distance measuring device 10 shown is Figure 1The system further comprises a control device according to the invention, not shown in detail here, for carrying out the method 100 according to the invention. The control device of the laser distance measuring device 10 includes a computer program comprising commands which, when executed by means of a processor device of the control device, cause the control device to execute the method 100 according to the invention. The computer program is stored on a computer-readable storage medium. The laser distance measuring device 10, comprising the imaging sensor device 12a, constitutes a system 200 according to the invention.

[0051] The process of the inventive method 100 for determining a position 58 of the first image area 50 in the correspondence image 52, wherein the first image area 50 corresponds to at least one second image area 56 in the correspondence image 52, is described in Figure 3 represented as a process diagram and is implemented using the Figure 2and 4a-4f explained.

[0052] In a first process step 102, the first image area 50, comprising at least one reference position 60, is provided (see dashed rectangle in Figure 2a In this embodiment, the provision is carried out by means of the camera 12 as the imaging sensor device 12a of the laser distance measuring device 10, as described above for the exemplary measurement configuration. In this embodiment, the provided first image area 50, cut out from the first image 24, is chosen to be somewhat larger so that sufficient reserve is available for the subsequent determination of signature values ​​64, since the determination involves not only a single image position 62 (here chosen as image point or pixel 54) but also its local surroundings. In process step 104, the corresponding image 52 is provided (see below). Figure 2bThe provision of the data is also carried out here by means of the camera 12 of the laser distance measuring device 10. It should be noted that the process steps 102 and 104 can be carried out at different times, whereby it may be advantageous to be able to temporarily store the first image area 50 and / or the corresponding image 52, whereby the provision is then carried out by reading from a memory. In the illustrations of the Figure 2 Individual image positions 62 and 80 are represented by pixel 54 (here by a grid). Each pixel position can thus be assigned a coordinate pair (xi, yi) that describes image position 62 and 80, respectively. As already mentioned, this shows Figure 2b already the result of the procedure, i.e. the second image area 56.

[0053] It should be noted that in the illustrated embodiment of the method according to the invention, 100 exemplary image data for the first image area 50 and the corresponding image 52 are shown, which are reduced in size to a few pixels 54 for better illustration (for example, in the case of the first image area 50 to 5×5 pixels, instead of, for example, comprising 41×51 pixels as in a typical application).

[0054] In process step 106, a plurality of signature values ​​64 (denotable by si) are determined for different image positions 62 (describable by pixel coordinates (xi, yi)) in the first image area 50. A signature value 64 characterizes a neighborhood of an image position 62 in the first image area 50. Each image position 62 has a relative offset 66 – describable as a distance (Δx i, Δy i) of the pixel coordinates – with respect to the reference position 60 (see also Figures 4a and 4bVarious methods are known to those skilled in the art for determining a signature value 64. For example, it can involve smoothing the image data, followed by several pairwise comparisons of pixel values ​​from image positions 62 (pixel positions) in a predefined relative arrangement, with binarization and concatenation of the individual results, whereby a binary sequence is determined. In this embodiment, the binary sequence has a length of 10 bits, corresponding to a value range from 0 to 1023. The majority of signature values ​​64 for different image positions 62 in the first image area 50 are determined in the form of a matrix, the so-called first signature value matrix 68. In the present embodiment, it is assumed that the first image area 50 is sufficiently large and that one signature value 64 is determined for each image position 62, here given by the individual pixels 54 of the image area 50.It should be noted that it is also possible in principle to determine more than one or less than one signature value 64 per image position 62 (i.e., 54 per pixel).

[0055] In Figure 4a A matrix of signature values ​​64 is shown, representing the first image area 50. This so-called first signature value matrix 68, also referred to as the first signature image, consists here, for example, of 5×5 matrix elements. This corresponds to 5×5 image positions 62 (pixels 54) in the first image area 50. There is a unique assignment between each matrix element of the signature value matrix 68 and an image position 62 in the first image area 50. The signature values ​​64 have a value range from 0 to 1023, as described.

[0056] In process step 108, relative offsets 66 of a plurality of image positions 62 – possibly also of all image positions 62 – in the first image area 50 are assigned to respective determined signature values ​​64. The relative offsets 66 of the plurality of image positions 62 in the first image area 50 are assigned to respective determined signature values ​​64 in an assignment rule 70, in particular in a lookup table. Such an assignment rule is described in Figure 4dAs shown, using a respective signature value 64, an address 72 of the assignment rule 70, in particular the lookup table, is defined. Specifically, the respective signature value 64 defines a table address where the respective entry is made. The content of the entry includes relative offsets 66 and weighting values ​​74 (denotable by w). As already described, it is conceivable that the address is not identical to a signature value, but can be indirectly determined from one or more signature values, e.g., by omitting components, combining components, performing bit operations, calculations, or the like.

[0057] To create this assignment rule, two further matrices are used in this embodiment, shown in Figure 4b and 4c In Figure 4bA matrix of relative offsets 66 is represented, which is the same size as the first signature value matrix 68. Each matrix element contains those relative offsets 66 that must be added at the respective position of a matrix element – ​​corresponding to an image position 62 – in order to reach the reference position 60. The reference position 60 is chosen as the center of the matrix.

[0058] For example, the entry '(2;1)' is in the first position from the left in the second row. This means that from this matrix element, one must move two fields to the right and one field down to reach the reference position 60. The assumed x-coordinate axis points to the right and the assumed y-coordinate axis points down, as is common in image processing. In this embodiment, the matrix of relative offsets 66 is intended to be fixed and unchanging and therefore does not need to be stored in variable memory.

[0059] In Figure 4cFurthermore, a matrix of weighting values ​​74 is shown, which is also as large as the first signature value matrix 68. In this embodiment, the weighting values ​​74 are integers and decrease substantially with increasing distance from the reference position 60 (i.e., from the center of the matrix or the first image area 50). Thus, a weighting value 74 is selected depending on the relative offset 66. With such a chosen matrix of weighting values ​​74, signature values ​​64 near the center of the signature value matrix 68 are given greater weight in the voting than signature values ​​64 at the edge of the signature value matrix 68. Small integers are used as weighting values ​​74, with the advantage that these can be described with shorter word lengths and result in lower storage and bandwidth requirements for transmission, which is relevant and advantageous for embedded systems.In this embodiment, the matrix of weighting values ​​74 is provided to be variable. In particular, zero is also permitted as a weighting value 74 in order to exclude certain signature values ​​64 from the vote, for example, if a signature value 64 is ambiguous. It should be noted that in a simplest embodiment, it is also possible for all weighting values ​​74 to have the value 1. In this case, the matrix of weighting values ​​74 can, in principle, be omitted.

[0060] Figure 4d Figure 70 shows an excerpt of the assignment rule 70, populated with example values, in particular the lookup table, which here contains the addresses 72 in the range from 0 to 1023. It depicts a population state that exists after the assignment rule 70 has been described using the signature value matrix 64. Figure 4a and based on the matrix of relative offsets 66 from Figure 4band optionally based on the matrix of weighting values ​​74 from Figure 4c has taken place. At address '1' of the allocation rule 70, a corresponding relative offset 66 with '(1;-1)' and a weighting value 74 with '3' are stored. The comparison with Figure 4a shows that the signature value '1' only occurs in the penultimate row in the second field of the signature value matrix 68. At the corresponding position of the matrix of relative offsets 66 ( Figure 4b ) the entry ,(-1;-1)' is located and at the corresponding position of the matrix of weighting values ​​74 is the entry ,3'.

[0061] The procedure for describing the allocation rule 70 is as follows: Element-wise and synchronous iteration (for example, in row or column direction) through all matrix elements of the 5×5 matrix elements - signature value matrix 68 ( Figure 4a ), Matrix of relative offsets 66 ( Figure 4b) and matrix of weighting values ​​74 ( Figure 4c ); thereby reading signature values ​​64 from the signature value matrix 68, relative offset 66 from the matrix of relative offsets 66 and weighting value 74 from the matrix of weighting values ​​74; jumping to address 72 of the assignment rule 70, which is defined using the signature value 64, and entering relative offset 66 and weighting value 74 - wherein in the Figure 4d In the illustrated embodiment, it is assumed for simplification that the signature value 64 directly represents the address 72 in the assignment rule 70.

[0062] Assignment rule 70 may contain addresses 72 where no entry has been made. This is in Figure 4drepresented by empty fields. In practice, however, something will actually be entered there, for example, arbitrary (because irrelevant) relative offsets 66 and / or weighting values ​​'0' – where '0' exemplifies the 'empty state'. The 'empty state' can mean that the corresponding signature value 64 did not occur in the signature value matrix 68, or that the corresponding signature value 64 occurred too often, so it was deleted to avoid ambiguity. In particular, it is conceivable to encode the latter case of deleted signature values ​​64 with a special value, e.g., '-1', to distinguish it from the 'empty state', in order to avoid ambiguity. The addresses 72 themselves do not require any storage space, as they are only implicitly present and fixed in ascending order.If the assignment rule 70 is to be used again, for example to search for another first image area 50 in another or the same correspondence image 52, it can be reset by setting all weighting values ​​74 and / or relative offsets 66 to zero. Alternatively, and in particular less time-consuming, it is possible to only go through and reset those signature values ​​64 that were used and thus changed in a previous execution of the procedure 100. These affected signature values ​​64 can be retrieved from the corresponding old (cached) first signature value matrix 68.

[0063] In this embodiment, repetitions of individual signature values ​​64 occur - here, for example, a clustering of the values ​​'333', '911' and '400' (marked by frames) in Figure 4a. In this embodiment, ambiguity handling is performed in process step 108 for such cases of ambiguity by combining or deleting ambiguities (not shown in detail here).

[0064] According to a first option, ambiguities of signature values ​​64 are deleted. This means that ambiguous signature values ​​64 are excluded from the subsequent procedure 100. This can be achieved during the writing process using the assignment rule 70 described below, for example, by writing a weighting value 74 of '-1', which marks the corresponding signature value 64 as excluded. The signature value 64 is then no longer available for subsequent correspondence searches. In practice, this could be implemented as follows: If, during an attempt to make an entry, it is determined that a weighting value w greater than '0' is already entered for an address, it is set to '-1'.If, during a further attempt to make an entry, we find that '-1' is already entered as the weighting value at this address, this entry will be cancelled and the weighting value '-1' will be left in place.

[0065] According to another option, ambiguities of signature values ​​64 are resolved by summarizing them. This can be done by weighting several relative offsets 66 in the assignment rule 70 by multiplying them with their respective associated weighting values ​​74: W i 1 ⋅ ΔX i 1 + W i 2 ⋅ ΔX i 2 + … / W i 1 + W i 2 + … ; W i 1 ⋅ ΔY i 1 + W i 2 ⋅ ΔY i 2 + … / W i 1 + W i 2 + … . For simplification, the division by (w i1 + w i2 + ...) can also be postponed to a later time (for example, after the last signature value 64 has been received).

[0066] In process step 110, a plurality of correspondence signature values ​​78 (denoted by sj) are determined for different image positions 80 (denotable by (uj, vj)) in the correspondence image 52, where a correspondence signature value 78 characterizes a neighborhood of an image position 80 in the correspondence image 52. The correspondence signature values ​​78 for image positions 80 in the correspondence image 52 are generated analogously, in particular according to the same procedure, as the signature values ​​64 for image positions 62 in the first image area 50. The result is a second signature value matrix 82 (also denoted as a second signature image), as shown in Figure 4eThis is shown. It should be noted that the second signature value matrix 82 has an "empty border" because determining correspondence signature values ​​78 requires a local neighborhood, which is not always the case near the edge of the image. Therefore, a valid correspondence signature value 78 cannot be determined everywhere there. For example, the "empty border" can be implemented either by omitting rows or columns of the second signature value matrix 82 or by marking correspondence signature values ​​78 as "invalid," e.g., by a special signature value or by a weighting value of '0'. It can also be advantageous to perform ambiguity handling for the second signature value matrix 82—regardless of whether such handling is performed for the first signature value matrix 68 or not. It should be noted that the second signature value matrix 82 does not necessarily have to exist in its entirety.In particular, it is not necessary to store the second signature value matrix 82. Instead, it may suffice to store only a fraction of it, e.g., only as many matrix rows of correspondence signature values ​​78 as are required for ambiguity handling—or even just a single correspondence signature value 78. This is because the correspondence signature value 78 is only needed briefly and can then be forgotten. This allows for the implementation of method 100 with very low memory requirements, making it particularly interesting for cost-critical embedded systems.

[0067] The following procedural steps 112-116 describe the execution of a vote with the aim of determining a voting matrix 84 (also referred to as a voting image or voting weight matrix), in particular of the size of the second signature value matrix 82 (possibly minus the empty margin, i.e., the voting matrix 84 contains high values ​​at those matrix elements where there is a high similarity to the desired position 58 and low values ​​at the remaining (dissimilar) positions. Since weighting values ​​must be entered at image positions (Δuj + uj, Δvj + vj) during the vote, it can also be advantageous to choose the voting matrix 84 to be precisely the size that allows for the entry of weighting values ​​74 at all times – i.e., that no weighting values ​​74 end up outside the voting matrix 84.

[0068] In process step 112, relative offsets 66 (now denotable as (Δu j ,Δv j )) to determined correspondence signature values ​​78 are determined using the assignment rule 70. Figure 4eThe diagram illustrates how the voting process works. The image shown is the second signature value matrix 82, which represents the content of the correspondence image 52. This second signature value matrix 82 is processed element by element, for example, row by row, starting at the top left. If a valid correspondence signature value 78 is present at the respective image position 80 (matrix position) – here, '709' – then the address 72 of the assignment rule 70 corresponding to the correspondence signature value 78, specifically the lookup table, is checked to see if a valid entry exists, i.e., in this example, whether the weighting value 74 stored there is greater than zero. If this is the case, the relative offsets 66 and the weighting value 74 stored there are read – here, the relative offset '(2;1)' and the weighting value '2'.Accordingly, the allocation rule 70 is accessed again during the voting process, but this time only in read-only mode. For each valid correspondence signature value 78, in particular from the second signature value matrix 82, the address 72 of the allocation rule 70, defined by the correspondence signature value 78, is accessed, and the stored relative offsets 66 and the weighting value 74 are read out there, if available.

[0069] In process step 114, target positions (denotable as (u 0 ,v 0 ) = (uj ,vj ) + (Δu j ,Δv j ) = (Δu j +uj ,Δv j +vj ) in the correspondence image 52) are calculated by adding the relative offset 66 (Δu j ,Δv j ) and the respective image position 80 (uj ,vj ) in the correspondence image 52. Optionally, scaling, preferably rotation, stretching / compression, and shearing, can also be compensated at this point. Weighting values ​​74 are assigned to target positions in the correspondence image 52 in the form of the reconciliation matrix 84 - see [reference]. Figure 4fThe voting matrix 84, in this embodiment, has the same size as the second signature value matrix 82 and is initialized before creation begins (for example, set to zero or 'empty' everywhere). At the determined target position 88 in the voting matrix 84, an incremental weighting value 86 (denotable by w') is then increased in process step 116 by incrementally assigning the weighting value 74, provided it is valid. Further details are provided below. Figure 4e Two lines and one column later, the correspondence signature value 78 of '1' is reached. At the corresponding address 72 of the allocation rule 70 (see... Figure 4eThe relative offset 66 is stored as (1;-1) and the weighting value ,3'. Thus, in this embodiment, the same target position 88 in the reconciliation matrix 84 results as previously with the correspondence signature value 78 of '709'. The previously stored incremental weighting value 86 of ,2' is now increased by ,3' to ,5' (value ,5' is in Figure 4f (as shown). This demonstrates how the expected relative arrangement of the correspondence signature values ​​78 (compare the relative arrangement of the signature values ​​'709' and '1' with respect to the target position 88 in Figure 4e with the relative arrangement of the signature values ​​of the same name 64 in relation to the reference position 60 in Figure 4aThis leads to an increase in the incremental weighting value 86, which ultimately results in the desired position 58. It should be noted that no further action is required on matrix elements of the second signature value matrix 82 without a valid correspondence signature value 78 (represented as empty fields). It is also conceivable to assign a special signature value such as '-1', '0', or '1024' to such invalid fields.

[0070] Due to quantization to a pixel grid, rounding inaccuracies, interpolation errors, changes in distance, and perspective shifts between the first image area 50 and the corresponding image 52, the incremental weighting values ​​86, which ideally should be concentrated in one image position 80 (i.e., in one matrix element), may be scattered or 'smeared' across several neighboring image positions 80. This scattering can be effectively compensated for by a post-processing step, for example, by applying a linear and spatially invariant smoothing filter to the tuning matrix 84. Such filtering leads to a clearer identification of the local extrema 90 to be searched for, especially maxima, and can facilitate the search.to reduce uncertainty in the search, because when considering the relationships, weak secondary extrema 92, especially secondary maxima, tend to be suppressed and strong extrema, especially maxima, tend to be emphasized.

[0071] Finally, in process step 118, position 58 (denotable as (u 0 ,v 0 )*) of the first image area 50 in the corresponding image 52 is determined by finding an extremum 90, in particular a maximum, in the incremental weighting values ​​86 of the voting matrix 84. It should be noted that the voting matrix 84 does not need to be explicitly or completely stored at any time. It may be sufficient to pass on matrix elements or voting values ​​for the "winner" (global extremum or maximum) of the vote and possibly for a second, third "winner" (local extrema or maxima), etc., to a further evaluation, in particular to a further process step 120 of the evaluation.Further analysis can, for example, check whether the incremental weighting value 86 or the incremental weighting values ​​86 are sufficiently unique for a winner and whether there is a sufficiently significant distance between an incremental weighting value 86 and another image position 80 in the corresponding image 52 (i.e., the second "winner"). Accordingly, position 58 of the first image area 50 in the corresponding image 52 is determined if the incremental weighting value 86 assigned to position 58 reaches a relative or absolute threshold.

[0072] It should be noted that the voting matrix 84 does not need to be stored in its entirety, but it can be useful to retain in memory the portion on which weighting values ​​74 can still be assigned. Since the vectors (Δu j ,Δv j ) of the offset 66 are typically not very long, the "range" of the voting process is very limited, so that weighting values ​​74 are only assigned on a relatively narrow strip of the voting matrix 84. Therefore, it is conceivable to evaluate the portion of the voting matrix 84 that is no longer within the range of the voting process—i.e., that is no longer within the range of the offset 66—beforehand, so that it no longer needs to be stored.

[0073] In Figure 5 are for the exemplary measurement constellation of the Figure 1Two voting matrices, 84a and 84b, are shown. The incremental weighting values ​​86 of each voting matrix 84a and 84b are represented as grayscale values ​​in a graphical representation, where a darker grayscale value corresponds to a higher value of the underlying incremental weighting value 86. Voting matrix 84a in Figure 5a concerns the determination of position 58a of the first image area 50 in the first image 24, while the voting matrix 84b in Figure 5bThis concerns the determination of position 58b of the first image area 50 in the corresponding image 52. It is clearly evident that a distinct extremum 90 exists in the reconciliation matrix 84a. A clearly defined extremum 90 is also apparent in the reconciliation matrix 84b, which marks position 58b of the first image area 50 in the corresponding image 52, with further minor extrema 92 being less visible. Finally, by examining the respective values ​​using a threshold value, it becomes possible to identify position 58b as the global extremum.

[0074] It should be noted that voting matrix 84a is not strictly necessary in practice. However, it is conceivable to use voting matrix 84a optionally to determine in advance (in the same image, i.e., in the first image 24) whether there are any unfavorable ambiguities and how close these are to a correct candidate. This prior knowledge could be used in the later evaluation of the correspondence image 52, for example, as information on how pronounced and how close a second "winner" is and / or whether a second "winner" is still included in the correspondence image or not.

[0075] In Figure 6A SoC 300 for carrying out the method 100 according to the invention is shown. The SoC 300 acquires a first image area 50 and a corresponding image 52 using an imaging sensor device 312a of the SoC 300. The SoC 300 is configured to determine a position 58 of the first image area 50 in the corresponding image 52 and, depending on this, to generate or output a signal 302.

Claims

1. Method for determining a position (58) of a first image region (50) of an initial image (24) in a correspondence image (52), comprising at least the following method steps: • providing the first image region (50) from the initial image (24) comprising at least one reference position (60), wherein the reference position (60) is arbitrarily selected, • providing the correspondence image (52), • determining a plurality of signature values si (64) for different image positions (xi,yi) (62) in the first image region (50), wherein a signature value sj (64) characterizes surroundings of an image position (xi,yi) (62) in the first image region (50) and wherein an image position (xi,yi) (62) has a relative offset (Δxi,Δyi) = (xr,yr) - (xi,yi) (66) with respect to the reference position (xr,yr) (60), • assigning relative offsets (Δxi,Δyi) (66) of a plurality of image positions (xi,yi) (62) in the first image region (50) to respectively determined signature values si (64) in an assignment specification (70), in particular in a lookup table, • determining a plurality of correspondence signature values sj (78) for different image positions (uj,vj) (80) in the correspondence image (52), wherein a correspondence signature value sj (78) characterizes surroundings of an image position (uj,vj) (80) in the correspondence image (52), • determining relative offsets (Δuj,Δvj) (66) for determined correspondence signature values sj (78) using the assignment and • calculating target positions (u0,v0) = (uj,vj) + (Δuj,Δvj) in the correspondence image (52) by summation of the relative offset (Δuj,Δvj) (66) and the respective image position (uj,vj) (80), • incrementally assigning a weighting value w (74) to a calculated target position by reading out from the assignment specification (70), wherein an incremental weighting value w' (86) is formed, • determining the position (u0,v0)* (58) in the correspondence image (52) which corresponds to the reference position (ur,vr) (58) in the first image region (50), by finding an extremum (90) in the incremental weighting values w' (86).

2. Method according to Claim 1, wherein the plurality of signature values si (64) for different image positions (xj,yj) (62) in the first image region (50) are determined in the form of a matrix or table and / or wherein the plurality of correspondence signature values sj (78) for different image positions (uj,vj) (80) of the correspondence image (52) are determined in the form of a matrix or table.

3. Method according to either of the preceding claims, wherein a respective signature value si (64) defines an address (72) of the assignment specification (70), in particular of the lookup table.

4. Method according to any of the preceding claims, wherein writing the assignment specification (70) comprises the following steps: • element-wise and synchronous iteration through all the matrix elements of a matrix with size 5 x 5 matrix elements; • signature value matrix (68), matrix of relative offsets (66) and matrix of weighting values (74); • reading out signature values (64) from the signature value matrix (68), relative offset (66) from the matrix of relative offsets (66) and weighting value (74) from the matrix of weighting values (74); • jumping to an address (72) of the assignment specification (70) which is defined using the signature value (64), and entering relative offset (66) and weighting value (74).

5. Method according to any of the preceding claims, wherein the weighting values w (74) are assigned to target positions in the correspondence image (52) in the form of a matrix or table.

6. Method according to any of the preceding claims, wherein a correspondence signature value sj (78) for an image position (uj,vj) (80) in the correspondence image (52) is formed analogously to, in particular according to the same specification as, a signature value si (64) for an image position (xi,yi) (62) in the first image region (50).

7. Method according to any of the preceding claims, wherein the assigned weighting value w (74) is selected depending on the relative offset (Δuj,Δvj) (66), in particular increases or decreases as the relative offset (66) increases.

8. Method according to any of the preceding claims, wherein a position (u0,v0)* (58) of the first image region (50) in the correspondence image (52) is determined if the incremental weighting value w' (86) assigned to the position (58) reaches a relative or absolute threshold value.

9. Method according to any of the preceding claims, wherein a relative offset (Δuj, Δvj) (66) for determined correspondence signature values sj (78) is calculated using the assignment and a scaling variable.

10. Method according to any of the preceding claims, wherein ambiguities in the plurality of determined signature values si (64) are eliminated by deleting and / or by merging the ambiguous signature values si (64), wherein the merging is carried out by weighted averaging of relative offsets (66).

11. Method according to any of the preceding claims, wherein the first image region (50) from the initial image (24) and / or the correspondence image (52) are / is captured and provided using at least one imaging sensor device (12a).

12. SoC (System-on-Chip) configured to carry out a method according to any of Claims 1 to 11, wherein the SoC (300) is configured to capture at least one first image region (50) and a correspondence image (52) and to generate an output signal (302) depending on the determined position (u0,v0)* (58) of the first image region (50) in the correspondence image (52).

13. Control device configured to carry out a method according to any of Claims 1 to 11.

14. System comprising at least one control device according to Claim 13 and at least one imaging sensor device (12a) configured to capture and provide a first image region (50) and a correspondence image (52).

15. Computer program comprising instructions which, when executed by means of a processor device of a control device, cause the control device to carry out a method according to any of the preceding Claims 1 to 11.

16. Computer-readable storage medium on which a computer program according to Claim 15 is stored.