WO2021174506A1 - Procédé et appareil de détection de bord d'image, et processeur de signal d'image - Google Patents

Procédé et appareil de détection de bord d'image, et processeur de signal d'image Download PDF

Info

Publication number
WO2021174506A1
WO2021174506A1 PCT/CN2020/078066 CN2020078066W WO2021174506A1 WO 2021174506 A1 WO2021174506 A1 WO 2021174506A1 CN 2020078066 W CN2020078066 W CN 2020078066W WO 2021174506 A1 WO2021174506 A1 WO 2021174506A1
Authority
WO
WIPO (PCT)
Prior art keywords
pixel
identifier
edge
target image
image
Prior art date
Application number
PCT/CN2020/078066
Other languages
English (en)
Chinese (zh)
Inventor
赵文扬
高明明
张�林
任靖晔
Original Assignee
深圳市大疆创新科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳市大疆创新科技有限公司 filed Critical 深圳市大疆创新科技有限公司
Priority to PCT/CN2020/078066 priority Critical patent/WO2021174506A1/fr
Publication of WO2021174506A1 publication Critical patent/WO2021174506A1/fr

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection

Definitions

  • This application relates to the field of image processing, and in particular to an image edge detection method, device, and image signal processor.
  • Edge refers to a collection of pixels with sharp changes in the gray levels of surrounding pixels in an image, and is the most basic feature of an image. Edge detection plays an important role in many computer vision tasks, and can be widely used in areas such as feature extraction, image recognition, image segmentation, image matching, and image classification. The quality of image edge detection determines the effect of subsequent image processing to a large extent. A good edge detection method can provide better guarantee for subsequent image processing.
  • one of the objectives of the embodiments of the present application is to provide an image edge detection method, device, and image signal processor.
  • an image edge detection method including:
  • each pixel in the target image corresponds to an initial edge identifier, and the initial edge identifier is used to describe the edge information contained in the pixel;
  • the edge image of the target image is determined.
  • an image edge detection device including an image rotation module and a scanning module
  • the image rotation module is used to rotate the target image by a specified angle; each pixel in the target image corresponds to an initial edge identifier, and the initial edge identifier is used to describe the edge information contained in the pixel;
  • the scanning module is configured to scan pixels in the rotated target image along a designated scanning direction, and determine a new edge identifier for each pixel;
  • the device is further configured to determine the edge image of the target image according to the new edge identifier of each pixel.
  • an image signal processor including the device according to any one of the second aspects, and the image signal processor is configured to process image signals collected by an image sensor.
  • Fig. 1A is a schematic diagram showing a target image according to an exemplary embodiment of the present application.
  • Fig. 1B is a schematic diagram of an edge image obtained by scanning in a related technology according to an exemplary embodiment of the present application.
  • Fig. 2 is a flowchart of a first image edge detection method according to an exemplary embodiment of the present application.
  • Fig. 3A is a flow chart showing a second method for image edge detection according to an exemplary embodiment of the present application.
  • Fig. 3B is a flowchart of a third image edge detection method according to an exemplary embodiment of the present application.
  • Fig. 4A is a schematic diagram obtained by rotating Fig. 1A clockwise by 90° according to an exemplary embodiment of the present application.
  • Fig. 4B is a schematic diagram of an edge image obtained after scanning Fig. 4A according to an exemplary embodiment of the present application.
  • Fig. 4C is a schematic diagram obtained by rotating Fig. 4B clockwise by 90° according to an exemplary embodiment of the present application.
  • Fig. 4D is a schematic diagram of an edge image obtained after scanning Fig. 4C according to an exemplary embodiment of the present application.
  • Fig. 4E is a schematic diagram obtained by rotating Fig. 4D clockwise by 90° according to an exemplary embodiment of the present application.
  • Fig. 4F is a schematic diagram of an edge image obtained after scanning Fig. 4E according to an exemplary embodiment of the present application.
  • Fig. 4G is a schematic diagram obtained by rotating Fig. 4F clockwise by 90° according to an exemplary embodiment of the present application.
  • Fig. 5 is a schematic diagram showing a pixel Q and its 3 ⁇ 3 neighboring pixels according to an exemplary embodiment of the present application.
  • Fig. 6 is a schematic diagram showing a pulsation unit according to an exemplary embodiment of the present application.
  • Fig. 7A is a schematic diagram of a single pulsation unit in cascade-connected pulsation units according to an exemplary embodiment of the present application.
  • Fig. 7B is a schematic diagram showing a cascaded pulsation unit according to an exemplary embodiment of the present application.
  • Fig. 8 is a structural diagram of an image edge detection device according to an exemplary embodiment of the present application.
  • Edge refers to a collection of pixels with sharp changes in the gray levels of surrounding pixels in an image, and is the most basic feature of an image. Edge detection plays an important role in many computer vision tasks, and can be widely used in areas such as feature extraction, image recognition, image segmentation, image matching, and image classification. The quality of image edge detection determines the effect of subsequent image processing to a large extent. A good edge detection method can provide better guarantee for subsequent image processing.
  • the canny edge detection algorithm firstly performs Gaussian smoothing on the image, and then calculates the gradient intensity and gradient direction of each pixel in the image, and preliminarily determines the edge nature of the pixel through the non-maximum value suppression step and the double threshold detection step , Divide all pixels into 3 categories, namely strong edge pixels, weak edge pixels and non-edge pixels. Finally, edge detection is completed by suppressing isolated weak edges; the so-called "suppression of isolated weak edges" means that if weak If an edge pixel is connected to a strong edge pixel, the weak edge pixel is considered to be a strong edge pixel and retained, otherwise, the weak edge pixel is considered to be a non-edge pixel and discarded.
  • the forward single scan method is usually used to extend in a single direction, and only the edge information in one direction can be traversed at a time.
  • Figure 1A In the original image of the "suppression of isolated weak edges" process, suppose that the black grid in Figure 1A represents pixels with strong edges, the gray grid represents pixels with weak edges, and the white grid represents pixels with non-edges.
  • the forward single scan method Single direction (assuming that the scanning direction is from left to right and top to bottom) to extend, first scan from the first pixel in the upper left corner of the image, first scan the first row of pixels from left to right, then scan After finishing the first row of pixels, select the second row from top to bottom, and then scan the second row of pixels from left to right, and so on, until the last row of pixels is scanned, because the scanning direction cannot be changed , Only the scanning result shown in Fig. 1B can be obtained, complete edge information cannot be detected, and the edge connectivity of the detection result is poor.
  • FIG. 2 is a flowchart of an image edge detection method according to an exemplary embodiment of the present application.
  • the image edge detection method of the embodiment of the present application rotates the target image, so that the target image can be scanned from different directions and dimensions of the target image using a forward single scan method to obtain complete edge information. , which helps to ensure the accuracy of the scan results.
  • the image edge detection method of the embodiments of the present application can be applied to electronic devices that require image processing, such as smart phones, computers, tablets, image capture devices (such as cameras), and movable devices (such as unmanned aerial vehicles, unmanned vehicles). , Unmanned boats or mobile robots), etc.
  • the method includes:
  • step S101 a target image is acquired; each pixel in the target image corresponds to an initial edge identifier, and the initial edge identifier is used to describe the edge information contained in the pixel.
  • step S102 after the target image is rotated by a specified angle, the pixels in the rotated target image are scanned along the specified scanning direction to determine a new edge identifier for each pixel.
  • step S103 the edge image of the target image is determined according to the new edge identifier of each pixel.
  • the initial edge identifier of the pixel may be determined in the following manner: Use a Gaussian filter to process the target image to filter out noise in the target image, and obtain a processed target image , And then calculate the gradient value and gradient direction of each pixel in the processed target image, and finally determine the initial edge identifier of the pixel based on the gradient direction and gradient value of the pixel, and the initial edge identifier Used to describe the edge information contained in the pixel.
  • the edge information contained in the pixel can be represented by the gradient value of the pixel, according to the order of the strength of the edge information contained in each pixel in the target image from strong to weak (gradient value from In order of large to small), the edge identifiers are sequentially divided into a first identifier, a second identifier, and a third identifier. It should be noted that the first identifier, the second identifier, and the third identifier are mutually exclusive. Are not the same. In an example, the first identifier indicates that the pixel is a strong edge pixel, the second identifier indicates that the pixel is a weak edge pixel, and the third identifier indicates that the pixel is a non-edge pixel.
  • the target image is rotated by a specified angle along a specified rotation direction, and then pixels in the rotated target image are scanned along the specified scanning direction to determine a new information about each pixel.
  • the edge identification, and finally the edge image of the target image can be determined according to the new edge identification of each pixel; in this embodiment, the target image is rotated and then scanned to achieve different directions and dimensions. The scanning process ensures that complete and accurate edge information of the target image is obtained.
  • the specified scanning direction includes, but is not limited to, a top-down direction, a bottom-up direction, a left-to-right direction or a right-to-left direction, or at least two of the above directions.
  • the target image is rotated by a specified angle along a specified rotation direction, and then each pixel in the rotated target image is detected along the specified scanning direction.
  • the initial edge identifier of the pixel meets the first specified condition, and the current edge identifier of any neighboring pixel corresponding to the pixel (according to the designated scanning direction, if the neighboring pixel corresponding to the pixel is in the pixel It has been scanned before, the current edge identifier of the neighboring pixel is the new edge identifier, if the neighboring pixel corresponding to the pixel has not been scanned before the pixel, the current edge identifier of the neighboring pixel Is the initial edge identification) that meets the second specified condition, then the edge identification of the pixel is changed to obtain the new edge identification of the pixel; wherein, the first specified condition includes the initial edge of the pixel The identifier is the second identifier, and the second designated condition includes that the current edge identifier of any neighboring pixel
  • the neighboring pixels of the pixel include all pixels in the 3 ⁇ 3 neighborhood of the pixel.
  • the scanning operation performed on the entire target image is a forward single scan.
  • the scanning direction is left-to-right and top-down scanning.
  • Figure 1A First scan from the first pixel in the upper left corner of the image, and scan the first row of pixels from left to right. After scanning the first row of pixels, select the second row of pixels from top to bottom, and then scan the second row of pixels from left to right, and so on, until the last row of pixels is scanned, it is visible, positive
  • the single scan process includes multiple line scans. Taking into account the problem that the single scan direction of the forward single scan algorithm leads to incomplete edge information of the acquired target image, in this embodiment, the target image is rotated by a specified angle to realize scanning of the rotated target image from different directions and dimensions.
  • the target image is rotated at least twice.
  • the target image needs to be scanned in a forward direction at least twice. For example, after the target image is rotated by a specified angle, the entire rotated target image is scanned once from left to right and from top to bottom. If after scanning the entire rotated target image at one time, the number of pixels in the rotated target image whose initial edge identification has been changed is 0, there is no need to perform the operations of rotating the target image and scanning the target image, and the end The edge detection process of the target image obtains the edge image of the target image based on the new edge identifier of each pixel in the target image.
  • FIG. 3A provides a schematic flowchart of a second image edge detection method according to an exemplary embodiment of the present application, and the method includes:
  • step S201 a target image is acquired.
  • step S202 the target image is rotated by a specified angle along a specified rotation direction.
  • each pixel in the rotated target image is sequentially detected along a designated scanning direction, and if the initial edge identifier of the pixel is detected as the second identifier, and the pixel corresponds to The current edge identifier of any neighboring pixel is the first identifier, the initial edge identifier of the pixel is changed from the second identifier to the first identifier, and the new edge identifier of the pixel is determined.
  • step S204 it is determined whether the number of pixels of the initial edge identifier changed in the target image scan is 0; if not, skip to step S202, if yes, skip to step S205.
  • the target image after the previous rotation is rotated by the specified angle along the specified rotation direction.
  • step S205 the edge image of the target image is determined according to the new edge identifier of each pixel.
  • the target image is repeatedly rotated by the designated angle in a designated rotation direction until the target image is changed from the second identification to the first identification.
  • the number of pixels is 0.
  • the pixels in the rotated target image are scanned along the specified scanning direction to determine the new edge identifier of each pixel;
  • After scanning the second target image determine the number of pixels in the target image whose initial edge identifier has been changed, that is, the number of pixels changed from the second identifier to the first identifier.
  • step S202 and step S203 If the number is 0, it means If the edge information of the target image has been completely obtained, there is no need to repeat the operations of step S202 and step S203, but jump to step S205, and determine the target image according to the new edge identifier of each pixel. Edge image, otherwise, repeat the operations of step S202 and step S203; this embodiment ensures that after multiple rotations and scans of the target image, complete edge information of the target image can be obtained.
  • FIG. 3B provides a third image according to an exemplary embodiment of this application
  • a schematic flow chart of an edge detection method the method includes:
  • step S301 a target image is acquired.
  • step S302 the target image is rotated by a specified angle along a specified rotation direction.
  • each pixel in the rotated target image is sequentially detected along a designated scanning direction, and if the initial edge identifier of the pixel is detected as the second identifier, and the pixel corresponds to The current edge identifier of any neighboring pixel is the first identifier, the initial edge identifier of the pixel is changed from the second identifier to the first identifier, and the new edge identifier of the pixel is determined.
  • step S304 it is determined whether the target image has been rotated at least one circle from the initial position and the number of pixels of the initial edge identifier is 0 during the target image scanning; if not, skip to step S302, if yes , Jump to step S305.
  • the target image after the previous rotation is rotated by the specified angle along the specified rotation direction.
  • step S305 the edge image of the target image is determined according to the new edge identifier of each pixel.
  • the target image may be repeatedly rotated by the specified angle in a specified rotation direction, until the target image is rotated at least one circle from the initial position and the target image is changed Since the number of pixels of the first identification is 0, each time the target image is rotated by a specified angle along the specified rotation direction, the pixels in the rotated target image are scanned along the specified scanning direction to determine each pixel A new edge identifier; after each scan, determine whether the number of pixels in the target image whose initial edge identifier has been changed is 0, that is, the target image is changed from the second identifier to the first Whether the number of identified pixels is 0 and whether the target image has been rotated at least one circle (360°) from the initial position, the number of which is 0, indicating that the edge information of the target image has been completely obtained, and there is no need to repeat the steps
  • the embodiment of the present application does not impose any restrictions on the specified rotation direction and the specified angle, and can be specifically set according to actual application scenarios;
  • the specified rotation direction may be clockwise or counterclockwise, and
  • the specified angle may be any value between 0° and 360°, where the specified angle may be 90°.
  • the specified rotation direction is clockwise, the specified angle is 90°, and the specified rotation direction is from top to bottom and from left to right, and it needs to satisfy that the target image is from the initial position
  • the first rotation and scanning process Rotate the target image in Figure 1A by 90° in a clockwise direction to obtain the image shown in Figure 4A.
  • Comparing the image shown in Figure 4A from top to bottom and from left to right Scan the pixels in the image line by line one by one (first scan from the first pixel in the upper left corner of the image, first scan the first line of pixels from left to right, and then scan the first line of pixels from top to bottom Select the second row of pixels in the direction, and then scan the second row of pixels from left to right, and so on, until the last row of pixels is scanned), if the initial edge identifier of the pixel meets the first specified condition (the The first designated condition includes that the initial edge identifier of the pixel is the second identifier), and the current edge identifier of any neighboring pixel corresponding to the pixel meets the second designated condition (the second designated condition includes the The current edge identifier of any neighboring pixel corresponding to the pixel is the first identifier), the initial edge identifier of the pixel is
  • the second rotation and scanning process Rotate the target image in Figure 4B by 90° in a clockwise direction to obtain the image shown in Figure 4C. If the initial edge identifier of the pixel meets the first specified condition, and the current edge identifier of any neighboring pixel corresponding to the pixel meets the second specified condition, the initial edge identifier of the pixel is changed from The second mark is changed to the first mark, and the new edge mark of the pixel is determined to be the first mark, thereby obtaining a scanned image as shown in FIG.
  • the third rotation and scanning process Rotate the target image in Figure 4D along a clockwise direction by 90° to obtain the image shown in Figure 4E, and compare the image shown in Figure 4E from top to bottom and from left to right. If the initial edge identifier of the pixel meets the first specified condition, and the current edge identifier of any neighboring pixel corresponding to the pixel meets the second specified condition, the initial edge identifier of the pixel is changed from The second mark is changed to the first mark, and the new edge mark of the pixel is determined to be the first mark, thereby obtaining a scanned image as shown in FIG.
  • the fourth rotation and scanning process Rotate the target image in Figure 4F by 90° in a clockwise direction to obtain the image shown in Figure 4G, and compare the image shown in Figure 4G from top to bottom and from left to right. If the initial edge identifier of the pixel meets the first specified condition, and the current edge identifier of any neighboring pixel corresponding to the pixel meets the second specified condition, the initial edge identifier of the pixel is changed from The second mark is changed to the first mark, and the new edge mark of the pixel is determined to be the first mark, thereby obtaining a scanned image as shown in FIG.
  • the image rotation and scanning process can be implemented by hardware.
  • the information of pixels in the rotated target image is sequentially read along the specified scanning direction, and the pixel information includes the initial edge of the pixel.
  • the pulsation unit is used to scan and detect the pixels, which is beneficial to increase the scanning speed.
  • the target neighborhood pixel may be an n ⁇ n neighborhood corresponding to the pixel.
  • n is an odd number.
  • the target neighborhood pixel may be one or more pixels in the 3 ⁇ 3 neighborhood of the pixel; further, in order to improve scanning efficiency, the target neighborhood pixel may be all neighbors corresponding to the pixel. The scanned part of the pixel in the domain, so that it can be quickly determined whether the pixel meets the condition based on the target neighbor pixel.
  • the number of pixels read each time can be determined according to the bit width of the data bus. The larger the bit width of the data bus, the more pixels are read each time, which is beneficial to increase the scanning speed.
  • the edge identifier is expressed in binary form, and the number of pixels read each time can be determined according to the number of bits of the edge identifier and the bit width of the data bus; at the position of the edge identifier In the case of fixed, the occupied bit width is also fixed. If the bit width of the data bus is larger, the number of pixels read each time is larger; when the bit width of the data bus is fixed, The smaller the representation position of the edge identifier, the smaller the bit width it occupies, and the greater the number of pixels read each time; in one example, the representation of the edge identifier is 2 bits, which occupies 2 bits of Bit width, the bit width of the data bus is 128 bits, and the number of pixels that can be read at a time is 64 bits.
  • the number of the pulsating units can be determined based on the number of pixels.
  • the pixels can be processed in parallel by multiple pulsating units, which is beneficial to Increase the scanning speed of the target image.
  • the pulsating units can be connected in a cascade manner, which is beneficial to increase the scanning speed.
  • the target image is used to perform a single forward scan
  • the specified scan direction is from top to bottom and from left to right
  • the target neighborhood pixel is the 3 ⁇ 3 neighborhood corresponding to the pixel.
  • the scanned part of all the neighboring pixels in the domain is taken as an example for description: read the information of the pixels in the rotated target image in the top-down and left-to-right directions, and the information of the pixels in the The information includes the initial edge identification of the pixel and the current edge identification of the target neighborhood pixel of the pixel.
  • FIG. 5 is a schematic diagram of pixel Q and all neighborhood pixels (a to h) in its 3 ⁇ 3 neighborhood.
  • the target neighborhood pixels are pixel a, pixel b, pixel c, and pixel d, and the initial edge identification of pixel Q and pixel a,
  • the current edge identifiers of the pixel b, the pixel c, and the pixel d are input into the pulsation unit, processed by the pulsation unit, and a new edge identifier of the pixel Q is output.
  • the pulsation unit is used to scan and detect the pixels, which is beneficial to increase the scanning speed.
  • the target neighboring pixels are all neighboring pixels in the 3 ⁇ 3 neighborhood corresponding to the pixel that have been scanned
  • the information of the pixels includes the initial edge identifier of the pixel and the current edge identifier of the target neighboring pixel of the pixel.
  • FIG. 5 is A schematic diagram of pixel Q and all neighboring pixels (a ⁇ h) in its 3 ⁇ 3 neighborhood. Since the designated scanning direction is from top to bottom and from left to right, the target neighborhood pixel is pixel a, pixel b. Pixel c and pixel d.
  • each pixel represents the edge identifier in binary form
  • 00 represents the third identifier
  • 01 represents the second identifier
  • 11 and 10 represent the first identifier
  • the 0th bit is denoted as pixel [0]
  • the first bit is denoted as pixel [ 1]
  • the edge identifier of the pixel is marked as "pixel [1] pixel [0]”
  • the edge identifier of the pixel Q in Figure 5
  • the 0th bit is marked as Q[0]
  • the first bit is marked as Q[ 1]
  • the edge identifier of pixel Q is marked as "Q[1]Q[0]”
  • the binary representation of the edge identifier of its neighboring pixels is the same, for example, the edge identifier of pixel a is marked as "a[1]a[0 ]", the edge identifier of pixel b is marked as "b[1]b[0]", the edge identifier of pixel c is marked as "c[1]c[
  • FIG. 6 is a schematic diagram of a single pulsation unit.
  • Q in [1]), Q out [0] Q in [0]" determines the pixel Q new edge identification Q out [1] Q out [0].
  • FIG. 7A is an example diagram of a single pulsation unit connected in cascade mode.
  • Figure 7B is a cascade connection of multiple pulsation units. Schematic diagram of the connection. The pulsation unit shown in FIG.
  • the input to the pulsation unit includes: a[1], b[1], c[1], d[1], Q in [1] Q in [0], where Q in [1] Q in [0] represents the initial pixel identifier of the pixel Q, processed by the pulsating unit, and outputting the new edge identifier of the pixel Q from the pulsating unit Q out [1] Q out [0] and the carry output pixel Q Q out [1], the Q out [1] is used as identification information of the edge pixel in a neighborhood of pixels.
  • the first detected pixel is pixel Q1, and its target neighboring pixels are a1, b1, c1, d1; the second detected pixel is Q2, and its target The neighboring pixels are a2, b2, c2, Q1; and so on, the Nth (N greater than 0) detected pixel is Qn, and the target neighboring pixels are an, bn, cn, Q(n-1) Give an example.
  • the pulsation unit is used to scan and detect the pixels, which is beneficial to increase the scanning speed.
  • an embodiment of the present application also provides an image edge detection device.
  • the device includes an image rotation module 41 and a scanning module 42.
  • the image rotation module 41 is configured to rotate the target image by a specified angle; each pixel in the target image corresponds to an initial edge identifier, and the initial edge identifier is used to describe the edge information contained in the pixel.
  • the scanning module 42 is configured to scan pixels in the rotated target image along a designated scanning direction, and determine a new edge identifier for each pixel.
  • the device is further configured to determine the edge image of the target image according to the new edge identifier of each pixel.
  • the scanning module 42 is specifically configured to: sequentially detect each pixel in the rotated target image along the specified scanning direction; if the initial edge identifier of the pixel is detected If the first specified condition is met, and the current edge identification of any neighboring pixel corresponding to the pixel meets the second specified condition, the initial edge identification of the pixel is changed.
  • the image rotation module 41 is specifically configured to: repeatedly rotate the target image along the specified rotation direction by the specified angle until the number of pixels of the initial edge identification is changed in the target image scan Up to 0.
  • the image rotation module 41 is specifically configured to: repeatedly rotate the target image by a specified angle along a specified rotation direction, until the target image is rotated at least one circle from the initial position and is changed during scanning of the target image.
  • the number of pixels of the initial edge identification is zero.
  • the edge identifiers are sequentially divided into a first identifier, a second identifier, and a third identifier; The first identifier, the second identifier and the third identifier are different from each other.
  • the first specified condition includes: the initial edge identifier of the pixel is the second identifier.
  • the second specified condition includes: the current edge identifier of any neighboring pixel corresponding to the pixel is the first identifier.
  • the edge information contained in the pixel is represented by the gradient value of the pixel.
  • the first identifier indicates that the gradient value of the pixel is greater than a first threshold.
  • the second identifier represents that the gradient value of the pixel is between the first threshold and the second threshold; the second threshold is less than the first threshold.
  • the third identifier indicates that the gradient value of the pixel is smaller than the second threshold.
  • the scanning module includes a pulsation unit.
  • the scanning module is specifically configured to: sequentially read the pixel information in the rotated target image along the designated scanning direction; the pixel information includes the initial edge identifier of the pixel and the pixel information The current edge identifier of the target neighboring pixel; the information of the pixel is input into the pulsation unit, processed by the pulsation unit, and the new edge identifier of the pixel is output.
  • the target neighboring pixel is a scanned part of all neighboring pixels corresponding to the pixel.
  • the number of pixels read by the scanning module each time is determined according to the bit width of the data bus.
  • the edge identifier is represented in a binary form.
  • the number of pixels read by the scanning module each time is determined according to the number of bits represented by the edge identifier and the bit width of the data bus.
  • the number of the pulsating units is determined based on the number of pixels.
  • the pulsation units are connected in a cascade manner.
  • the designated scanning direction includes at least a left-to-right direction, a right-to-left direction, a top-down direction, or a bottom-up direction.
  • the device further includes an initial edge identification determination module.
  • the initial edge identification determination module is used to process the target image by using a Gaussian filter to obtain the processed target image; calculate the gradient value and gradient direction of each pixel in the processed target image; The gradient direction and gradient value of the pixel determine the initial edge identifier of the pixel.
  • the device embodiments described above are merely illustrative.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in One place, or it can be distributed to multiple network units.
  • Some or all of the modules can be selected according to actual needs to achieve the objectives of the solutions of the embodiments. Ordinary technicians in this neighborhood can understand and implement without creative work.
  • an embodiment of the present application also provides an image signal processor, which includes any one of the above-mentioned devices, and the image signal processor is used to process image signals collected by an image sensor.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

Des modes de réalisation de la présente invention concernent un procédé et un appareil de détection de bord d'image, ainsi qu'un processeur de signal d'image. Après l'obtention d'une image cible, l'image cible est tournée selon un angle déterminé le long d'une direction de rotation déterminée ; ensuite, des pixels de l'image cible tournée sont balayés le long d'une direction de balayage déterminée afin de déterminer un nouvel identifiant de bord pour chaque pixel ; enfin, une image de bord de l'image cible peut être déterminée en fonction du nouvel identifiant de bord de chaque pixel. Dans les modes de réalisation de l'invention, l'image cible est tournée et balayée pour mettre en oeuvre un processus de balayage à partir de dimensions de direction différentes, ce qui permet d'assurer que des informations de bord complètes et précises de l'image cible sont obtenues.
PCT/CN2020/078066 2020-03-05 2020-03-05 Procédé et appareil de détection de bord d'image, et processeur de signal d'image WO2021174506A1 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/078066 WO2021174506A1 (fr) 2020-03-05 2020-03-05 Procédé et appareil de détection de bord d'image, et processeur de signal d'image

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/078066 WO2021174506A1 (fr) 2020-03-05 2020-03-05 Procédé et appareil de détection de bord d'image, et processeur de signal d'image

Publications (1)

Publication Number Publication Date
WO2021174506A1 true WO2021174506A1 (fr) 2021-09-10

Family

ID=77613796

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/078066 WO2021174506A1 (fr) 2020-03-05 2020-03-05 Procédé et appareil de détection de bord d'image, et processeur de signal d'image

Country Status (1)

Country Link
WO (1) WO2021174506A1 (fr)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117686096A (zh) * 2024-01-30 2024-03-12 大连云智信科技发展有限公司 一种基于目标检测的畜禽动物体温检测方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5604821A (en) * 1992-02-28 1997-02-18 The University Of South Florida Structure and method for dynamic scene analysis
CN101493932A (zh) * 2009-03-05 2009-07-29 西安电子科技大学 基于形态Haar小波纹理梯度提取的分水岭纹理图像分割方法
CN101763512A (zh) * 2009-12-11 2010-06-30 西安电子科技大学 高分辨率遥感图像中道路目标的半自动检测方法
CN104272323A (zh) * 2013-02-14 2015-01-07 Lsi公司 使用至少一个附加图像的图像增强和边缘验证方法和设备
CN106682678A (zh) * 2016-06-24 2017-05-17 西安电子科技大学 一种基于支撑域的图像角点检测和分类方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5604821A (en) * 1992-02-28 1997-02-18 The University Of South Florida Structure and method for dynamic scene analysis
CN101493932A (zh) * 2009-03-05 2009-07-29 西安电子科技大学 基于形态Haar小波纹理梯度提取的分水岭纹理图像分割方法
CN101763512A (zh) * 2009-12-11 2010-06-30 西安电子科技大学 高分辨率遥感图像中道路目标的半自动检测方法
CN104272323A (zh) * 2013-02-14 2015-01-07 Lsi公司 使用至少一个附加图像的图像增强和边缘验证方法和设备
CN106682678A (zh) * 2016-06-24 2017-05-17 西安电子科技大学 一种基于支撑域的图像角点检测和分类方法

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117686096A (zh) * 2024-01-30 2024-03-12 大连云智信科技发展有限公司 一种基于目标检测的畜禽动物体温检测方法
CN117686096B (zh) * 2024-01-30 2024-04-16 大连云智信科技发展有限公司 一种基于目标检测的畜禽动物体温检测方法

Similar Documents

Publication Publication Date Title
CN108805023B (zh) 一种图像检测方法、装置、计算机设备及存储介质
CN109509200B (zh) 基于轮廓提取的棋盘格角点检测方法以及计算机可读存储介质
WO2014160433A2 (fr) Systèmes et procédés pour classifier des objets dans des images numériques capturées à l'aide de dispositifs mobiles
EP3617938B1 (fr) Procédé et dispositif de traitement de ligne de voie
WO2020135056A1 (fr) Dispositif et procédé de traitement d'image
US9245194B2 (en) Efficient line detection method
CN114820594B (zh) 基于图像检测板材封边缺陷的方法、相关设备及存储介质
CN111681256A (zh) 图像边缘检测方法、装置、计算机设备和可读存储介质
CN111144337B (zh) 火灾检测方法、装置及终端设备
JP2011134012A (ja) 画像処理装置、その画像処理方法及びプログラム
CN114022503A (zh) 检测方法及检测***、设备和存储介质
CN108229583B (zh) 一种基于主方向差分特征的快速模板匹配的方法及装置
CN113785181A (zh) Oled屏幕点缺陷判定方法、装置、存储介质及电子设备
CN112437948A (zh) 图像诊断支援***及图像诊断支援装置
CN112419207A (zh) 一种图像矫正方法及装置、***
WO2021174506A1 (fr) Procédé et appareil de détection de bord d'image, et processeur de signal d'image
CN113052162B (zh) 一种文本识别方法、装置、可读存储介质及计算设备
CN113283439A (zh) 基于图像识别的智能计数方法、装置及***
WO2024016686A1 (fr) Procédé et appareil de détection de coins
CN110147765A (zh) 一种图像处理方法及装置
CN112883762A (zh) 一种活体检测方法、装置、***及存储介质
KR102452511B1 (ko) 도면의 요소 이미지 검출 방법 및 장치
CN116993654A (zh) 摄像头模组缺陷检测方法、装置、设备、存储介质及产品
US10176399B1 (en) Method and apparatus for optical character recognition of dot text in an image
CN112784737B (zh) 结合像素分割和线段锚的文本检测方法、***及装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20923084

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20923084

Country of ref document: EP

Kind code of ref document: A1