WO2019127503A1 - 一种图像处理方法、设备及系统 - Google Patents

一种图像处理方法、设备及系统 Download PDF

Info

Publication number
WO2019127503A1
WO2019127503A1 PCT/CN2017/120229 CN2017120229W WO2019127503A1 WO 2019127503 A1 WO2019127503 A1 WO 2019127503A1 CN 2017120229 W CN2017120229 W CN 2017120229W WO 2019127503 A1 WO2019127503 A1 WO 2019127503A1
Authority
WO
WIPO (PCT)
Prior art keywords
pixel
rotated
pixels
rotation angle
image
Prior art date
Application number
PCT/CN2017/120229
Other languages
English (en)
French (fr)
Inventor
阳光
韩琨
Original Assignee
深圳配天智能技术研究院有限公司
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 深圳配天智能技术研究院有限公司 filed Critical 深圳配天智能技术研究院有限公司
Priority to PCT/CN2017/120229 priority Critical patent/WO2019127503A1/zh
Priority to CN201780028397.7A priority patent/CN109496326B/zh
Publication of WO2019127503A1 publication Critical patent/WO2019127503A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00—Geometric image transformations in the plane of the image
    • G06T3/60—Rotation of whole images or parts thereof

Definitions

  • the present invention relates to the field of image processing, and in particular, to an image processing method, device and system.
  • image resampling that is, interpolation calculation
  • image resolution is very low
  • the reduction in the amount of information due to image rotation will greatly affect image recognition, greatly reducing the accuracy of image recognition.
  • the technical problem to be solved by the present invention is to provide an image processing method, device and system, which can solve the problem of a decrease in the amount of information due to interpolation during the rotation of an existing image.
  • a technical solution adopted by the present invention is to provide an image processing method, including: acquiring a vector model, and obtaining, by using a vector model, a pixel to be rotated and a pair of adjacent pixels in the image rotation process Corresponding relationship between the influence weight of the corresponding pixel and the rotation angle; acquiring the image block to be rotated, the rotation center, and the target rotation angle; determining, by the vector model, the to-be-rotated image block to be rotated according to the rotation center and the target rotation angle
  • the pixel and its neighboring pixels have an influence weight on the rotated corresponding pixel, and the pixel value of the rotated corresponding pixel is calculated in a pixel fusion manner according to the influence weight.
  • an image processing apparatus including: a communication circuit for acquiring an image block to be rotated, a rotation center, and a target rotation angle; and a processor coupled to the communication
  • the circuit is configured to obtain a vector model, and the corresponding relationship between the weight of the pixel to be rotated and the adjacent pixel of the rotated pixel and the rotation angle during the rotation of the image can be obtained by using the vector model;
  • the rotation center and the target rotation angle determine the influence weight of the pixel to be rotated and the adjacent pixel in the image block to be rotated on the rotated corresponding pixel, and calculate the pixel of the rotated corresponding pixel in a pixel fusion manner according to the influence weight value.
  • an image processing system including: the image processing device and the photographing device as described above, the image processing device being connected to the photographing device for photographing the photographing device The image to be rotated is processed.
  • the beneficial effects of the present invention are: different from the prior art, in some embodiments of the present invention, by acquiring a vector model, it is possible to obtain a corresponding pixel after the pixel to be rotated and its adjacent pixel pair are rotated during image rotation.
  • Corresponding relationship between the influence weight and the rotation angle obtaining the image block to be rotated, the rotation center and the target rotation angle, and using the vector model to determine the pixel to be rotated in the image block to be rotated according to the rotation center and the target rotation angle
  • the adjacent pixels affect the weight of the corresponding pixels after the rotation, and calculate the pixel values of the rotated corresponding pixels in a pixel fusion manner according to the influence weight, so that the influence of the surrounding pixels can be merged into the rotated pixels without performing
  • the interpolation operation does not greatly reduce the amount of information, thereby reducing the influence on image recognition and improving the accuracy of image recognition.
  • FIG. 1 is a schematic flow chart of a first embodiment of an image processing method according to the present invention.
  • FIG. 2 is a schematic flow chart of a second embodiment of an image processing method according to the present invention.
  • FIG. 3 is a schematic diagram of a vector model in an image processing method of the present invention.
  • FIG. 4 is a schematic flow chart of a third embodiment of an image processing method according to the present invention.
  • FIG. 5 is a schematic flow chart of a fourth embodiment of an image processing method according to the present invention.
  • FIG. 6 is a schematic flow chart of a fifth embodiment of an image processing method according to the present invention.
  • FIG. 7 is a schematic diagram of a scene in which an image block to be rotated has an oblique setting in an image to be rotated;
  • step S137 is a schematic flowchart of step S137 in the fifth embodiment of the image processing method of the present invention.
  • FIG. 9 is a schematic structural diagram of a first embodiment of an image processing apparatus according to the present invention.
  • FIG. 10 is a schematic structural diagram of a second embodiment of an image processing apparatus according to the present invention.
  • FIG. 11 is a schematic structural diagram of an embodiment of an image processing system of the present invention.
  • FIG. 12 is a schematic structural diagram of an embodiment of a device having a storage function according to the present invention.
  • the first embodiment of the image processing method of the present invention includes:
  • S11 acquiring a vector model, and obtaining, by using the vector model, a correspondence relationship between a weight of the pixel to be rotated and a corresponding pixel of the adjacent pixel after the rotation of the image and the rotation angle;
  • the vector model may include a pixel position model obtained by rotating the image to be rotated by using the center position of the image to be rotated as a rotation center, and the image model may be obtained during the image rotation process by analyzing the position model. Correspondence between the weight of the pixel to be rotated and its neighboring pixels on the corresponding pixel after the rotation and the rotation angle.
  • the image to be rotated may be a certain area in the image to be rotated, for example, an area containing the workpiece to be identified, or an entire image to be rotated, which is not specifically limited herein.
  • the target rotation angle may be a rotation angle manually input by the user, or may be a preset rotation angle, or a rotation angle that is favorable for observation after the image to be rotated is recognized, and the angle value may be based on actual processing or observation. Depending on the needs, there is no specific limit here.
  • the center of rotation may be the center position of the image block to be rotated, or may be other positions, such as a location selected by the user. For example, when the user does not input the center of rotation, the center of the image to be rotated is rotated by the center of rotation. When the user specifies the center of rotation, the user-specified position is used as the center of rotation.
  • the pixel value corresponding to the pixel to be rotated and its neighboring pixel may be the original pixel value of the pixel to be rotated and its neighboring pixel, or may be the pixel value obtained after the pixel to be rotated and its adjacent pixel are processed.
  • the pixel fusion mode may be a weighted summation or a proportional weighting calculation method, and the specific calculation manner may be determined according to actual needs, and is not specifically limited herein.
  • the image processing apparatus can directly acquire the pre-saved vector model and the target rotation angle, such as user input. 45 degrees, that is, rotate 45 degrees clockwise, and then analyze the vector model according to the target rotation angle, and can obtain the pixel before rotation and the phase after rotating the image block to be rotated (such as the entire image to be rotated) clockwise by 45 degrees.
  • the proportion of the adjacent pixels in the corresponding pixels after the rotation so that the influence of the pixel values of the pixels before rotation and their adjacent pixels on the pixel values of the corresponding pixels after the rotation can be obtained.
  • the image processing device can calculate the pixel values of the pixels before rotation and their neighboring pixels in combination with the corresponding influence weights by using a pixel fusion method (such as weighted summation), and finally obtain the pixel values of the rotated corresponding pixels.
  • a pixel fusion method such as weighted summation
  • the position specified by the user as the rotation center is first acquired, and then the position is substituted into the rotation center of the vector model, thereby further determining the to-be-targeted rotation angle according to the target rotation angle.
  • the pixels to be rotated and their adjacent pixels in the rotated image block are weighted by the corresponding pixels after the rotation, and the pixel values of the rotated corresponding pixels are calculated in a pixel fusion manner according to the influence weight.
  • the pixel values of the pixels before rotation and their neighboring pixels may be processed according to actual needs, and corresponding pixel values are obtained, and then the pixel values of the corresponding pixels after rotation are calculated, for example, the edge is relative to a pixel to be rotated in a row/column direction in which the pixel is to be rotated, and the pixel to be rotated and the adjacent pixel of the image block to be rotated are first calculated when the edge of the image to be rotated is parallel to the row/column direction of the pixel.
  • the pixel value corresponding to the pixel and then calculating the pixel value of the rotated corresponding pixel by using the corresponding pixel value after the inverse calculation.
  • the vector model may directly include the correspondence obtained after the analysis, for example, the correspondence between the rotation angle and the influence weight of the pixel to be rotated at different positions is saved as a correspondence table, so as to facilitate subsequent When the image processing is performed, the influence weight corresponding to the target rotation angle of the pixel at the specified position is directly found in the correspondence table, thereby improving the processing efficiency.
  • the vector model may also be a real-time calculated model according to the obtained target rotation angle and the position of the target pixel, which is not specifically limited herein.
  • a correspondence relationship between the weight of the pixel to be rotated and the adjacent pixel of the rotated pixel and the rotation angle during the image rotation process can be obtained, and the image block to be rotated is obtained.
  • a rotation center and a target rotation angle and the vector model is used to determine the influence weight of the pixel to be rotated and the adjacent pixel in the image block to be rotated on the rotated corresponding pixel according to the rotation center and the target rotation angle, and according to the influence weight
  • the pixel value of the corresponding pixel after rotation is calculated by the pixel fusion method, so that the influence of the surrounding pixels can be merged into the rotated pixel without interpolation, so that the amount of information is not greatly reduced, thereby reducing image recognition. Influence and improve the accuracy of image recognition.
  • Step S11 further includes:
  • S111 a proportional relationship between a pixel to be rotated represented by a predetermined shape and an adjacent pixel thereof and a corresponding area of the corresponding pixel represented by the predetermined shape after being rotated by a rotation angle is used as the influence weight, and the influence weight is The rotation angle and the pixel position are associated to form a vector model.
  • the predetermined shape may be a square shape, a rectangular shape, or another shape such as a polygon, and may be specifically determined according to actual needs, and is not specifically limited herein.
  • the adjacent pixel is an adjacent pixel in the image block to be rotated before the pixel to be rotated is rotated.
  • a pixel to be rotated and a neighboring pixel thereof are represented by a square, wherein adjacent pixels are included in a row/column direction of a pixel arrangement, and a pixel to be rotated in a diagonal direction Adjacent pixels.
  • the pixel to be rotated represented by the square (taking central pixel A as an example) and its adjacent pixels B1 to B8 counterclockwise 45 After the degree of rotation, the proportional relationship between the overlapping area of the corresponding pixel C (square C indicated by a broken line in FIG.
  • the weight is added, and the obtained influence weight is associated with the rotation angle (for example: -45 degrees / 45 degrees counterclockwise) and the position of the pixel, and finally the vector model is obtained. According to the above manner, a vector model corresponding to each rotation angle can be obtained.
  • the method may further include:
  • S1101 Determine, according to an image rotation center and a rotation angle of the image to be rotated, a corresponding pixel of each pixel to be rotated represented by a predetermined shape after rotating the rotation angle around a center of the image rotation.
  • the image rotation center is the center of the image to be rotated.
  • S1102 Determine a proportional relationship between an area to be rotated and an adjacent pixel of the to-be-rotated image after the rotation angle is rotated around a center of the image rotation center and an overlapping area of the corresponding pixel.
  • a rotation center a center position of the image to be rotated
  • a rotation angle such as a counterclockwise rotation of 45 degrees
  • the overlapping area of the adjacent pixels B4, A, B7 and the corresponding pixel G, and the pixel respectively calculates the ratio of the overlapping area of B6, B4, A, and B7 to the total area of the pixel G, that is, the pixel to be rotated B6 and The proportional relationship between the adjacent pixels B4, A, and B7 and the overlapping area of the corresponding pixels G after being rotated by a rotation angle (rotated counterclockwise by 45 degrees). According to the above manner, the corresponding pixel after each pixel to be rotated is rotated, and the proportional relationship between the pixel to be rotated and its adjacent pixels after being rotated by the rotation angle and the overlapping area of the corresponding pixel.
  • each rotation angle and each pixel to be rotated For each rotation angle and each pixel to be rotated, according to the above manner, corresponding pixels of each pixel to be rotated rotated according to each rotation angle are obtained, and each pixel to be rotated and its adjacent pixels are rotated at a rotation angle. The proportional relationship between the area of overlap with the corresponding pixel after rotation. Then, the above step S111 is continued, that is, a vector model corresponding to each rotation angle and each pixel to be rotated can be obtained.
  • the correspondence between the influence weight and the rotation angle and the pixel position may be expressed by a look-up table method or a formula, that is, the correspondence between the influence weight and the rotation angle and the pixel position is saved in one or more lookup tables. In the middle, or the correspondence between the influence weight and the rotation angle and the pixel position is saved as one or more formulas. After obtaining the target rotation angle and the pixel position, the table or formula calculation method can be used to obtain the influence quickly. Weights to improve image processing efficiency.
  • the vector model may be set according to actual requirements, for example, the correspondence between the influence weight and a certain angle range is set, which is not specifically limited herein.
  • step S13 further includes:
  • V mid is the pixel value of the corresponding pixel after rotation
  • V m ' id is the pixel value corresponding to the pixel to be rotated
  • a is the influence weight corresponding to the pixel to be rotated
  • V i is the ith phase of the pixel to be rotated
  • b i is the influence weight corresponding to the i-th adjacent pixel
  • n is the number of adjacent pixels.
  • the adjacent pixels include at least four adjacent pixels disposed adjacent to the pixel to be rotated along a row arrangement direction and a column arrangement direction of pixels in the image block to be rotated.
  • adjacent pixels are arranged adjacent to the pixel to be rotated by four adjacent pixels along the row arrangement direction and the column arrangement direction of the pixels in the image block to be rotated.
  • the image processing device acquires the influence weight a of the pixel value V m ' id of the pixel A to be rotated on the pixel value V mid of the rotated corresponding pixel C according to the vector model, and its adjacent four pixels B2, B4, B5 and Effects of weight of the pixel value V mid corresponding pixel C after B7 pixel value V 1 ⁇ V 4 the rotation weight b after 1 ⁇ b 4, it is possible according to the above formula to calculate the corresponding pixels C after the rotary pixel value V mid .
  • the values of the influence weights b 1 to b 4 may be the same or different, and may be different depending on actual conditions. For example, when the areas where the pixels B2, B4, B5, and B7 overlap with the pixel C are equal, b 1 to b 4 The values are the same.
  • the adjacent pixels may further include adjacent pixels disposed adjacent to the pixel to be rotated in a diagonal direction of the pixel arrangement, such as B1, B3, B6, and B8 in FIG.
  • This embodiment can also be combined with the second embodiment of the image processing method of the present invention.
  • Step S13 further includes:
  • S133 Perform weighted summation on the actual pixel values of the pixels to be rotated and their neighboring pixels in the rotated image block according to the influence weight, and then use the calculated weighted pixel value as the pixel value of the rotated corresponding pixel.
  • the image processing device when the vector model uses the lookup table to represent the correspondence between the influence weight and the rotation angle and the pixel position, the image processing device obtains the target rotation angle and the pixel position, and may be based on the pixel to be rotated. And the position of the neighboring pixels in the lookup table finds the influence weight of each pixel under the target rotation angle, and then weights the actual pixel values of the pixel to be rotated and its adjacent pixels in the rotated image block according to the acquired influence weight And, finally, the pixel value of the corresponding pixel after the rotation can be calculated.
  • the pixel to be rotated A and its adjacent four pixels B2, B4, B5, and B7 are taken as an example, wherein the target rotation angle can be obtained by using a vector model lookup table (
  • the corresponding pixel A has a weighting effect of 7/9 on the rotated pixel C
  • B2, B4, B5, and B7 have a weighting effect on the rotated pixel C of 1/18
  • the actual pixel values of B4, B5, and B7 are 128, 220, 160, 100, and 40, respectively, and the weighted sum can be obtained by using the above formula (1):
  • the pixel value V mid of the rotated pixel C can be obtained.
  • the pixel values of other pixels in the image block to be rotated can also be calculated by the above formula (1), and finally the pixel value of the entire rotated image can be obtained.
  • the adjacent pixels of other pixels may further include pixels arranged adjacent to each other in the diagonal direction of the pixel arrangement, and the influence weights of the adjacent pixels may also be different, and the adjacent pixels in the vector model are rotated and rotated.
  • the overlapping area ratio of the corresponding pixels is determined, and is not specifically limited herein.
  • This embodiment can also be combined with the second or third embodiment of the image processing method of the present invention or a combination thereof.
  • the edge of the image block to be rotated has an angle with the row/column direction of the pixel arrangement
  • the edge of the workpiece has an angle with the row direction of the pixel arrangement, that is, the workpiece is The image is tilted, and the angle is the tilt angle.
  • the pixel in the image block to be rotated needs to be processed, and the initial pixel of the pixel to be rotated when the edge of the workpiece is aligned with the row direction of the pixel arrangement is inversely calculated.
  • the value using the initial pixel value and the angle of the tilt angle and the target rotation angle superimposed, can accurately calculate the pixel value of the rotated corresponding pixel, avoid the edge blur problem after the rotation process due to the tilt, and is beneficial to improve subsequent image recognition. Precision.
  • Step S13 further includes:
  • the reference direction is a row arrangement direction or a column arrangement direction of pixels in the image block to be rotated.
  • edge detection may be performed on the rotated image to obtain an image block to be rotated.
  • the edge detection method may detect the connected area as the image block to be rotated by using a connected domain detection method, and the image block to be rotated may be an image of the workpiece to be detected or the workpiece to be detected and the detection station, which may be actual. Demand settings, not specifically defined here.
  • the image block to be rotated acquired after the edge detection of the image to be rotated includes D and E.
  • the direction is tilted, for example, to determine that the blocks D and E are near the edge of the bottom of the image. with Whether it is consistent with the x-axis or y-axis direction shown in Figure 7, if they are inconsistent, there is an angle, as shown in Figure 7, with If there are angles ⁇ and ⁇ with respect to the x-axis direction, it can be determined that the blocks D and E are inclined with respect to the preset reference direction.
  • S135 detecting a tilt angle of the image block to be rotated if the setting is inclined with respect to a preset reference direction;
  • the tilt angles of the blocks D and E may be respectively detected, that is, the detection and acquisition of the clip are performed.
  • the values of the angles ⁇ and ⁇ , wherein the edges of the blocks D and E can be acquired by means of edge detection or the like with The angle between the direction of the direction and the x-axis direction ⁇ and ⁇ , or can be detected with The slope of the slope is calculated by the angles ⁇ and ⁇ of the x-axis direction.
  • the specific detection method can be determined according to actual needs.
  • the first group of influence weights corresponding to the tilt angles ⁇ and ⁇ may be obtained by using a vector model, which is specifically obtained by a table lookup or a formula calculation, where No specific restrictions. Then, according to the actual pixel value of the pixel to be rotated and its neighboring pixels in the block D and the first group of weight values corresponding to the tilt angle ⁇ , the edge of the image block D to be rotated can be inversely calculated.
  • the initial pixel value can be estimated by the formula (1) in the above step S131.
  • the initial pixel values of the pixels to be rotated and their neighboring pixels in the block E can also be calculated in the same manner.
  • step S137 includes:
  • S1371 The inverse pixel calculation is performed to obtain the actual pixel value of the pixel to be rotated by weighting and summing the initial pixel values of the pixel to be rotated and its neighboring pixels calculated by the first group of influence weights.
  • the initial pixel value of the pixel to be rotated and its adjacent pixels may be established by using the above formula (1). And the equation between the actual pixel value, and then the equation between the initial pixel value and the actual pixel value of one adjacent pixel and its neighboring pixels can be established, and the initial of other pixels and their adjacent pixels can also be established.
  • S138 acquire, by using a vector model, a second set of influence weights corresponding to the superimposed angle of the tilt angle and the target rotation angle;
  • S139 Perform weighted summation on the inversely derived pixel to be rotated and the initial pixel value of the adjacent pixel according to the second set of influence weights, and then calculate the obtained weighted pixel value as the pixel value of the rotated corresponding pixel.
  • the superposition angle ⁇ +30 of the tilt angle ⁇ and the target rotation angle (eg, 30 degrees) can be found by the vector model.
  • Corresponding second group of influence weights so that the inverse pixel of the pixel to be rotated and the initial pixel value of the adjacent pixel can be weighted and summed by using the above formula (1), and finally the pixel value of the corresponding pixel after rotation can be obtained.
  • the calculation can also be performed in the same manner as described above.
  • the edge-tilted image block to be rotated may also be directly calculated by using the method provided by any one of the first to fourth embodiments of the image processing method of the present invention, that is, The vector model is directly obtained by the target rotation angle, and the rotated pixel value is calculated. It is not necessary to inverse the initial pixel value, and finally the rotated image is obtained.
  • the initial pixel value when the edge of the image block to be rotated is consistent with the row direction of the pixel arrangement is inversely calculated, and the initial pixel value is utilized.
  • the angle of the tilt angle and the target rotation angle superimposed the pixel value of the corresponding pixel after the rotation can be accurately calculated, and the edge blurring problem after the rotation processing due to the tilt is avoided, which is advantageous for further improving the accuracy of the subsequent image recognition.
  • the first embodiment 10 of the image processing apparatus of the present invention includes:
  • a communication circuit 102 configured to acquire an image block to be rotated, a rotation center, and a target rotation angle
  • the processor 101 is coupled to the communication circuit 102 for acquiring a vector model.
  • the vector model can obtain the influence weight and the rotation angle of the pixel to be rotated and the adjacent pixel of the rotated pixel during the image rotation. Corresponding relationship between the two, according to the rotation center and the target rotation angle, determining the influence weight of the pixel to be rotated and its neighboring pixels in the image block to be rotated on the rotated corresponding pixel, and pixel fusion according to the influence weight The method calculates the pixel value of the corresponding pixel after the rotation.
  • the processor 101 may also acquire the vector model through the communication circuit 102, and the vector model may be stored inside the image processing device 10 or may be stored in an external device connected to the image processing device 10 ( In the drawings, it is not specifically limited herein.
  • the processor 101 is further configured to: when the vector model is acquired, overlap the pixel to be rotated and its neighboring pixels represented by the predetermined shape after rotating by the rotation angle and the corresponding pixel represented by the predetermined shape The proportional relationship between the areas is used as the influence weight, and the influence weight is associated with the rotation angle and the pixel position, and finally a vector model is formed.
  • the correspondence between the influence weight and the rotation angle may be represented by a look-up table method or a formula.
  • the processor 101 is further configured to: when a pixel to be rotated represented by a predetermined shape and an adjacent pixel thereof are rotated by a rotation angle and a ratio of overlapping areas of corresponding pixels represented by a predetermined shape
  • each pixel to be rotated represented by the predetermined shape is determined to be at the center of the image rotation according to the image rotation center and the rotation angle Rotating the corresponding pixels after the rotation angle for the center, and determining a proportional relationship between the rotation area of each pixel to be rotated and its adjacent pixels after rotating the rotation angle around the rotation center of the image and the overlapping area of the corresponding pixel.
  • the processor 101 is further configured to perform pixel fusion by using the above formula (1) when calculating the pixel value of the rotated corresponding pixel.
  • the adjacent pixels include at least four adjacent pixels disposed adjacent to the pixel to be rotated along a row arrangement direction and a column arrangement direction of pixels in the image block to be rotated.
  • the processor 101 is further configured to: acquire, by using a vector model, an influence weight corresponding to the target rotation angle and the pixel position, and treat, according to the influence weight, the actual pixel of the pixel to be rotated and the adjacent pixel thereof in the rotated image block.
  • the values are weighted and summed, and the weighted pixel values are calculated as the pixel values of the rotated corresponding pixels.
  • the processor 101 is further configured to: determine whether an edge of the image block to be rotated is tilted with respect to a preset reference direction, and when the edge of the image block to be rotated is tilted with respect to a preset reference direction Detecting a tilt angle of the image block to be rotated, acquiring a first set of influence weights corresponding to the tilt angle from the vector model, according to actual pixel values of the pixel to be rotated and its neighboring pixels in the image block to be rotated, and the The first set of weight values inversely derives initial pixel values of the pixels to be rotated and their neighboring pixels in the image block to be rotated when the edge of the image block to be rotated is set along the reference direction, and then acquires the same by the vector model a second set of influence weights corresponding to a superposition angle of the tilt angle and the target rotation angle, and weighting and summing the initial pixel values of the pixels to be rotated and the adjacent pixels thereof calculated according to the second set of influence weights, and then
  • the processor 101 when calculating the initial pixel value of the pixel to be rotated and its neighboring pixels, is configured to inversely calculate the initial pixel of the pixel to be rotated and its adjacent pixel by backward estimation The values are weighted and summed by the first set of influence weights to obtain the actual pixel values of the pixels to be rotated.
  • the image processing apparatus 10 can more accurately calculate the pixel value of the rotated corresponding pixel, thereby avoiding Since the tilting causes the edge blurring problem after the rotation processing, it is advantageous to further improve the accuracy of subsequent image recognition.
  • the specific function implementation process of the processor 101 may refer to the methods provided in the first to fifth embodiments of the image processing method of the present invention, and is not repeated here.
  • the image processing device may be a server, a mobile terminal, or a circuit or chip integrated with the image processing function described above.
  • the image processing device may further include a memory, a display, and the like according to actual needs, and is not specifically limited herein.
  • the image processing device can obtain a correspondence between the weight of the pixel to be rotated and the adjacent pixel of the rotated pixel and the rotation angle during the image rotation, and utilize the communication.
  • the circuit obtains the image block to be rotated, the rotation center and the target rotation angle, and uses the vector model to determine the influence of the pixel to be rotated and its neighboring pixels in the image block to be rotated on the corresponding pixel after rotation according to the rotation center and the target rotation angle. Weighting, and calculating the pixel value of the rotated corresponding pixel in a pixel fusion manner according to the influence weight, so that the influence of the surrounding pixels can be merged into the rotated pixel without interpolation, so the information amount is not greatly reduced. , thereby reducing the impact on image recognition and improving the accuracy of image recognition.
  • the second embodiment 20 of the image processing apparatus of the present invention is similar in structure to the image processing apparatus 10 of the present invention, and the same portions are not described herein again, except that the image processing apparatus 20 of the present invention further
  • the camera assembly 103 is coupled to the processor 101 for capturing an image and transmitting the image to be rotated after the shooting to the processor 101 for the processor 101 to perform image processing.
  • the photographing component 103 may include components such as a camera, an encoding chip, and the like, which are specifically determined according to actual needs, and are not specifically limited herein.
  • the image processing device 20 further includes a memory 104 coupled to the processor 101 for storing data or instructions required by the processor 101 to execute an instruction, such as a vector model, to facilitate subsequent image processing by the processor 101.
  • a memory 104 coupled to the processor 101 for storing data or instructions required by the processor 101 to execute an instruction, such as a vector model, to facilitate subsequent image processing by the processor 101.
  • the image processing device may be a camera having an image processing function or an image processing device having an imaging function.
  • an embodiment 30 of the image processing system of the present invention includes: an image processing device 301 and a photographing device 302, wherein the image processing device 301 is connected to the photographing device 302 for performing an image to be rotated taken by the photographing device 302. deal with.
  • the structure of the image processing device 301 can refer to the first embodiment of the image processing device of the present invention, and is not repeated here.
  • the image processing device acquires the image to be rotated by the photographing device, and obtains a correspondence between the weight of the pixel to be rotated and the corresponding pixel of the adjacent pixel after the rotation of the image and the rotation angle by using a vector model.
  • the device 50 having a storage function internally stores a program 501, which is implemented to implement the first to fifth embodiments of the image processing method according to the present invention.
  • the method provided by any one of them and any combination of non-conflicting.
  • the device 50 having a storage function may be a portable storage medium such as a USB flash drive or an optical disk, or may be a base station or a separate component that can be integrated into the base station, such as a baseband board or the like.

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Image Processing (AREA)

Abstract

一种图像处理方法、设备及系统，该方法包括：获取一矢量模型，通过所述矢量模型能够获得在图像旋转过程中待旋转像素及其相邻像素对旋转后的对应像素的影响权重与旋转角度之间的对应关系(S11)；获取待旋转图像区块、旋转中心和目标旋转角度(S12)；利用所述矢量模型根据所述旋转中心以及目标旋转角度确定所述待旋转图像区块中的待旋转像素及其相邻像素对旋转后的对应像素的影响权重，并根据所述影响权重以像素融合方式计算旋转后的对应像素的像素值(S13)。通过上述方式，不会大幅度降低信息量，进而减少对图像识别的影响，提高图像识别的准确性。

Description

一种图像处理方法、设备及系统 【技术领域】
本发明涉及图像处理领域，尤其是涉及一种图像处理方法、设备及系统。
【背景技术】
现有的图像处理过程中，图像旋转过程中，需要经过图像重采样，即插值运算，这样容易降低图像的信息量，一般情况下不会对图像识别产生影响。但是，当图像分辨率很低时，由于图像旋转带来的信息量降低会对图像识别造成很大影响，大大降低图像识别的准确性。
【发明内容】
本发明主要解决的技术问题是提供一种图像处理方法、设备及系统，能够解决现有图像旋转过程中由于插值导致的信息量降低的问题。
为了解决上述问题，本发明采用的一个技术方案是：提供一种图像处理方法，包括：获取一矢量模型，通过矢量模型能够获得在图像旋转过程中待旋转像素及其相邻像素对旋转后的对应像素的影响权重与旋转角度之间的对应关系；获取待旋转图像区块、旋转中心和目标旋转角度；利用该矢量模型根据旋转中心以及目标旋转角度确定该待旋转图像区块中的待旋转像素及其相邻像素对旋转后的对应像素的影响权重，并根据该影响权重以像素融合方式计算旋转后的对应像素的像素值。
为了解决上述问题，本发明采用的另一个技术方案是：提供一种图像处理设备，包括：通信电路，用于获取待旋转图像区块、旋转中心和目标旋转角度；处理器，耦接该通信电路，用于获取一矢量模型，通过矢量模型能够获得在图像旋转过程中待旋转像素及其相邻像素对旋转后的对应像素的影响权重与旋转角度之间的对应关系；利用该矢量模型根据旋转中心以及目标旋转角度确定该待旋转图像区块中的待旋转像素及其相邻像素对旋转后的对应像素的影响权重， 并根据该影响权重以像素融合方式计算旋转后的对应像素的像素值。
为了解决上述问题，本发明采用的又一个技术方案是：提供一种图像处理系统，包括：如上所述的图像处理设备和拍摄设备，图像处理设备连接该摄设备，用于对拍摄设备拍摄的待旋转图像进行处理。
本发明的有益效果是：区别于现有技术的情况，本发明的部分实施例中，通过获取一矢量模型，能够获得在图像旋转过程中待旋转像素及其相邻像素对旋转后的对应像素的影响权重与旋转角度之间的对应关系，获取待旋转图像区块、旋转中心和目标旋转角度，利用该矢量模型根据旋转中心以及目标旋转角度确定待旋转图像区块中的待旋转像素及其相邻像素对旋转后的对应像素的影响权重，并根据该影响权重以像素融合方式计算旋转后的对应像素的像素值，从而可将周围像素的影响融合进旋转后的像素，而不需要进行插值运算，因此不会大幅度降低信息量，进而减少对图像识别的影响，提高图像识别的准确性。
【附图说明】
图1是本发明图像处理方法第一实施例的流程示意图；
图2是本发明图像处理方法第二实施例的流程示意图；
图3是本发明图像处理方法中矢量模型的示意图；
图4是本发明图像处理方法第三实施例的流程示意图；
图5是本发明图像处理方法第四实施例的流程示意图；
图6是本发明图像处理方法第五实施例的流程示意图；
图7是待旋转图像中存在倾斜设置的待旋转图像区块的场景示意图；
图8是本发明图像处理方法第五实施例中步骤S137的流程示意图；
图9是本发明图像处理设备第一实施例的结构示意图；
图10是本发明图像处理设备第二实施例的结构示意图；
图11是本发明图像处理系统一实施例的结构示意图；
图12是本发明具有存储功能的设备一实施例的结构示意图。
【具体实施方式】
下面结合附图和实施例对本发明进行详细说明。
如图1所示，本发明图像处理方法第一实施例包括：
S11：获取一矢量模型，通过矢量模型能够获得在图像旋转过程中待旋转像素及其相邻像素对旋转后的对应像素的影响权重与旋转角度之间的对应关系；
其中，该矢量模型可以包括预先保存的将待旋转图像以该待旋转图像的中心位置为旋转中心旋转各个角度后得到的像素位置模型，通过对该位置模型的分析，可以得到在图像旋转过程中待旋转像素及其相邻像素对旋转后的对应像素的影响权重与旋转角度之间的对应关系。
S12：获取待旋转图像区块、旋转中心和目标旋转角度；
其中，该待旋转图像区块可以是待旋转图像中的某些区域，例如包含待识别工件的区域，也可以是整个待旋转图像，此处不做具体限定。
该目标旋转角度可以是用户手动输入的旋转角度，也可以是预先设定的旋转角度，又或者是对待旋转图像进行识别后检测出的有利于观测的旋转角度，角度数值可以根据实际处理或观测需求而定，此处不做具体限定。
该旋转中心可以是待旋转图像区块的中心位置，也可以是其他位置，例如用户所选择的位置。例如，当用户不输入旋转中心时，默认以该待旋转图像的中心位置为旋转中心进行旋转。当用户指定旋转中心时，则以用户指定的位置作为旋转中心进行旋转。
S13：利用该矢量模型根据旋转中心以及目标旋转角度确定待旋转图像区块中的待旋转像素及其相邻像素对旋转后的对应像素的影响权重，并根据该影响权重以像素融合方式计算旋转后的对应像素的像素值。
其中，该待旋转像素及其相邻像素所对应的像素值可以是该待旋转像素及其相邻像素的原始像素值，也可以是待旋转像素及其相邻像素经过处理后获取的像素值；该像素融合方式可以是加权求和或者比例加权等计算方式，具体计 算方式可以视实际需求而定，此处不做具体限定。
具体地，在一个应用例中，当需要对图像以默认旋转中心(如待旋转图像的中心位置)进行旋转处理时，图像处理设备可以直接获取预先保存的矢量模型以及目标旋转角度，如用户输入的45度，即顺时针旋转45度，然后根据目标旋转角度，分析该矢量模型，可以得到待旋转图像区块(如整个待旋转图像)顺时针旋转45度后，旋转前的像素及其相邻像素在旋转后的对应像素中所占的比例，从而可以得到旋转前的像素及其相邻像素的像素值对旋转后的对应像素的像素值的影响权重。然后，图像处理设备可以利用像素融合方式(如加权求和)对旋转前的像素及其相邻像素的像素值结合对应的影响权重进行计算，最后得到旋转后的对应像素的像素值。该待旋转图像区块中的每个像素均可以采用上述方式计算得到旋转后的对应像素的像素值，从而可以得到旋转后的图像区块。
当需要对图像以用户指定的旋转中心进行旋转时，首先获取用户指定的作为旋转中心的位置，然后将该位置代入到该矢量模型的旋转中心，从而进一步利用该矢量模型根据目标旋转角度确定待旋转图像区块中的待旋转像素及其相邻像素对旋转后的对应像素的影响权重，并根据该影响权重以像素融合方式计算旋转后的对应像素的像素值。
在其他应用例中，可以根据实际需要对旋转前的像素及其相邻像素的像素值进行处理后，获得对应的像素值，然后再计算旋转后的对应像素的像素值，例如对边缘相对于像素排列行/列方向处于倾斜状态的待旋转图像区块，先反算该待旋转图像区块边缘平行于像素排列行/列方向时，该待旋转图像区块的待旋转像素及其相邻像素所对应的像素值，然后再以反算后的该对应像素值计算旋转后的对应像素的像素值。
在其他实施例中，该矢量模型也可以直接包括分析后得到的该对应关系，例如将旋转角度与不同位置的待旋转像素的影响权重之间的对应关系保存为一个对应关系表，以便于后续图像处理时直接在该对应关系表中找到指定位置的 像素在目标旋转角度对应的影响权重，从而提高处理效率。当然，该矢量模型也可以是根据获取的目标旋转角度以及目标像素的位置来实时计算得到的模型，此处不做具体限定。
本实施例中，通过获取一矢量模型，能够获得在图像旋转过程中待旋转像素及其相邻像素对旋转后的对应像素的影响权重与旋转角度之间的对应关系，获取待旋转图像区块、旋转中心和目标旋转角度，利用该矢量模型根据旋转中心以及目标旋转角度确定待旋转图像区块中的待旋转像素及其相邻像素对旋转后的对应像素的影响权重，并根据该影响权重以像素融合方式计算旋转后的对应像素的像素值，从而可将周围像素的影响融合进旋转后的像素，而不需要进行插值运算，因此不会大幅度降低信息量，进而减少对图像识别的影响，提高图像识别的准确性。
如图2所示，本发明图像处理方法第二实施例是在本发明图像处理方法第一实施例的基础上，步骤S11进一步包括：
S111：将以预定形状表示的待旋转像素及其相邻像素在按旋转角度进行旋转后与以该预定形状表示的对应像素的重叠面积之间的比例关系作为该影响权重，并将该影响权重与该旋转角度以及像素位置进行关联，以形成矢量模型。
其中，该预定形状可以是正方形，也可以是长方形，还可以是多边形等其他形状，具体可以根据实际需求而定，此处不做具体限定。该相邻像素是待旋转像素旋转前在该待旋转图像区块中的相邻像素。
具体地，结合图3所示，以正方形表示的待旋转像素及其相邻像素为例，其中，相邻像素包括在像素排列的行/列方向上，以及对角方向上与该待旋转像素相邻的像素。当需要对该待旋转图像区块进行逆时针45度旋转，如用户输入-45度时，该正方形表示的待旋转像素(以中心像素A为例)及其相邻像素B1～B8逆时针45度旋转后，与以正方形表示的对应像素C(图3中以虚线表示的正方形C)的重叠面积之间的比例关系分别为7/9、0、1/18、0、1/18、1/18、0、1/18和0，即如图3所示，对于中心像素A旋转后的对应像素C，只有像素A、像素 B2、B4、B5和B7与其重叠，从而以上述比例关系作为该待旋转像素A及其相邻像素B1～B8对旋转后的对应像素C的影响权重，根据上述方式，可以得到每个位置的待旋转像素及其相邻像素对旋转后的对应像素的影响权重，并将得到的影响权重与该旋转角度(例如：-45度/逆时针45度)以及像素的位置进行关联，最后得到该矢量模型。根据上述方式，可以得到每个旋转角度对应的矢量模型。
可选地，步骤S111之前，还可以包括：
S1101：根据待旋转图像的图像旋转中心和旋转角度，确定以预定形状表示的每个待旋转像素在以该图像旋转中心为中心旋转该旋转角度后的对应像素。其中，该图像旋转中心为待旋转图像的中心。
S1102：确定该待旋转图像中每个待旋转像素及其相邻像素在以该图像旋转中心为中心旋转该旋转角度后与该对应像素的重叠面积之间的比例关系。
具体地，结合图3所示，在一个应用例中，当需要对待旋转图像建立矢量模型时，首先根据旋转中心(待旋转图像的中心位置)和旋转角度(如逆时针旋转45度)，确定该待旋转图像中，每个待旋转像素旋转后的对应像素。以图3中待旋转像素B6为例，该像素B6逆时针旋转45度后，得到的对应像素是G(图3中的以虚线表示的正方形G)，然后，可以根据该像素B6和其相邻像素B4、A、B7与该对应像素G的重叠面积，像素分别计算B6、B4、A和B7与像素G的重叠面积占该像素G的总面积的比例，即可以得到待旋转像素B6及其相邻像素B4、A、B7在按旋转角度(逆时针旋转45度)进行旋转后与该对应像素G的重叠面积之间的比例关系。根据上述方式，可以得到每个待旋转像素旋转后的对应像素，以及每个待旋转像素及其相邻像素在按旋转角度进行旋转后与该对应像素的重叠面积之间的比例关系。
对于每个旋转角度和每个待旋转像素，均可以根据上述方式，得到每个待旋转像素根据每个旋转角度旋转后的对应像素，以及每个待旋转像素及其相邻像素在按旋转角度进行旋转后与该对应像素的重叠面积之间的比例关系。然后， 继续执行上述步骤S111，即可以得到每个旋转角度和每个待旋转像素对应的矢量模型。
本实施例中，可以用查询表方式或公式方式表示影响权重与旋转角度以及像素位置之间的对应关系，即将影响权重与旋转角度以及像素位置之间的对应关系保存在一个或多个查询表中，或者将影响权重与旋转角度以及像素位置之间的对应关系保存为一个或多个公式，在获取目标旋转角度以及像素位置后，采用查表或者公式计算的方式，即可以快速得到该影响权重，提高图像处理效率。
在其他实施例中，还可以根据实际需求设置该矢量模型，例如设置影响权重与某个角度范围的对应关系，此处不做具体限定。
如图4所示，本发明图像处理方法第三实施例是在本发明图像处理方法第一实施例的基础上，步骤S13进一步包括：
S131：利用以下公式(1)进行像素融合：
Figure PCTCN2017120229-appb-000001
其中，V mid为旋转后的对应像素的像素值，V m' id为待旋转像素所对应的像素值，a为待旋转像素所对应的影响权重，V i为待旋转像素的第i个相邻像素所对应的像素值，b i为第i个相邻像素所对应的影响权重，n为相邻像素的数量。
其中，相邻像素至少包括沿待旋转图像区块中的像素的行排列方向和列排列方向与待旋转像素相邻设置的四个相邻像素。
具体地，结合图3所示，在一个应用例中，相邻像素以沿待旋转图像区块中的像素的行排列方向和列排列方向与待旋转像素相邻设置的四个相邻像素为例，图像处理设备根据矢量模型获取待旋转像素A的像素值V m' id对旋转后的对应像素C的像素值V mid的影响权重a，及其相邻四个像素B2、B4、B5和B7的像素值V 1～V 4对旋转后的对应像素C的像素值V mid的影响权重b 1～b 4后，则可以根据上述公式，计算该旋转后的对应像素C的像素值V mid。其中，该影响权重b 1～b 4的值可以相同，也可以不同，具体视实际情况而定，例如当像素B2、B4、B5和 B7与像素C重叠的面积相等时，b 1～b 4的值相同。
当然，在其他实施例中，该相邻像素还可以包括像素排列的对角方向上与该待旋转像素相邻设置的相邻像素，如图3中的B1、B3、B6和B8。
本实施例还可以与本发明图像处理方法第二实施例相结合。
如图5所示，本发明图像处理方法第四实施例是在本发明图像处理方法第一实施例的基础上，步骤S13进一步包括：
S132：通过矢量模型获取与目标旋转角度以及像素位置对应的影响权重；
S133：根据该影响权重对待旋转图像区块中的待旋转像素及其相邻像素的实际像素值进行加权求和，进而将计算获得的加权像素值作为旋转后的对应像素的像素值。
具体地，在一个应用例中，当矢量模型采用用查询表方式表示影响权重与旋转角度以及像素位置之间的对应关系时，图像处理设备获取目标旋转角度以及像素位置后，可以根据待旋转像素及其相邻像素的位置在查询表中查找各像素在目标旋转角度下的影响权重，然后根据获取的影响权重对待旋转图像区块中待旋转像素及其相邻像素的实际像素值进行加权求和，最终可以计算出旋转后的对应像素的像素值。
例如，图3中的待旋转图像区块中，以待旋转像素A和其相邻的四个像素B2、B4、B5和B7为例，其中，利用矢量模型查表可以得到该目标旋转角度(如45度)对应的像素A对旋转后的像素C的影响权重为7/9，而B2、B4、B5和B7对旋转后的像素C的影响权重均为1/18，且像素A、B2、B4、B5和B7的实际像素值分别为128、220、160、100和40，则可以利用上述公式(1)进行加权求和：
Figure PCTCN2017120229-appb-000002
可以得到旋转后的像素C的像素值V mid。待旋转图像区块中的其他像素的像素值也可以采用上述公式(1)进行计算，最终可以得到整个旋转后的图像的像素值。具体计算过程中，其他像素的相邻像素还可以包括沿像素排列的对角方向相邻 设置的像素，各个相邻像素对应的影响权重也可以不同，具体视矢量模型中相邻像素与旋转后的对应像素的重叠面积比例关系而定，此处不做具体限定。
本实施例还可以与本发明图像处理方法第二或第三实施例或者其组合相结合。
在其他实施例中，当待旋转图像区块的边缘与像素排列的行/列方向存在夹角时，例如拍摄工件时，该工件的边缘与像素排列的行方向有夹角，即该工件在图像中是倾斜的，该夹角即为倾斜角度，此时，需要将该待旋转图像区块中的像素进行处理，反算该工件边缘与像素排列的行方向一致时待旋转像素的初始像素值，利用该初始像素值以及倾斜角度和目标旋转角度叠加后的角度，可以较准确地计算旋转后的对应像素的像素值，避免由于倾斜导致旋转处理后边缘模糊问题，有利于提高后续图像识别的精度。
具体如图6所示，本发明图像处理方法第五实施例是在本发明图像处理方法第一实施例的基础上，步骤S13进一步包括：
S134：判断待旋转图像区块的边缘是否相对于预设的参考方向倾斜设置；
其中，该参考方向为待旋转图像区块中的像素的行排列方向或列排列方向。
具体地，在一个应用例中，可以先对待旋转图像进行边缘检测，获取待旋转图像区块。其中，边缘检测方法可以采用连通域检测等方式检测连通的区域作为待旋转图像区块，该待旋转图像区块可以是拍摄的待检测工件或者待检测工件和检测台的图像，具体可以视实际需求设置，此处不做具体限定。
结合图7所示，对待旋转图像边缘检测后获取的待旋转图像区块包括D和E，此时，分别判断待旋转图像区块D和E的边缘是否相对于像素的行排列方向或列排列方向倾斜设置，例如分别判断区块D和E靠近图像底部的边缘
Figure PCTCN2017120229-appb-000003
和
Figure PCTCN2017120229-appb-000004
是否与图7所示的x轴或y轴方向一致，若不一致，即存在夹角，如图7所示，
Figure PCTCN2017120229-appb-000005
和
Figure PCTCN2017120229-appb-000006
与x轴方向均存在夹角α和β，则可以判定该区块D和E相对于预设的参考方向倾斜设置。
S135：若相对于预设的参考方向倾斜设置，则检测该待旋转图像区块的倾 斜角度；
具体地，结合图7所示，上述应用例中，判定该区块D和E相对于预设的参考方向倾斜设置后，可以分别检测该区块D和E的倾斜角度，即检测获取该夹角α和β的值，其中，可以利用边缘检测等方式获取该区块D和E的边缘
Figure PCTCN2017120229-appb-000007
和
Figure PCTCN2017120229-appb-000008
所在方向与x轴方向的夹角α和β的值，或者可以通过检测
Figure PCTCN2017120229-appb-000009
和
Figure PCTCN2017120229-appb-000010
的斜率，计算其与x轴方向的夹角α和β的值，具体检测方式可以根据实际需求而定。
S136：从矢量模型获取与该倾斜角度对应的第一组影响权重；
S137：根据该待旋转图像区块中的待旋转像素及其相邻像素的实际像素值以及第一组权重值反向推算出当该待旋转图像区块的边缘沿参考方向设置时该待旋转图像区块内的待旋转像素及其相邻像素的初始像素值；
具体地，在上述应用例中，获取该倾斜角度α和β后，可以利用矢量模型获取该倾斜角度α和β对应的第一组影响权重，具体通过查表或公式计算等方式获取，此处不做具体限定。然后，根据区块D中的待旋转像素及其相邻像素的实际像素值以及该倾斜角度α对应的第一组权重值，可以反向推算出当该待旋转图像区块D的边缘
Figure PCTCN2017120229-appb-000011
沿参考方向x轴方向设置时该待旋转图像区块D内的待旋转像素及其相邻像素的初始像素值。其中，可以利用上述步骤S131中的公式(1)估算出该初始像素值。区块E中的待旋转像素及其相邻像素的初始像素值也可以采用相同的方式计算。
可选地，如图8所示，步骤S137包括：
S1371：通过反向推算使得反向推算出的待旋转像素及其相邻像素的初始像素值经第一组影响权重进行加权求和后能够得到待旋转像素的实际像素值。
具体地，在上述应用例中，对于每个边缘倾斜设置的待旋转图像区块，如图7中的D，可以利用上述公式(1)，建立待旋转像素及其相邻像素的初始像素值和实际像素值之间的方程，然后同样可以建立其中一个相邻像素及其相邻像素的初始像素值和实际像素值之间的方程，同样还可以建立周边其他像素及 其相邻像素的初始像素值和实际像素值之间的方程，由此可以建立由多个方程组成的方程组，进而通过对该方程组中某个或某些像素的未知像素值(例如旋转中心像素的初始像素值)进行赋值的方式，可以估算出该方程组的解，从而可以比较准确的反向推算出待旋转像素及其相邻像素的初始像素值。
S138：通过矢量模型获取与该倾斜角度与目标旋转角度的叠加角度对应的第二组影响权重；
S139：根据第二组影响权重对反向推算出的待旋转像素及其相邻像素的初始像素值进行加权求和，进而将计算获得加权像素值作为旋转后的对应像素的像素值。
具体地，对于每个边缘倾斜设置的待旋转图像区块，如图7中的区块D，可以通过矢量模型查找到该倾斜角度α和目标旋转角度(如30度)的叠加角度α+30对应的第二组影响权重，从而可以利用上述公式(1)对反向推算出的待旋转像素及其相邻像素的初始像素值进行加权求和，最终可以得到旋转后的对应像素的像素值。类似的，对于图7中的区块E，也可以采用上述相同的方式进行计算。
可选地，在其他实施例中，该边缘倾斜的待旋转图像区块也可以采用本发明图像处理方法第一至第四实施例中任一实施例或其组合提供的方法直接进行计算，即以该目标旋转角度直接获取矢量模型后计算旋转后的像素值，不需要反算初始像素值，最终得到旋转后的图像。
本实施例中，通过将倾斜设置的待旋转图像区块中的像素进行处理，反算出该待旋转图像区块的边缘与像素排列的行方向一致时的初始像素值，并利用该初始像素值以及倾斜角度和目标旋转角度叠加后的角度，可以较准确地计算旋转后的对应像素的像素值，避免由于倾斜导致旋转处理后边缘模糊问题，有利于进一步提高后续图像识别的精度。
如图9所示，本发明图像处理设备第一实施例10包括：
通信电路102，用于获取待旋转图像区块、旋转中心和目标旋转角度；
处理器101，耦接该通信电路102，用于获取一矢量模型，通过该矢量模型能够获得在图像旋转过程中待旋转像素及其相邻像素对旋转后的对应像素的影响权重与旋转角度之间的对应关系，利用该矢量模型根据旋转中心以及目标旋转角度确定待旋转图像区块中的待旋转像素及其相邻像素对旋转后的对应像素的影响权重，并根据该影响权重以像素融合方式计算旋转后的对应像素的像素值。
其中，该处理器101也可以是通过该通信电路102获取该矢量模型，该矢量模型可以是保存在该图像处理设备10的内部，也可以是保存在与该图像处理设备10连接的外部设备(图未示)中，此处不做具体限定。
可选地，该处理器101进一步用于：在获取该矢量模型时，将以预定形状表示的待旋转像素及其相邻像素在按旋转角度进行旋转后与以预定形状表示的对应像素的重叠面积之间的比例关系作为影响权重，并将影响权重与旋转角度以及像素位置进行关联，最终形成矢量模型。
其中，在所述矢量模型中，可以采用查询表方式或公式方式表示影响权重与旋转角度之间的对应关系。
可选地，该处理器101进一步用于：在将以预定形状表示的待旋转像素及其相邻像素在按旋转角度进行旋转后与以预定形状表示的对应像素的重叠面积之间的比例关系作为影响权重，并将所述影响权重与旋转角度以及像素位置进行关联，以形成矢量模型之前，根据图像旋转中心和旋转角度，确定以预定形状表示的每个待旋转像素在以该图像旋转中心为中心旋转该旋转角度后的对应像素，并确定每个待旋转像素及其相邻像素在以该图像旋转中心为中心旋转该旋转角度后与对应像素的重叠面积之间的比例关系。
可选地，该处理器101进一步用于：在计算旋转后的对应像素的像素值时，通过上述公式(1)进行像素融合。
其中，相邻像素至少包括沿待旋转图像区块中的像素的行排列方向和列排列方向与待旋转像素相邻设置的四个相邻像素。
可选地，该处理器101进一步用于：通过矢量模型获取与目标旋转角度以及像素位置对应的影响权重，根据该影响权重对待旋转图像区块中的待旋转像素及其相邻像素的实际像素值进行加权求和，进而将计算获得加权像素值作为旋转后的对应像素的像素值。
可选地，该处理器101还用于：判断待旋转图像区块的边缘是否相对于预设的参考方向倾斜设置，并在待旋转图像区块的边缘相对于预设的参考方向倾斜设置时，检测该待旋转图像区块的倾斜角度，从矢量模型获取与该倾斜角度对应的第一组影响权重，根据待旋转图像区块中的待旋转像素及其相邻像素的实际像素值以及该第一组权重值反向推算出当待旋转图像区块的边缘沿参考方向设置时待旋转图像区块内的待旋转像素及其相邻像素的初始像素值，然后，通过矢量模型获取与该倾斜角度与目标旋转角度的叠加角度对应的第二组影响权重，根据该第二组影响权重对反向推算出的待旋转像素及其相邻像素的初始像素值进行加权求和，进而将计算获得加权像素值作为旋转后的对应像素的像素值。
其中，可选地，在计算该待旋转像素及其相邻像素的初始像素值时，该处理器101用于通过反向推算使得反向推算出的待旋转像素及其相邻像素的初始像素值经该第一组影响权重进行加权求和后能够得到待旋转像素的实际像素值。
由此，当该待旋转图像区块相对于预设的参考方向，例如像素排列的行/列方向倾斜设置时，该图像处理设备10可以较准确地计算旋转后的对应像素的像素值，避免由于倾斜导致旋转处理后边缘模糊问题，有利于进一步提高后续图像识别的精度。
本实施例中，该处理器101的具体功能实现过程可以参考本发明图像处理方法第一至第五实施例所提供的方法，此处不再重复。
本实施例中，该图像处理设备可以是服务器、移动终端或者集成有上述图像处理功能的电路或芯片等。
在其他实施例中，该图像处理设备还可以视实际需求包括存储器、显示器 等，此处不做具体限定。
本实施例中，图像处理设备通过获取一矢量模型，能够获得在图像旋转过程中待旋转像素及其相邻像素对旋转后的对应像素的影响权重与旋转角度之间的对应关系，并利用通信电路获取待旋转图像区块、旋转中心和目标旋转角度，利用该矢量模型根据旋转中心以及目标旋转角度确定待旋转图像区块中的待旋转像素及其相邻像素对旋转后的对应像素的影响权重，并根据该影响权重以像素融合方式计算旋转后的对应像素的像素值，从而可将周围像素的影响融合进旋转后的像素，而不需要进行插值运算，因此不会大幅度降低信息量，进而减少对图像识别的影响，提高图像识别的准确性。
具体地，如图10所示，本发明图像处理设备第二实施例20与本发明图像处理设备10结构类似，相同之处此处不再赘述，不同之处在于，本发明图像处理设备20还包括：拍摄组件103，耦接该处理器101，用于拍摄图像，并将拍摄后的待旋转图像传输给处理器101，以便于处理器101进行图像处理。
其中，该拍摄组件103可以包括摄像头、编码芯片等部件，具体视实际需求而定，此处不做具体限定。
可选地，图像处理设备20还包括：存储器104，耦接该处理器101，用于存储处理器101执行指令所需要的数据或指令，例如矢量模型，以便于处理器101后续进行图像处理。
本实施例中，图像处理设备可以是具有图像处理功能的摄像机或者具有摄像功能的图像处理设备。
如图11所示，本发明图像处理系统一实施例30包括：图像处理设备301和拍摄设备302，其中该图像处理设备301连接该拍摄设备302，用于对拍摄设备302拍摄的待旋转图像进行处理。
其中，该图像处理设备301的结构可以参考本发明图像处理设备第一实施例，此处不再重复。
本实施例中，图像处理设备通过拍摄设备获取待旋转图像，利用一矢量模 型获得在图像旋转过程中待旋转像素及其相邻像素对旋转后的对应像素的影响权重与旋转角度之间的对应关系，并获取待旋转图像区块、旋转中心和目标旋转角度，利用该矢量模型根据旋转中心以及目标旋转角度确定待旋转图像区块中的待旋转像素及其相邻像素对旋转后的对应像素的影响权重，并根据该影响权重以像素融合方式计算旋转后的对应像素的像素值，从而可将周围像素的影响融合进旋转后的像素，而不需要进行插值运算，因此不会大幅度降低信息量，进而减少对图像识别的影响，提高图像识别的准确性。
如图12所示，本发明具有存储功能的设备一实施例中，具有存储功能的设备50内部存储有程序501，该程序501被执行时实现如本发明图像处理方法第一至第五实施例中任一个以及任意不冲突的组合所提供的方法。
其中，具有存储功能的设备50可以是便携式存储介质如U盘、光盘，也可以是基站或可集成于基站中的独立部件，例如基带板等。
本实施例中，具有存储功能的设备中存储的程序被执行时，通过获取一矢量模型，能够获得在图像旋转过程中待旋转像素及其相邻像素对旋转后的对应像素的影响权重与旋转角度之间的对应关系，获取待旋转图像区块、旋转中心和目标旋转角度，利用该矢量模型根据旋转中心以及目标旋转角度确定待旋转图像区块中的待旋转像素及其相邻像素对旋转后的对应像素的影响权重，并根据该影响权重以像素融合方式计算旋转后的对应像素的像素值，从而可将周围像素的影响融合进旋转后的像素，而不需要进行插值运算，因此不会大幅度降低信息量，进而减少对图像识别的影响，提高图像识别的准确性。
以上所述仅为本发明的实施方式，并非因此限制本发明的专利范围，凡是利用本发明说明书及附图内容所作的等效结构或等效流程变换，或直接或间接运用在其他相关的技术领域，均同理包括在本发明的专利保护范围内。

Claims (20)

  1. 一种图像处理方法，其特征在于，包括：
    获取一矢量模型，通过所述矢量模型能够获得在图像旋转过程中待旋转像素及其相邻像素对旋转后的对应像素的影响权重与旋转角度之间的对应关系；
    获取待旋转图像区块、旋转中心和目标旋转角度；
    利用所述矢量模型根据所述旋转中心以及目标旋转角度确定所述待旋转图像区块中的待旋转像素及其相邻像素对旋转后的对应像素的影响权重，并根据所述影响权重以像素融合方式计算旋转后的对应像素的像素值。
  2. 根据权利要求1所述的方法，其特征在于，所述获取一矢量模型的步骤包括：
    将以预定形状表示的待旋转像素及其相邻像素在按所述旋转角度进行旋转后与以所述预定形状表示的所述对应像素的重叠面积之间的比例关系作为所述影响权重，并将所述影响权重与所述旋转角度以及像素位置进行关联，以形成所述矢量模型。
  3. 根据权利要求2所述的方法，其特征在于，所述将以预定形状表示的待旋转像素及其相邻像素在按所述旋转角度进行旋转后与以所述预定形状表示的所述对应像素的重叠面积之间的比例关系作为所述影响权重，并将所述影响权重与所述旋转角度以及像素位置进行关联，以形成所述矢量模型之前，包括：
    根据图像旋转中心和所述旋转角度，确定以所述预定形状表示的每个所述待旋转像素在以所述图像旋转中心为中心旋转所述旋转角度后的所述对应像素；
    确定每个所述待旋转像素及其相邻像素在以所述图像旋转中心为中心旋转所述旋转角度后与所述应像素的重叠面积之间的比例关系。
  4. 根据权利要求2所述的方法，其特征在于，在所述矢量模型中，以查询表方式或公式方式表示所述影响权重与所述旋转角度以及所述像素位置之间的对应关系。
  5. 根据权利要求1所述的方法，其特征在于，所述利用所述矢量模型根据所述旋转中心以及目标旋转角度确定所述待旋转图像区块中的待旋转像素及其相邻像素对旋转后的对应像素的影响权重，并根据所述影响权重以像素融合方式计算旋转后的对应像素的像素值的步骤包括：
    通过以下公式进行像素融合：
    Figure PCTCN2017120229-appb-100001
    其中，V mid为旋转后的对应像素的像素值，V′ mid为所述待旋转像素所对应的像素值，a为所述待旋转像素所对应的影响权重，V i为所述待旋转像素的第i个相邻像素所对应的像素值，b i为第i个相邻像素所对应的影响权重，n为所述相邻像素的数量。
  6. 根据权利要求5所述的方法，其特征在于，所述相邻像素至少包括沿所述待旋转图像区块中的像素的行排列方向和列排列方向与所述待旋转像素相邻设置的四个相邻像素。
  7. 根据权利要求1所述的方法，其特征在于，所述利用所述矢量模型根据所述旋转中心以及目标旋转角度确定所述待旋转图像区块中的待旋转像素及其相邻像素对旋转后的对应像素的影响权重，并根据所述影响权重以像素融合方式计算旋转后的对应像素的像素值的步骤包括：
    通过所述矢量模型获取与所述目标旋转角度以及像素位置对应的影响权重；
    根据所述影响权重对所述待旋转图像区块中的待旋转像素及其相邻像素的实际像素值进行加权求和，进而将计算获得的加权像素值作为旋转后的对应像素的像素值。
  8. 根据权利要求1所述的方法，其特征在于，所述利用所述矢量模型根据所述旋转中心以及目标旋转角度确定所述待旋转图像区块中的待旋转像素及其相邻像素对旋转后的对应像素的影响权重，并根据所述影响权重以像素融合方式计算旋转后的对应像素的像素值的步骤包括：
    判断所述待旋转图像区块的边缘是否相对于预设的参考方向倾斜设置；
    若相对于预设的参考方向倾斜设置，则检测所述待旋转图像区块的倾斜角度；
    从所述矢量模型获取与所述倾斜角度对应的第一组影响权重；
    根据所述待旋转图像区块中的待旋转像素及其相邻像素的实际像素值以及所述第一组权重值反向推算出当所述待旋转图像区块的边缘沿所述参考方向设置时所述待旋转图像区块内的待旋转像素及其相邻像素的初始像素值；
    通过所述矢量模型获取与所述倾斜角度与所述目标旋转角度的叠加角度对应的第二组影响权重；
    根据所述第二组影响权重对反向推算出的待旋转像素及其相邻像素的初始像素值进行加权求和，进而将计算获得加权像素值作为旋转后的对应像素的像素值。
  9. 根据权利要求8所述的方法，其特征在于，所述根据所述待旋转图像区块中的待旋转像素及其相邻像素的实际像素值以及所述第一组权重值反向推算出当所述待旋转图像区块的边缘沿所述参考方向设置时所述待旋转图像区块内的待旋转像素及其相邻像素的初始像素值的步骤包括：
    通过反向推算使得反向推算出的待旋转像素及其相邻像素的初始像素值经所述第一组影响权重进行加权求和后能够得到所述待旋转像素的实际像素值。
  10. 根据权利要求8所述的方法，其特征在于，所述参考方向为所述待旋转图像区块中的像素的行排列方向或列排列方向。
  11. 一种图像处理设备，其特征在于，包括：
    通信电路，用于获取待旋转图像区块、旋转中心和目标旋转角度；
    处理器，耦接所述通信电路，用于获取一矢量模型，通过所述矢量模型能够获得在图像旋转过程中待旋转像素及其相邻像素对旋转后的对应像素的影响权重与旋转角度之间的对应关系；利用所述矢量模型根据所述旋转中心以及目标旋转角度确定所述待旋转图像区块中的待旋转像素及其相邻像素对旋转后的对应像素的影响权重，并根据所述影响权重以像素融合方式计算旋转后的对应 像素的像素值。
  12. 根据权利要求11所述的设备，其特征在于，所述处理器进一步用于：将以预定形状表示的待旋转像素及其相邻像素在按所述旋转角度进行旋转后与以所述预定形状表示的所述对应像素的重叠面积之间的比例关系作为所述影响权重，并将所述影响权重与所述旋转角度以及像素位置进行关联，以形成所述矢量模型。
  13. 根据权利要求12所述的设备，其特征在于，所述处理器进一步用于：在所述矢量模型中，以查询表方式或公式方式表示所述影响权重与所述旋转角度以及所述像素位置之间的对应关系。
  14. 根据权利要求12所述的设备，其特征在于，所述处理器进一步用于：在将以预定形状表示的待旋转像素及其相邻像素在按所述旋转角度进行旋转后与以所述预定形状表示的所述对应像素的重叠面积之间的比例关系作为所述影响权重，并将所述影响权重与所述旋转角度以及像素位置进行关联，以形成所述矢量模型之前，根据图像旋转中心和所述旋转角度，确定以所述预定形状表示的每个所述待旋转像素在以所述图像旋转中心为中心旋转所述旋转角度后的所述对应像素，并确定每个所述待旋转像素及其相邻像素在以所述图像旋转中心为中心旋转所述旋转角度后与所述应像素的重叠面积之间的比例关系。
  15. 根据权利要求11所述的设备，其特征在于，所述处理器进一步用于：
    通过以下公式进行像素融合：
    Figure PCTCN2017120229-appb-100002
    其中，V mid为旋转后的对应像素的像素值，V′ mid为所述待旋转像素所对应的像素值，a为所述待旋转像素所对应的影响权重，V i为所述待旋转像素的第i个相邻像素所对应的像素值，b i为第i个相邻像素所对应的影响权重，n为所述相邻像素的数量。
  16. 根据权利要求15所述的设备，其特征在于，所述相邻像素至少包括沿所述待旋转图像区块中的像素的行排列方向和列排列方向与所述待旋转像素相 邻设置的四个相邻像素。
  17. 根据权利要求11所述的设备，其特征在于，所述处理器进一步用于：通过所述矢量模型获取与所述目标旋转角度以及像素位置对应的影响权重；根据所述影响权重对所述待旋转图像区块中的待旋转像素及其相邻像素的实际像素值进行加权求和，进而将计算获得加权像素值作为旋转后的对应像素的像素值。
  18. 根据权利要求11所述的设备，其特征在于，所述处理器还用于：
    判断所述待旋转图像区块的边缘是否相对于预设的参考方向倾斜设置，并在所述待旋转图像区块的边缘相对于预设的参考方向倾斜设置时，检测所述待旋转图像区块的倾斜角度；
    从所述矢量模型获取与所述倾斜角度对应的第一组影响权重；
    根据所述待旋转图像区块中的待旋转像素及其相邻像素的实际像素值以及所述第一组权重值反向推算出当所述待旋转图像区块的边缘沿所述参考方向设置时所述待旋转图像区块内的待旋转像素及其相邻像素的初始像素值；
    通过所述矢量模型获取与所述倾斜角度与所述目标旋转角度的叠加角度对应的第二组影响权重；
    根据所述第二组影响权重对反向推算出的待旋转像素及其相邻像素的初始像素值进行加权求和，进而将计算获得加权像素值作为旋转后的对应像素的像素值。
  19. 根据权利要求18所述的设备，其特征在于，所述处理器还用于：通过反向推算使得反向推算出的待旋转像素及其相邻像素的初始像素值经所述第一组影响权重进行加权求和后能够得到所述待旋转像素的实际像素值。
  20. 一种图像处理系统，其特征在于，包括：如权利要求11-19任一项所述的图像处理设备和拍摄设备，所述图像处理设备连接所述拍摄设备，用于对所述拍摄设备拍摄的待旋转图像进行处理。
PCT/CN2017/120229 2017-12-29 2017-12-29 一种图像处理方法、设备及系统 WO2019127503A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2017/120229 WO2019127503A1 (zh) 2017-12-29 2017-12-29 一种图像处理方法、设备及系统
CN201780028397.7A CN109496326B (zh) 2017-12-29 2017-12-29 一种图像处理方法、设备及系统

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2017/120229 WO2019127503A1 (zh) 2017-12-29 2017-12-29 一种图像处理方法、设备及系统

Publications (1)

Publication Number Publication Date
WO2019127503A1 true WO2019127503A1 (zh) 2019-07-04

Family

ID=65687635

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/120229 WO2019127503A1 (zh) 2017-12-29 2017-12-29 一种图像处理方法、设备及系统

Country Status (2)

Country Link
CN (1) CN109496326B (zh)
WO (1) WO2019127503A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116433515A (zh) * 2023-03-30 2023-07-14 格兰菲智能科技有限公司 图像旋转抗锯齿方法、装置、计算机设备和存储介质

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110223245B (zh) * 2019-05-16 2021-07-16 华南理工大学 基于深度神经网络的模糊图片清晰化处理方法及系统
CN110807765B (zh) * 2019-09-27 2023-05-02 许昌许继软件技术有限公司 一种基于图像处理的悬垂绝缘子串倾斜检测方法与系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1755707A (zh) * 2004-09-30 2006-04-05 德鑫科技股份有限公司 一种倾斜图像自动校正方法
CN103325099A (zh) * 2013-07-11 2013-09-25 北京智诺英特科技有限公司 图像校正方法和装置
US8606030B2 (en) * 2010-07-05 2013-12-10 Hon Hai Precision Industry Co., Ltd. Image correction device and method
CN104537609A (zh) * 2014-11-28 2015-04-22 上海理工大学 一种旋转图像校正方法
CN104809703A (zh) * 2015-04-22 2015-07-29 上海理工大学 一种简单的图像角度校正方法

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP4378060B2 (ja) * 2001-02-22 2009-12-02 キヤノンマーケティングジャパン株式会社 画像処理装置及びその方法並びに記憶媒体
CN100487522C (zh) * 2005-01-13 2009-05-13 奥林巴斯映像株式会社 模糊校正方法及摄像装置
JP4947314B2 (ja) * 2007-10-26 2012-06-06 村田機械株式会社 画像処理装置、画像読取装置、画像処理方法及び画像処理プログラム
CN105023260A (zh) * 2014-04-22 2015-11-04 Tcl集团股份有限公司 一种全景图像融合方法及融合装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1755707A (zh) * 2004-09-30 2006-04-05 德鑫科技股份有限公司 一种倾斜图像自动校正方法
US8606030B2 (en) * 2010-07-05 2013-12-10 Hon Hai Precision Industry Co., Ltd. Image correction device and method
CN103325099A (zh) * 2013-07-11 2013-09-25 北京智诺英特科技有限公司 图像校正方法和装置
CN104537609A (zh) * 2014-11-28 2015-04-22 上海理工大学 一种旋转图像校正方法
CN104809703A (zh) * 2015-04-22 2015-07-29 上海理工大学 一种简单的图像角度校正方法

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116433515A (zh) * 2023-03-30 2023-07-14 格兰菲智能科技有限公司 图像旋转抗锯齿方法、装置、计算机设备和存储介质

Also Published As

Publication number Publication date
CN109496326A (zh) 2019-03-19
CN109496326B (zh) 2023-06-02

Similar Documents

Publication Publication Date Title
CN111862224B (zh) 确定相机与激光雷达之间外参的方法和装置
WO2021196548A1 (zh) 距离确定方法、装置及系统
WO2017076106A1 (zh) 图像的拼接方法和装置
US8842906B2 (en) Body measurement
JP2019079553A (ja) ビジョンシステムでラインを検出するためのシステム及び方法
KR102612530B1 (ko) 장면의 초해상도 이미지를 생성하기 위한 광학 이미지 안정화 움직임
CN108629810B (zh) 双目相机的标定方法、装置及终端
WO2019127503A1 (zh) 一种图像处理方法、设备及系统
CN107516322B (zh) 一种基于对数极空间的图像物体大小和旋转估计计算方法
US8531505B2 (en) Imaging parameter acquisition apparatus, imaging parameter acquisition method and storage medium
CN110278408A (zh) 图像处理装置、图像处理方法和存储介质
CN106469435B (zh) 图像处理方法、装置及设备
KR20130090615A (ko) 이미지 자동 보정을 위한 구도 기반 노출 측정 방법 및 장치
CN111383252A (zh) 多相机目标追踪方法、系统、装置及存储介质
CN109661815A (zh) 存在相机阵列的显著强度变化的情况下的鲁棒视差估计
CN109598674B (zh) 一种图像拼接方法及装置
WO2013149866A2 (en) Method and device for transforming an image
CN113407030B (zh) 视觉定位方法及相关装置、设备和存储介质
CN111192308A (zh) 图像处理方法及装置、电子设备和计算机存储介质
KR101348681B1 (ko) 영상탐지시스템의 다중센서 영상정렬방법 및 이를 이용한 영상탐지시스템의 다중센서 영상정렬장치
JP7192526B2 (ja) 画像処理装置、画像処理方法及びプログラム
CN108432229B (zh) 一种通过自拍实现他人拍照效果的方法及拍照设备
CN109961393A (zh) 基于插值和迭代优化算法的亚像素配准与拼接技术
CN113506343B (zh) 基于多源数据的色坐标估计方法、系统、设备及介质
JP2003091726A (ja) 反射パラメータ取得装置、反射成分分離装置、反射パラメータ取得用プログラム、及び反射成分分離用プログラム

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: 17936016

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: 17936016

Country of ref document: EP

Kind code of ref document: A1