• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    An image processing method and apparatus for locating at least one element (2, 102, 202) in an image, the element having a known shape, the method comprising the steps of: - forming a first distribution (C1) comprising a first identifiable signature (21, 22) related to first features (11, 12) of the known form and a second distribution (C2) comprising a second identifiable signature (23, 24) related to second features (13, 14) of the form, - identifying the first signature in the first distribution and the second signature in the second distribution and detecting the respective positions of the first signature in the first distribution and the second signature in the second distribution, - deriving from the position of the first signature in the first distribution and the position of the second signature in the second distribution, of a position of the element (2, 102, 202) in the image.
    公开号:BE1024091B1
    申请号:E2015/5451
    申请日:2015-07-15
    公开日:2017-11-13
    发明作者:Gil José De；David Rebetez；Mathieu Gaillard
    申请人:Mycartis Nv；
    IPC主号:
    专利说明:

    Method for precise location of elements in an image
    DESCRIPTION
    TECHNICAL FIELD AND PRIOR ART
    The present invention relates generally to the field of image processing and more particularly to methods arranged for locating items in an image.
    In computer imaging, it is sometimes necessary to detect the precise position of a specific item with a known shape.
    The template matching method can be used, for example, to detect shapes in a digital image. Such a method is described in "Template Matching Techniques in Computer Vision: Theory and Practice", by R. Brunelli, ISBN 978-0-470-51706-2, 2009 and uses a convolution mask that is dependent on a particular known property of an image that one wants to detect.
    Another method is the Hough transformation method, which is particularly suitable for locating shapes with elliptical or circle lines.
    The computing time is a major problem of the known image processing methods arranged to locate items, particularly when high resolution images are being processed or when, for some applications, it is necessary to repeat the localization operation multiple times for several different images.
    In the case, for example, of the template matching method, the image resolution must often be lowered by pre-filtering the image in the frequency space, which requires at least the implementation of a heavy FFT (FFT as "Fast Fourier Transform") operation .
    In the case of the Hough transformation method, an edge detection step is generally required. This step further increases the sensitivity to image sharpness and / or noise.
    The problem that arises is to find a new image processing method for locating objects of known shape in an image, and which is improved in terms of computation time and noise sensitivity or image quality.
    DESCRIPTION OF THE INVENTION
    The present invention provides an image processing method for determining the position of at least one element in an image, the element having a known shape.
    The method comprises the following steps: a) forming a first distribution consisting of a first identifiable signature related to first characteristics of the known form and a second distribution consisting of a second identifiable signature related to second characteristics of said form, the first distribution is formed by an operation comprising: forming a first series of selected pixel lines of the image and associating with each selected pixel line of the first series a first index value and a first distribution value, the first distribution value being dependent on respective values of the pixels from which the selected line is formed, the first distribution being formed by the first index values and the first distribution values, the second distribution being formed by an operation comprising: forming a second series of selected pixel lines and having each selected line of the second series associating a second index value and a second distribution value depending on respective values of the pixels from which the selected line is formed, the second distribution being formed by the second index values and the first distribution values, the first series and the second series are selected to form a grid so as to form a coordinate system of the image that makes it possible to identify positions in the image and so that the set of first distribution and second distribution enable the processing of a step consisting of: b) the identifying the first signature in the first distribution and the second signature in the second distribution to derive position data according to the coordinate system by identifying the pixel lines of the first series corresponding to the first signature in the first distribution and the pixel lines of the second series corresponding to the second signature in d the second distribution, c) deriving the position of the element in the image from the position data according to the coordinate system.
    This method allows faster detection because 1D sequences (the first and second distributions) are processed instead of a 2D image.
    The process for forming the first and second distribution, called collapse processing, has reduced the sensitivity to noise since the collapse tends to average high-frequency variations, and the sensitivity to image sharpness is reduced since edge detection is not necessary.
    A signature is defined in the present description as one or more distinctive set (s) of values of a distribution (or function) or one or more distinctive part (s) of the graphical representation of a distribution (or function) that is related ( are) to, or is (are) representative of, characteristic feature (s) of the shape of an element.
    The collapse processing at step a) may be such that the value associated with a given line of pixels is a function of the sum of respective values of pixels included in said given line.
    According to a possible embodiment, collapse processing can be performed such that the value associated with a given line of pixels is a function of an average of respective values of pixels included in said given line.
    According to a possible embodiment, collapse processing can be performed such that the value associated with a given line of pixels is a function of a maximum of respective values of pixels included in said given line.
    According to a possible embodiment, the value may further be a function of a weighting factor.
    According to a possible implementation, the identification of the first signature in the first distribution and of the second signature in the second distribution may comprise the steps of: - calculating a convolution between the first distribution and a first predetermined template that performs a first function is comprising a model of the first signature, - calculating a convolution between the second distribution and a second predetermined template that is a second function comprising a model of the second signature.
    According to a possible embodiment, the lines of the first series and of the second series may correspond to horizontal lines and vertical lines of the image, respectively.
    Advantageously, when the orientation of the element in the image is known or when at least an estimate of the orientation is known for step a), the respective orientation of the lines of the first series of lines and of the second series is selected during step a) depending on the known orientation.
    According to a possible embodiment, the element prior to step a) further has a known orientation relative to the horizontal direction and vertical direction of the image, the respective orientation of the lines of the first series of lines and of the second series selected during step a ) depend on the known orientation.
    According to a possible implementation, the lines of the first series and / or the second series may be non-linear lines.
    The present invention further provides a computer program product comprising program code instructions stored on a computer-usable medium by means of computer processing means comprising readable programming means for performing an image processing method according to the present invention.
    Furthermore, a digital data storage device readable by computer processing means, comprising code instructions, is provided with a computer program as previously defined.
    Furthermore, an imaging system comprising an image acquisition apparatus and an image processing apparatus for performing an image processing method according to the present invention is provided.
    In another aspect, the present invention further provides an image processing apparatus for determining the position of at least one element in the image with a known shape, the apparatus comprising processing means, the processing means comprising: - means for forming a first distribution comprising a first identifiable signature related to first features of the known shape and a second distribution comprising a second identifiable signature related to second features of the shape, the first distribution being formed by an operation comprising: forming a first set of selected pixel lines of the image and associating a first index value and a first distribution value with each selected pixel line of the first series, the first distribution value being dependent on respective values of the pixels forming the selected line, the first distribution being formed by the first index value and the first distribution values, the second distribution being formed by an operation comprising: forming a second series of selected pixel lines and associating with each selected line of the second series a second index value and a second distribution value depending on respective values of the pixels from which the selected line is formed, the second distribution being formed by the second index values and the first distribution values, the first series and the second series being selected to form a grid forming a coordinate system of the image forming the allows to identify positions in the image, - means for identifying the first signature in the first distribution and the second signature in the second distribution so as to derive position data according to the coordinate system by identifying the pixel lines of the first series corresponding to the first signat hour in the first distribution and the pixel lines of the second series corresponding to the second signature in the second distribution, means for deriving the position of the element in the image from the position data according to the coordinate system.
    The present invention further provides a molecular diagnostic platform designed to receive a support with a reference mark, wherein the platform is provided with detection means for detecting the position of the support using an image of the support, the detection means: are arranged to implement an image processing method as previously defined and / or
    - an image processing system according to the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
    The present invention can be better understood upon reading the following description of exemplary embodiments provided purely for indicative and non-limiting purposes, with reference to the accompanying drawings, in which: Figure 1 shows a cartridge that can be imaged via a camera and that includes a reference mark that can be used as a positioning element whose exact position in the image can be detected by using an image processing method according to the invention; Figure 2 represents a 2D image obtained from the cartridge of Figure 1A, comprising a reference mark whose position can be detected by using an image processing method according to the invention; Figure 3 shows two graphical representations of distributions each comprising identifiable signatures representative of reference mark features, such distributions being formed by an operation called collapse processing implemented during the image processing method of the present invention; Figure 4 shows an example of a flow chart illustrating steps of an image processing method according to the invention; Figures 5A, 5B represent a first line of image pixels selected along a first collapse direction and a second line of image pixels selected along a second collapse direction during a collapse processing; Figure 6 shows a template for comparison with an identifiable signature of a distribution obtained during an image processing method according to the invention; Figure 7 shows the collapse processing step of an image processing method according to the invention and illustrates the formation of two 1D distributions from a 2D image by adding the gray value of the pixels, the distributions comprising a signature portion representative of the shape of an item, the signature being used to detect the exact position of this point on the 2D image; Figure 8 represents a dashed line of image pixels to illustrate a collapse processing method in which some pixels of a selected line are not processed; Figure 9 shows different areas of pixels that are processed according to a collapse processing method with different densities of processed pixels; Figure 10 shows a non-linear line of image pixels taken along a non-linear collapse direction and being processed according to a collapse processing method; Figure 11 represents a variant of the processing of Figure 2, wherein collapse processing is performed according to collapse direction that are different from horizontal and vertical directions of the image; Figure 12 represents a system for image acquisition and image processing that is arranged to implement an image processing method according to the invention.
    Identical, comparable or equivalent parts in the different figures have the same numerical references because of the coherence between the figures.
    The various components as shown in the figures are not necessarily shown in a uniform scale to make the figures more readable.
    In addition, in the following description, terms that depend on the orientation of the image such as "vertical", "horizontal" apply to an image oriented in the manner illustrated in the figures.
    DETAILED DESCRIPTION OF SPECIFIC
    EMBODIMENTS
    Figure 1 shows a cartridge that is provided with a micro-fluid system that is intended to be filled with a clinical specimen and that can be used in a molecular diagnostic platform (not shown), for example as designed by the applicant.
    The molecular diagnostics platform includes a collection of instruments (not illustrated), each of which may be designed to receive such a cartridge. The cartridge 1 comprises at least one reference mark 2 with a known shape and geometry, for example in the form of a cross, and which is used for alignment. The cartridge is thus inserted into an instrument and positioned by mechanical means using the reference mark 2. The positional accuracy of the cartridge is improved by an image processing method and image processing apparatus implemented according to the invention.
    Figure 2 shows an example of a 2D image with defects 5 such as dust and / or optical artifacts that can make detecting the reference mark 2 more difficult. The image processing method and apparatus implemented according to the invention make it possible to determine in the image I the precise position of the reference mark 2, the shape of which is known in advance (i.e. for applying the image processing method) but whose precise position must be determined. For example, in the case that the obtained image I is a digital image in shades of gray, the value of each pixel carries intensity information ranging from black (e.g. with value 0) at the weakest intensity to white at the strongest intensity (e.g. with value 255).
    To determine the precise position of the reference mark 2, a two-dimensional (2D) image I of an area known to contain the reference mark 2 is first obtained.
    An image processing method is then performed to determine the precise position of the reference mark 2 in the image I.
    An example of such an image processing method is described below with reference to Figures 3, 4 and 5.
    After obtaining the image I (step S0 in Figure 4), then an operation referred to as "projection" or "collapse" is performed in the image I to create a first distribution and a second distribution.
    The first and second distributions are associated with a first and second direction called "collapse direction" of the distribution and are formed based on the pixel values of an area of the image I including the shape of the reference mark.
    Each distribution is a function that belongs to a series of indexes also called "bins", a series of values also called "bin values".
    Each index identifies a given line in the image I along the collapse direction of the distribution.
    Each bin value associated with an index depends on pixel values of the given line.
    The two collapse directions can be parallel to the vertical Y direction and the horizontal X direction of the image I, respectively.
    According to a possible embodiment, the series of lines taken along the collapse direction can be vertical lines or horizontal lines of the image I, in which case, each index in the distribution can be an X coordinate of a corresponding vertical line, respectively a Y coordinate of a corresponding horizontal line.
    The first and second distribution are 1D sequences that include information regarding the shape of the mark 2 and that make it possible to detect its position in the image.
    The first distribution is formed by choosing a first collapse towards D1 (step S10 in Figure 4).
    This first collapse direction can be selected as a function of a known estimate of the orientation of the reference mark 2 or by knowing in advance the orientation of the reference mark 2.
    The first collapse direction can be further selected depending on the arrangement of observable features or patterns of the shape of the reference mark 2, so that when a first series of pixel lines is directed in this collapse direction, the first series has a recognizable feature of the shape of the reference mark. By "recognizable" is meant here a characteristic of a shape that can be distinguished from the background of the image and / or the surrounding shapes by image processing.
    In the case that the reference mark 2 has, for example, a cross shape, the first collapse direction D1 can be a direction parallel to a strip la of the crossing strips la, lb which form the cross shape. In the example shown in Figure 3, the first collapse direction D1 is further parallel to the horizontal direction (X direction in Figure 3) of image I.
    The collapse processing (step Sn in Figure 4) is performed on a first series Ai of pixel lines Ri oriented according to the first collapse direction D1.
    In the example illustrated in Figs. 3 and 5A, a horizontal pixel line Ri (where the first collapse direction, in this example, is parallel to the horizontal direction) is associated with a value Vi that is calculated by the respective values of the pixels pi that sum the horizontal line Ri forms. The result of this calculation is stored in association with the index value i of the line Ri. To form the first distribution, each line of the first series or at least some lines of the first series is processed.
    The first distribution therefore belongs to an index i (also called "bin") for each processed line Ri, a value Vi (also called "bin value") that is dependent on the sum of pixel values taken along the first collapse direction D1.
    In a case where the collapse direction D1 is parallel to the X direction of the image, the indexes i can easily correspond to the Y coordinate of each line of pixels Ri taken into consideration.
    As previously indicated, the first distribution comprises an identification part or distinctive identifiers representative of distinguishable features of the shape of the reference mark. In the graphical representation C1 of the first distribution shown in Figure 3, these identifiers are represented as peaks 21, 22.
    The peaks 21, 22 are representative of edge lines 11, 12 of the first strip 1a of the cross shape and thus form a first identifiable signature of the cross shape. By identifying this first signature and locating this first signature in the first distribution, ie determining the indexes i of the peaks, it is possible to know which line R 1 is the closest basis for first strip 1a and therefore, in this case where the index i is favorably chosen to correspond to the Y coordinate of each R1 line, to derive a position of the Y coordinate of the first strip la. Before, or simultaneously with, or after forming the first distribution, a second distribution is formed by choosing a second collapse direction D2 (step S20 in Figure 6).
    The first collapse direction D1 and the second collapse direction D2 are such that the pixel lines selected according to these directions D1, D2 form a grid 30 through which a coordinate system is formed. Thus, the pixel lines R1 taken along the collapse direction D1 and the pixel lines Rj taken along the collapse direction D2 intersect at one point such that each given pair of indexes i and j identifies a unique point M of the image.
    The second collapse direction can be a direction parallel to the second strip 1b that forms the cross of the reference mark and parallel to the vertical direction (indicated as Y-direction in Figure 3) of the image I.
    Before, or simultaneously with, or after the collapse processing on the first series A 1, a collapse processing to form the second distribution is performed on a second series A2 of pixel lines aligned according to the second collapse direction D2 (Step S21).
    In the example of collapse processing illustrated in Figures 3 and 5B, each vertical pixel line Rj is associated with a value Vj that is calculated by summing the respective values of the pixels that form the vertical line Rj.
    The result of this calculation is stored in conjunction with the index value j of the line Rj. To form the second distribution, each line of the second series A2 or at least some lines of the second series is processed.
    The second distribution associates a coordinate taken along the X direction with a bin value that is the accumulation of pixel values taken along the second collapse direction.
    In the graphical representation C2 of the second distribution shown in Figure 3, peaks 23, 24 are representative of a second identifiable signature of the second strip 1b are border lines 13, 14.
    After step Sn and before or simultaneously with, or after step S2i, an identification step (step Si2 in Figure 4) of the first signature in the first distribution can be performed.
    The first signature is one or more distinctive sets of values of the first distribution or one or more distinctive part (s) of the graphic representation of the first distribution that is (are) related to, or is (are) representative of, characteristic (s) and) of the shape of the element when collapsed in the first direction.
    The search and location of the first signature in the first distribution can be done via a signal processing method comprising: calculating a convolution with a first mask representative of a model of the first signature and finding the maximum of this calculation. This can be complemented with prior processing of the distribution to correct known effects in the image, such as normalizing the value level of the pixels or suppressing skew in the distribution to compensate for light effects.
    Figure 6 shows an example of a template Ti used to identify the first signature. In this specific template example, the distance Δ between two peaks 123, 124 is an important feature, while the peak amplitude does not have to be very precise.
    Before, or simultaneously with or after the location of the first signature, the second signature of the shape is identified and located in the second distribution (Step S22 in Figure 4).
    The search for the second signature can also be performed by calculating the convolution with a second mask representative of a template of the second signature and finding the maximum of this calculation. This can also be complemented with prior processing of the distribution to correct known effects in the image, such as normalizing the value level of the pixels or suppressing skew in the distribution to compensate for light effects.
    Thus, in the example of Figure 3, the position of the first signature in the first distribution and the position of the second signature in the second distribution are detected by determining indexes i, j of the distinguished features 11, 12, 13, 14 of the mark 2 so as to give the coordinates to these characteristics.
    These indexes make it possible to derive the exact position of the reference mark 2 in the image I (step S30).
    The above embodiment is described in the context of a specific application consisting of detecting the correct position of a mark on a specific type of support.
    However, an image processing system according to the invention can be applied for detections on other types of supports of other types of elements with different geometries or different shapes.
    Fig. 7 shows another example of item 102 whose position on an image can be detected by performing an image processing method according to the invention.
    The item 102 here comprises an L-shaped zone 103 with a certain gray level and a black zone 104 in the L-shaped zone 103. The shape of the item 102 and the orientation are known beforehand, i.e. before image processing.
    Both collapse directions D1 and D2 are selected such that a first signature and a second signature of the shape to be detected are recognizable when the L-shape is processed by collapse processing, respectively according to the first collapse direction and the second collapse direction.
    In this example, the first collapse direction and the second collapse direction correspond to the direction of a strip 102a and the direction of a strip 102b forming the L-shape. The first collapse direction D1 and the second collapse direction D2 are furthermore parallel to the horizontal X direction and parallel to the vertical Y direction of the image, respectively.
    Next, a first series of lines N (with N as an integer) oriented in the first collapse direction is processed to form a first distribution whose graphical representation C10 is given in Figure 7. The first distribution contains the pixels of the detectable L-shape or at least an identifiable property of the shape.
    In a particular embodiment of this collapse method, only certain pixels pi of selected pixel lines Ri are processed (processed pixels are checked in Figure 8) to save computing time.
    For saving computing time, the collapse process can also only take into account certain lines of a series of pixel lines. Indeed, a process with lower resolution may in some cases be sufficient to implement a distribution with an identifiable signature.
    An uneven density of lines can even be considered, with more lines being processed by collapse processing in certain zones 31 of an image, for example in the center of the image than other zones 32 of the image (Figure 9).
    This density can be adjusted depending on the prior knowledge of the characteristics of the shape to be detected.
    To form a second distribution whose graphical representation C2o is further given, a second series of M lines (with M as an integer that can be equal to N) is also processed. The M lines are oriented according to a second collapse towards D2.
    In the specific example of Figure 7, the collapse processing is performed by adding the value of the pixels of a line of pixels.
    Alternatively, the collapse processing can be performed by associating a value with each edited row of pixels depending on: a weighted sum of the pixel values of the edited row, or as a function of a maximum pixel value in this edited row, or an average pixel value in this edited row.
    Next, a first signature of the L shape and the position of the first signature in the distribution is detected, while in the second distribution a second signature of the L shape and its position are detected.
    Such a detection is carried out, for example, by calculating a convolution of the first distribution with a first predetermined template and a convolution of the second distribution with a second predetermined template.
    From the determined position of the first signature in the first distribution and the second signature in the second distribution, the position of the L-shape in the 2D image is subsequently detected.
    According to a variant of the above-described embodiment of the image processing method, collapse directions other than the horizontal X direction and the vertical Y direction of the image can be selected, particularly when the shape to be located has identifiable features that are not oriented according to X and Y directions.
    An example of such a variant is shown in Figure 10, where item 102 with L-shape has an orientation such that the respective directions of the 2 strips 102a, 102b forming the L-shape make a 45 ° angle with the horizontal X direction and vertical Y direction of the image. In this variant, collapse directions D'1 and D'2 are chosen depending on the 2 lines that form the L-shape and therefore make an angle of 45 ° with the horizontal X direction and vertical Y direction of the image.
    In this case, the collapse process to form the first and second distribution can use an algorithm, such as Bresenham, to select pixels along a line for oblique collapse directions.
    The image processing method is not limited to two collapse directions to locate an item in an image, and can be executed with more collapse directions, depending on the number of features that must be identified to derive the precise position of this item.
    In some cases, especially when the shape of the item is more complicated to detect and the signature of the shape is more difficult to distinguish from the background, k collapse directions (with k> 2) can be used to form k distributions .
    The image processing method is not limited to linear collapse directions. After all, as long as a first and second series of pixel lines forming a coordinate system of the image as defined above are used to perform a first and second collapse, one can derive a position in the image I from the knowledge of the signatures in the image. first and second distributions as a result of the collapse processes.
    Figure 11 shows a processed line Ri of pixels along a non-linear collapse "direction" D1. For example, radial collapses can be implemented for specific images with specific circular features.
    The method described above is not limited to the processing of grayscale and can be further applied to a color image. In this case, the pixel "values" and bin "values" as explained above can be a vector of values for red, green, and blue components (or any other color scheme such as YUV). The collapse operation, such as the sum of pixel values, can therefore be performed on vectors such that processing of the vector values of a line of pixels Ri results in a vector value Vi for bin i in the distribution.
    Alternatively, the above-mentioned distributions can be formed for each color, the values for red, green, blue components (or other color schemes such as YUV) being processed independently of each other to produce a first and second distribution for the red component, a first and a second form a second distribution for the green component, and a first and a second distribution for the blue component.
    Alternatively, only a few color components can be edited to form only first and second distributions for these selected colors.
    In another embodiment, the first and second distributions are implemented by combining all color components of the pixels. In that case, a different weight can be assigned to the color components. For example, the blue color component of pixels can be associated with a higher weight when it is known that the item to be located is blue.
    A filtering of the colored image can be performed prior to collapse processing and forming the first and second distribution.
    Fig. 12 is an example of an imaging system according to the invention, comprising an image acquisition device 410 connected to a processing device 420. The acquisition device 410 may be, for example, an optical system configured to obtain both transmission and fluorescence images.
    The processing apparatus 420 is suitably configured to process images from the acquisition apparatus 410, and to implement an image processing method according to the invention and as described, for example, above.
    The processing apparatus 420 includes a calculation section with all electronic components, software, or other necessary image processing means.
    The processing apparatus 420 includes, for example, a programmable processor, a memory module, and at least one input that is coupled to a bus system.
    The processor may, for example, comprise a field-programmable gate-array circuit (FPGA) or a microprocessor, or a CPU processor or a workstation processor. The processing device 420 may be or include a computer operable through a touch screen.
    The memory module may, for example, comprise a hard disk and / or a ROM, and / or a RAM and / or a magnetic or optical storage system.
    Image processing algorithms as described above can be stored in the memory module to perform image processing in accordance with one of the embodiments of the present invention.
    A program for carrying out the method according to the invention can further be recorded on a medium (e.g. CD-ROM or DVD-ROM, or removable USB media, or magnetic media or hard disk or memory card, such as SDRAM) that can be read by the processing device 420.
    This program can also be stored in a server 430 at a remote location that can communicate with the processing device 420. Thus, the processing device 420 can be further connected to a network 440 over which data relating to the image processing method is transmitted or received.
    The image processing method according to the invention can find application in various fields such as microscopy, automation, artificial vision.
    Another application example is the accurate detection of items that are arranged approximately in one location, such as an object on a conveyor belt or positioned in a recording space with relative inaccuracy.
    The image processing method can further be used, for example, to align a device, or to pick up, observe or manipulate an item.
    权利要求:
    Claims (14)
    [1]
    CONCLUSIONS
    An image-processing method for determining the position of at least one element (2,102) in an image, the element having a known shape, the method comprising steps performed by at least one processor, the steps comprising: a) forming a first distribution (C1) comprising a first identifiable signature (21, 22) related to first features (11, 12) of the known form and a second distribution (C2) comprising a second identifiable signature (23, 24 ) related to second features (13, 14) of the known shape, the first distribution being formed by a first operation comprising: forming a first series (A1) of selected pixel lines (R1) of the image and having it with each selected pixel line (R1) of the first set associate a first index value (i) and a first distribution value (Vi), the first distribution value being dependent on respective values of the pixels representing the forming an electric line (R1), the first distribution (C1) being formed by the first index values (i) and the first distribution values (Vi), the second distribution being formed by a second operation comprising: forming a second series ( A2) of selected pixel lines (Rj) of the image and associating with each selected line (Rj) of the second set a second index value (j) and a second distribution value (Vj) depending on respective values of the pixels that the selected line (Rj), wherein the second distribution (C2) is formed by the second index values (j) and the second distribution values (Vj), the first series (A1) and the second series (A2) being chosen to form a grid of selected pixel lines (R1, Rj) for forming a coordinate system, b) identifying the first signature in the first distribution and the second signature in the second distribution to derive position data according to said coordinate nsystem by identifying the pixel lines of the first series corresponding to the first signature in the first distribution and the pixel lines of the second series corresponding to the second signature in the second distribution, c) deriving from the position data according to the coordinate system of the position of the element (2, 102,202) in the image.
    [2]
    The method of claim 1, wherein processing is such that the value associated with a given line of pixels is a function of a sum of the respective values of pixels included in the given line.
    [3]
    The method of claim 1 or 2, wherein the operation is such that the value associated with a given line of pixels is a function of an average of respective values of pixels included in the given line.
    [4]
    The method of any one of claims 1 to 3, wherein the processing is such that the value associated with a given line of pixels is a function of a maximum of respective values of pixels in the given line.
    [5]
    The method of any one of claims 2-4, wherein the value is further a function of a weighting factor.
    [6]
    The method according to any of claims 1 to 5, wherein identifying the first signature in the first distribution and the second signature in the second distribution comprises the steps of: - calculating a convolution between the first distribution and a first predetermined template that is a first function comprising a model of the first signature, - calculating a convolution between the second distribution and a second predetermined template that is a second function comprising a model of the second signature.
    [7]
    The method of any one of claims 2-6, wherein the lines of the first series and the lines of the second series correspond to horizontal rows and vertical rows of the image, respectively.
    [8]
    The method of any one of claims 2-6, wherein before step a), the element further has a known orientation relative to the horizontal direction and vertical direction of the image, and wherein the respective orientation of the lines of the first series and the lines of the second series selected during step a) depend on the known orientation.
    [9]
    The method of any one of claims 2-6, wherein the lines of the first series and / or the second series are non-linear lines.
    [10]
    A computer program comprising program code instructions for enabling computer processing means to perform an image processing method according to any of claims 1 to 9.
    [11]
    A computer program product comprising program code instructions stored in a medium to be used by a computer by computer processing means, comprising readable programming means for performing an image processing method according to any one of claims 1 to 9.
    [12]
    A digital data storage device readable by computer processing means, comprising code instructions from a computer program according to claim 11.
    [13]
    An image processing device for determining the position of at least one element (2, 102) in an image, the element having a known shape, the image processing device comprising processing means, the processing means comprising: - means for forming a first distribution (C1) comprising a first identifiable signature (21, 22) related to first features (11, 12) of the known form and a second distribution (C2) comprising a second identifiable signature (23, 24) related to second features (13 14) of the shape, the first distribution being formed by an operation comprising: forming a first series (A1) of selected pixel lines (R1) of the image and associating with each selected pixel line (R1) of the first series of a first index value (i) and a first distribution value (Vi), the first distribution value being dependent on respective values of the pixels that the selected line ( R1), wherein the first distribution (C1) is formed by the first index values (i) and the first distribution values (Vi), the second distribution being formed by an operation comprising: forming a second series (A2) of selected pixels lines (Rj) and associating with each selected line (Rj) of the second set a second index value (j) and a second distribution value (Vj) depending on respective values of the pixels forming the selected line (Rj), the second distribution (C2) is formed by the second index values (j) and the second distribution values (Vj), the first series (A1) and the second series (A2) being chosen to form a grid of selected pixel lines (Ri, Rj ) forming a coordinate system, - means for identifying the first signature in the first distribution and the second signature in the second distribution so as to derive position data according to the coordinate system by the identif icing the pixel lines of the first series corresponding to the first signature in the first distribution and the pixel lines of the second series corresponding to the second signature in the second distribution, - means for deriving the position from the position data according to the coordinate system of the element (2, 102,202) in the image.
    [14]
    A molecular diagnostic platform designed to receive a support with a reference mark, the platform being equipped with detection means for detecting the position of the support using an image of the support, the detection means being: - arranged to implement an image processing method according to any of claims 1-9 and / or comprise an image processing system according to claim 13.
    类似技术:
    公开号 | 公开日 | 专利标题
    US8086041B2|2011-12-27|Pattern evaluation method, pattern matching method and computer readable medium
    JP3524853B2|2004-05-10|Pattern inspection apparatus, pattern inspection method, and recording medium
    JP5699788B2|2015-04-15|Screen area detection method and system
    KR20170005379A|2017-01-12|Using high resolution full die image data for inspection
    KR102021945B1|2019-09-17|Determination of localised quality measurements from a volumetric image record
    JP2011017705A|2011-01-27|Pattern inspection device, pattern inspection method, and recording medium
    JP2016502750A5|2016-11-24|
    TWI492165B|2015-07-11|Pattern matching method and device
    US8731274B2|2014-05-20|Method and system for wafer registration
    US10545490B2|2020-01-28|Method of inspecting a specimen and system thereof
    JP4865147B2|2012-02-01|Pattern measuring method and measuring system using display microscope image
    CN110286124B|2022-02-18|Machine vision-based refractory brick measuring system
    JP5548780B2|2014-07-16|Image processing apparatus and computer program
    CN113688807B|2022-02-08|Self-adaptive defect detection method, device, recognition system and storage medium
    JP6609954B2|2019-11-27|DNA chip image spot effectiveness determination apparatus, DNA chip image spot effectiveness determination method, and DNA chip image spot effectiveness determination program
    JP2014228357A|2014-12-08|Crack detecting method
    BE1024091B1|2017-11-13|Method for precise location of elements in an image
    JP4346379B2|2009-10-21|Defect inspection method
    JP5155938B2|2013-03-06|Pattern contour detection method
    JP2018096908A|2018-06-21|Inspection device and inspection method
    JP2010133744A|2010-06-17|Defect detection method, and visual inspection device using the same
    JP5389456B2|2014-01-15|Defect inspection apparatus and defect inspection method
    JP2006031167A|2006-02-02|Barcode recognition device, barcode recognition method and barcode recognition program
    Roy et al.2013|Abnormal regions detection and quantification with accuracy estimation from MRI of brain
    Lam et al.2010|Segmentation of center brains and optic lobes in 3D confocal images of adult fruit fly brains
    同族专利:
    公开号 | 公开日
    EP3096266A1|2016-11-23|
    BE1024091A1|2017-11-10|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    US8401250B2|2010-02-19|2013-03-19|MindTree Limited|Detecting objects of interest in still images|
    法律状态:
    2018-02-08| FG| Patent granted|Effective date: 20171113 |
    2020-04-16| MM| Lapsed because of non-payment of the annual fee|Effective date: 20190731 |
    优先权:
    申请号 | 申请日 | 专利标题
    EP15168521.1|2015-05-20|
    EP15168521.1A|EP3096266A1|2015-05-20|2015-05-20|Method for precise localization of elements on an image|
    [返回顶部]