WO2020118896A1 - 一种图像修复方法、图像修复系统及平板探测器 - Google Patents

一种图像修复方法、图像修复系统及平板探测器 Download PDF

Info

Publication number
WO2020118896A1
WO2020118896A1 PCT/CN2019/075643 CN2019075643W WO2020118896A1 WO 2020118896 A1 WO2020118896 A1 WO 2020118896A1 CN 2019075643 W CN2019075643 W CN 2019075643W WO 2020118896 A1 WO2020118896 A1 WO 2020118896A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
repair
defect
pixel
defects
Prior art date
Application number
PCT/CN2019/075643
Other languages
English (en)
French (fr)
Inventor
翟永立
张楠
方志强
Original Assignee
上海奕瑞光电子科技股份有限公司
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 上海奕瑞光电子科技股份有限公司 filed Critical 上海奕瑞光电子科技股份有限公司
Priority to US16/757,775 priority Critical patent/US11461877B2/en
Publication of WO2020118896A1 publication Critical patent/WO2020118896A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/77—Retouching; Inpainting; Scratch removal
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/211—Selection of the most significant subset of features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10116—X-ray image

Definitions

  • the invention relates to the field of X-ray detector imaging and image repair, in particular to an image repair method, an image repair system and a flat panel detector.
  • defective pixels refers to a detector unit that does not respond to X-ray intensity or responds poorly.
  • the main causes of defective pixels are the defects of the scintillator layer itself, the defects of the photodiode and thin film transistor itself, the damage of the driving circuit itself or the poor splicing of the detector plate, etc.
  • defective pixels can be divided into four types: dead pixels, under-responsive pixels, over-responsive pixels and chaotic pixels.
  • defective pixels in the connected domain of the image can be divided into three types: isolated point defects, cluster defects and bad line defects. These defective pixels exhibit different shapes when the detector is imaged, which not only reduces the yield of the detector when it is shipped from the factory, but also affects the judgment of the real situation of the object being photographed.
  • an image processing method used in the prior art mainly uses spline interpolation, bilinear interpolation, and three-spline interpolation to repair defective pixels. These repair methods consider defective pixel neighbors.
  • the gray value of normal pixels in the domain can restore the details of the image to a certain extent, but it will cause blur in the details of the edge class, and does not consider the details such as the texture of the image.
  • the embodiments of the present invention provide an image repair method, an image repair system, and a flat panel detector, so as to effectively repair the image output by the detector, and to protect the image details well.
  • an embodiment of the present invention provides an image restoration method, which includes the following steps:
  • the defective pixels are divided into three categories: isolated point defects, cluster defects and bad line defects;
  • the second image is sequentially repaired with isolated point defects, cluster defects, and bad line defects in the order of the defective pixels from small to large.
  • the isolated point defect repair is performed by using the method of neighborhood weighted average, and the grayscale of the repaired pixel is as follows:
  • F is the original image data in the isolated pixel defect 3 ⁇ 3 pixel neighborhood
  • N is the isolated pixel defect 3 ⁇ 3 pixel neighborhood
  • K is the weighted average coefficient corresponding to the normal pixel in the isolated pixel defect 3 ⁇ 3 pixel neighborhood .
  • the source of the weighted average includes:
  • the gray gradient of the normal pixels is calculated, wherein the normal pixel with a large gradient value corresponds to a high weighting coefficient ratio and a small gradient value
  • the weighting coefficients corresponding to the normal pixels are correspondingly smaller;
  • a method of combining level set and template matching is used to repair the cluster defect, wherein the level set method determines the repair order of the defective points of the cluster defect; the template matching method needs to be repaired To repair the current dead pixels.
  • determining the repair order of the defective points of the cluster defects includes the following steps:
  • Extract the cluster defect and determine the defect boundary of the cluster defect initialize the defect mark F and the arrival time T, and mark the normal pixel points, edge points and defect points of the defect boundary;
  • the bad point in the neighborhood of 3 ⁇ 3 pixels of the edge point with the smallest T value is determined, and the bad point in the neighborhood of 3 ⁇ 3 pixels of the edge point with the smallest T value is the current bad point that needs to be repaired.
  • repairing the current bad spot includes the following steps:
  • search area sequentially select search templates with the same size as the target template, calculate the similarity between the target template and the search template, and store the calculation results;
  • the repair step includes the steps of determining the repair order of the defective points of the cluster defect and repairing the determined defective points;
  • the repairing step further includes: updating the T value and F value of the repaired bad point after repairing the determined bad point;
  • the method of template matching is used to repair the bad line defect, and the repair is performed according to the order in which the bad points in the bad line defect are stored in the connected domain analysis.
  • the search range avoids the area where the bad line defect is located, and the search template adaptively changes according to the width of the bad line defect.
  • the method further includes the steps of performing background correction and gain correction on the output second image before repairing the output second image.
  • an embodiment of the present invention provides an image repair system, including:
  • the image acquisition module is used to acquire the first image
  • An image analysis module configured to perform connected domain analysis on the first image acquired by the image acquisition module, extract all defective pixels in the first image, and analyze the defects according to the size and shape of the connected domain Pixels are divided into three categories: isolated point defects, cluster defects and bad line defects;
  • the image repair module is configured to repair the second image according to the type of the defective pixel determined by the image analysis module.
  • the image repair system further includes an image acquisition module, which performs signal reading, amplification and analog-to-digital conversion on the output image, and transmits the amplified and converted image signal to the image repair module.
  • an image acquisition module which performs signal reading, amplification and analog-to-digital conversion on the output image, and transmits the amplified and converted image signal to the image repair module.
  • the image repair system further includes an image preprocessing module, the image preprocessing module includes a background correction module and a gain correction module, the background correction module performs background correction on the output image, and the gain correction module For performing gain correction on the output image.
  • the image preprocessing module includes a background correction module and a gain correction module
  • the background correction module performs background correction on the output image
  • the gain correction module For performing gain correction on the output image.
  • an embodiment of the present invention provides a flat panel detector including an image display module and the image repair system according to the second aspect of the present invention, the image repair system integrated in the flat panel detector
  • the image display module in the device outputs and displays the image repaired by the image repair system.
  • the image repair method, image repair system and flat panel detector provided by the present invention have the following beneficial effects:
  • the method of the present invention performs statistics and analysis of defective pixels on the first image, such as the defective template in the detector, sorts the defective pixels, and repairs the second image in the order of the defective pixels from small to large, and Different methods are used to repair different defective pixels.
  • This repair method takes into account the similarity of neighboring pixels and conforms to the distribution characteristics of the image grayscale, which can not only effectively repair the defective pixels, but also retain the details of the original image well.
  • the method of the present invention performs pixel statistics and classification on the defect template in the detector, and repairs the image output by the detector in real time according to the classification, which can realize rapid repair of the defective pixel of the image.
  • the image repair method described in the present invention is only related to the type and size of defective pixels, such as defective dot defects, cluster defects, and defective line defects. It can be applied to flat panel detectors, so it can significantly improve work efficiency and reduce labor costs.
  • the image repair system described in the present invention can be integrated into a flat panel detector without requiring changes to the hardware design, which can reduce product costs and increase the yield of shipments. And in the subsequent upgrade and maintenance, you only need to optimize and upgrade the software.
  • FIG. 1 shows a flowchart of an image restoration method according to Embodiment 1 of the present invention.
  • FIG. 2 shows a flowchart of the level set method in the method shown in Embodiment 3 of the present invention.
  • FIG. 3 shows a flowchart of a template matching method in the method shown in Embodiment 3 of the present invention.
  • FIG. 4 shows a flowchart of a repair method according to a preferred embodiment of Embodiment 3 of the present invention.
  • FIG. 5 shows a flowchart of a repair method according to another preferred embodiment of Embodiment 3 of the present invention.
  • 6a and 6b respectively show a defective image with cluster defects and an image repaired according to the method of Embodiment 3.
  • FIG. 8 is a schematic diagram of an image processing system provided by Embodiment 5 of the present invention.
  • Embodiment 9 is a schematic diagram of a flat panel detector provided in Embodiment 6 of the present invention.
  • This embodiment provides an image repair method. As shown in FIG. 1, the method includes the following steps:
  • the defective pixels are divided into three categories: isolated point defects, cluster defects and bad line defects;
  • the first image is preferably a defect template in an instrument such as a detector.
  • the first image such as the above defect template, may be subjected to connected domain analysis in order from left to right and from top to bottom.
  • the second image is sequentially repaired for isolated point defects, cluster defects, and bad line defects according to the size of the defective pixels, from small to large. Image repair in this order from small to large can achieve good repair results.
  • background correction and gain correction are first performed on the second image.
  • the method of this embodiment performs pixel statistics and classification on the defect template in the first image, such as a detector, and repairs the second image, such as the real-time output of the detector, according to the classification, which can quickly repair the defective pixels of the image.
  • this embodiment provides a method for repairing isolated point defects.
  • the field is adopted Weighted average method is used to repair outliers.
  • the above-mentioned weighted average method of neighborhood is used to implement the repair of isolated point defects.
  • the grayscale of isolated points repaired by this method is shown in the following formula (1):
  • F is the original image data in the isolated pixel defect 3 ⁇ 3 pixel neighborhood
  • N is the isolated pixel defect 3 ⁇ 3 pixel neighborhood
  • K is the weighted average coefficient corresponding to the normal pixel in the isolated pixel defect 3 ⁇ 3 pixel neighborhood .
  • the sources of the weighted average include the following two:
  • the gray distribution of all normal pixels in the 3 ⁇ 3 pixel neighborhood, the gray gradient of the normal pixels is calculated, wherein the normal pixel with a large gradient value corresponds to a high proportion of weighting coefficients and the gradient with a small gradient value
  • the weighting coefficients corresponding to the normal pixels are correspondingly smaller;
  • the gray distribution and position distribution of normal pixels in the neighborhood of the defective pixel are fully considered, and the weighted average pair corresponding to the normal pixel is used Repair isolated points.
  • This method can effectively repair the defects of isolated points, and can restore the details of the original image well.
  • this embodiment provides a method for repairing cluster defects.
  • a method of combining level sets and template matching is used to repair the cluster defects.
  • the level set method determines the repair order of the defective points of the cluster defects; the template matching method repairs the current defective points determined to be repaired.
  • using the level set method to determine the repair order of the defective points of the cluster defects includes the following steps: extracting the cluster defects and determining the cluster defects The defect boundary of F, initialize the defect mark F and the arrival time T, and mark the normal pixels, edge points and defect points of the defect boundary; where the arrival time T is obtained based on the level set principle, used to determine the first cluster Which defect of the defect is to be repaired.
  • the bad points in the domain are the current bad points that need to be repaired.
  • a template matching method is used to repair the current bad point.
  • extract the current bad point to be repaired from the defect data then, as shown in Figure 3, select a target template with the determined current bad point as the center, and set the search area with the current bad point as the center , For example, select a certain range centered on the current dead point as the search area.
  • search templates with the same size as the target template are sequentially selected, the similarity between the target template and the search template is calculated, and the data of the calculation result is stored.
  • the method for repairing the cluster defect further includes the following steps: after determining the defect boundary of the cluster defect, determine whether the defect boundary is empty, and if so, , It means that the defect boundary does not need to be repaired; if not, a repair step is performed, which also includes the sequence of determining the defect repair of the cluster defect shown in FIG. 2 and the process of repairing the determined defect shown in FIG. 3. Repeat the above determination and repair steps until the defect boundary is completely empty.
  • the repairing step further includes: after repairing the determined dead points, updating the T and F values of the repaired dead points, and then fixing the repaired dead points Add the defect boundary and delete the edge point with the smallest T value in the defect boundary.
  • FIG. 6a the black area is the cluster defect in the image
  • FIG. 6b is the image after repairing the cluster defect in FIG. 6a using the above method of this embodiment. It can be seen that the method of this embodiment fixes the cluster defects in the image very well, and the edges of the image are very clear.
  • this embodiment provides a method for repairing bad line defects.
  • This method also uses the template matching method provided in Embodiment 3 to directly repair the position where the bad point in the bad line defect is stored in the connected domain analysis.
  • the repair sequence of the broken lines is generally from left to right and from top to bottom.
  • the search range for bad line defects should try to avoid the area where the bad line is located.
  • the search template needs to be adaptively changed.
  • the black horizontal line in the image is the bad line defect in the image
  • FIG. 7b is the image after repairing the bad line defect in FIG. 7a using the template matching method. It can be seen from this that the method of this embodiment repairs bad line defects in the image well, the bad lines are effectively repaired, and the details of the image are well preserved.
  • the system includes an image acquisition module for acquiring a first image; and an image analysis module for performing an analysis on the first image acquired by the image acquisition module Connected domain analysis, extract all defective pixels in the first image, and divide the defective pixels into three categories according to the size and shape of the connected domain: isolated point defects, cluster defects and bad line defects; image repair The module is configured to repair the second image according to the type of the defective pixel determined by the image analysis module.
  • the image repair system further includes an image acquisition module that reads, magnifies, and analog-to-digital converts the second image signal, and The image signal is sent to the image processing module of the image repair system, and there is an image processing module for subsequent repair.
  • the image repair system further includes an image preprocessing module.
  • the image preprocessing module includes a background correction module and a gain correction module. Background correction and gain correction.
  • This embodiment provides a flat panel detector.
  • the flat panel detector includes the image repair system described in Embodiment 5, and the image repair system is integrated in the flat panel detector.
  • the flat panel detector further includes an image display module, which outputs and displays the repaired image.
  • the image acquisition module of the flat panel detector includes a scintillator, a thin film transistor (Thin Film Transistor, TFT), and a photodiode array.
  • the scintillator may include cesium iodide, which is commonly used and well-known in the art Scintillation crystal, the scintillator converts X-rays into visible light, and then the TFT and photodiode array convert the visible light into an electrical signal, which is then read, amplified, and analog-to-digital converted and transmitted to the image processing module.
  • the image processing module repairs the formed image.
  • the flat panel detector of this embodiment integrates the image repair system without changing the hardware design, which can reduce the product design and production costs and improve the shipment yield. In the subsequent upgrade and maintenance, it is only necessary to optimize and upgrade the software.
  • the image repair method, system, detector, server, and computer-readable storage medium provided by the foregoing embodiments have the following beneficial effects:
  • the method of the present invention performs statistics and analysis of defective pixels on the first image, such as the defective template in the detector, sorts the defective pixels, and repairs the second image in the order of the defective pixels from small to large, and Different methods are used to repair different defective pixels.
  • This repair method takes into account the similarity of neighboring pixels and conforms to the distribution characteristics of the image grayscale, which can not only effectively repair the defective pixels, but also retain the details of the original image well.
  • the method of the present invention performs pixel statistics and classification on the defect template in the detector, and repairs the image output by the detector in real time according to the classification, which can realize rapid repair of the defective pixel of the image.
  • the image repair method described in the present invention is only related to the type and size of defective pixels, such as defective dot defects, cluster defects, and defective line defects. It can be applied to flat panel detectors, so it can significantly improve work efficiency and reduce labor costs.
  • the image repair system described in the present invention can be integrated into a flat panel detector without requiring changes to the hardware design, which can reduce product costs and increase the yield of shipments. And in the subsequent upgrade and maintenance, you only need to optimize and upgrade the software.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Quality & Reliability (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Image Processing (AREA)
  • Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Multimedia (AREA)

Abstract

提供一种图像修复方法、图像修复系统及平板探测器，方法包括：获取第一图像，对第一图像进行连通域分析，提取所述第一图像中所有的缺陷像素；根据连通域的大小和形状将所述缺陷像素分为孤立点缺陷、团簇缺陷及坏线缺陷；输出第二图像并根据缺陷像素的类型对第二图像进行修复。该方法不仅可以有效修复缺陷像素，也能很好地保留原有图像的细节。根据对缺陷像素的分类对探测器实时输出的第二图像进行修复，实现对图像缺陷像素的快速修复。该方法只与缺陷像素的类型大小有关，适用性强，显著提高工作效率。图像修复系统集成在探测器中，不需要对硬件设计做改动，可以降低产品成本，后续的升级和维护中，只需对软件进行优化升级。

Description

一种图像修复方法、图像修复系统及平板探测器 技术领域
本发明涉及X射线探测器成像及图像修复领域，具体涉及一种图像修复方法、图像修复系统及平板探测器。
背景技术
对于X射线平板探测器，在探测器进行生产组装的过程中，由于各种各样的原因，探测器在成像时不可避免地会产生缺陷像素，即坏像素，在成像时不能真实的反映出像素单元接受的辐射能量。缺陷像素是指对X射线强度不响应或响应不良的探测器单元。产生缺陷像素的主要原因有闪烁体层自身的缺陷、光电二极管和薄膜晶体管自身的缺陷、驱动电路本身的损坏或探测器板材拼接不良等。根据缺陷像素对X射线响应的不同，缺陷像素可以分为死像素、响应不足像素、响应过度像素和响应混乱像素四种。而根据缺陷像素在图像连通域中分布的不同，又可以分为三种：孤立点缺陷、团簇缺陷和坏线缺陷。这些缺陷像素在探测器成像时表现出不同的形态，既降低了探测器出厂时的良率，也会影响到对所拍摄物体真实情况的判断。
当前，消除探测器缺陷像素的方法主要是两种方法：一是改善探测器生产流程和工艺，减少探测器缺陷像素的产生，但是此类方法的手段繁琐、复杂并且代价较高；另一类就是采用图像处理的方法，尽可能的还原探测器缺陷处的图像细节。例如现有技术中采用的一种图像处理的方法，针对缺陷像素主要使用样条插值法、双线性插值法以及三样条插值法等插值方法进行修复，这些修复方法考虑的是缺陷像素邻域内的正常像素点灰度值，可以在一定程度上还原图像的细节，但是在边缘类的细节上会造成模糊，也没有考虑图像的纹理等细节信息。
发明内容
有鉴于此，本发明的实施例提供了一种图像修复方法、图像修复系统及平板探测器，以便有效修复探测器输出的图像，并且使图像细节得到很好的保护。
根据第一方面，本发明实施例提供了一种图像修复方法，该包括如下步骤：
获取第一图像；
对所述第一图像进行连通域分析，提取所述第一图像中所有的缺陷像素；
根据所述连通域的大小和形状将所述缺陷像素分为三类：孤立点缺陷、团簇缺陷及坏线缺陷；
输出第二图像并根据所述缺陷像素的类型对所述第二图像进行修复。
优选地，按照所述缺陷像素从小到大的顺序对所述第二图像依次进行孤立点缺陷修复、团簇缺陷修复及坏线缺陷修复。
优选地，采用邻域加权平均的方法进行所述孤立点缺陷修复，修复后的像素灰度如下式所示：
Figure PCTCN2019075643-appb-000001
其中，F为孤立点缺陷3×3像素邻域内的原图数据，N为孤立点缺陷3×3像素邻域区域，K为孤立点缺陷3×3像素邻域内正常像素点对应的加权平均系数。
优选地，所述加权平均数的来源包括：
所述3×3像素邻域内所有正常像素点的灰度分布，计算出所述正常像素点的灰度梯度，其中梯度值大的所述正常像素点对应的加权系数比例高，梯度值较小的所述正常像素点对应的加权系数相应较小；
所述3×3像素邻域内所有正常像素点的位置分布，计算出所述正常像素点到所述孤立缺陷点的距离，距离小的所述正常像素点的加权系数比例高，距离较大的所述正常像素点对应的加权系数相应较小。
优选地，采用水平集和模板匹配相结合的方法对所述团簇缺陷进行修复，其中，所述水平集方法确定所述团簇缺陷的坏点的修复顺序；所述模板匹配方法对需要修复的当前坏点进行修复。
优选地，确定所述团簇缺陷的坏点的修复顺序包括以下步骤：
提取所述团簇缺陷并确定所述团簇缺陷的缺陷边界，初始化缺陷标记F和到达时间T，标记出所述缺陷边界的正常像素点、边缘点及缺陷点；
确定所述缺陷边界的所述边缘点中T值最小的所述边缘点；
确定T值最小的所述边缘点3×3像素邻域内的坏点，T值最小的所述边缘点的所述3×3像素邻域内的坏点即为需要修复的所述当前坏点。
优选地，对所述当前坏点进行修复包括以下步骤：
以所述当前坏点为中心选取目标模板；
以所述当前坏点为中心选取搜索区域；
在所述搜索区域内依次选取与目标模板同等大小的搜索模板，计算所述目标模板与所述搜索模板的相似性，并对计算结果进行存储；
搜索计算结果中相似性最大的数据，用所述相似性最大的数据所对应的搜索模板的中心像素值替代所述当前坏点的像素值。
优选地，还包括如下步骤：
判断所述缺陷边界是否为空，如果是，则无需修复；
如果否，则执行修复步骤，所述修复步骤包括确定所述团簇缺陷的坏点的修复顺序及对确定的所述坏点进行修复的步骤；
重复执行上述判定步骤和修复步骤，直至所述缺陷边界全部为空。
优选地，所述修复步骤还包括：修复完确定的所述坏点后，更新修复的所述坏点的T值和F值；
将修复后的所述坏点加入所述缺陷边界，并删除所述缺陷边界中T值最小的所述边缘点。
优选地，采用模板匹配的方法对所述坏线缺陷进行修复，按照所述坏线缺陷中的坏点在连通域分析中所存放的位置顺序进行修复。
优选地，对所述坏线缺陷进行修复时搜索范围避开所述坏线缺陷所在的区域，并且搜索模板根据所述坏线缺陷的宽度自适应变化。
优选地，所述方法还包括在对输出的所述第二图像进行修复之前，对输出的所述第二图像进行本底校正和增益校正的步骤。
根据第二方面，本发明实施例提供了一种图像修复系统，包括：
图像获取模块，用于获取第一图像；
图像分析模块，用于对所述图像获取模块获取的所述第一图像进行连通域分析，提取所述第一图像中所有的缺陷像素，并根据所述连通域的大小和形状将所述缺陷像素分为三类：孤立点缺陷、团簇缺陷及坏线缺陷；
图像修复模块，用于根据所述图像分析模块确定的所述缺陷像素的类型对第二图像进行修复。
优选地，图像修复系统还包括图像采集模块，对所述输出图像进行信号读取、放大及模数转换，并将经放大、转换的图像信号传送至所述图像修复模块。
优选地，图像修复系统还包括图像预处理模块，所述图像预处理模块包括本底校正模块和增益校正模块，所述本底校正模块对所述输出图像进行本底校正，所述增益校正模块用于对所述输出图像进行增益校正的。
根据第三方面，本发明实施例提供了一种平板探测器，该平板探测器包括图像显示模块以及本发明上述第二方面所述的图像修复系统，所述图像修复系统集成在所述平板探测器中图像显示模块对由所述图像修复系统修复过的图像进行输出显示。
本发明提供的图像修复方法、图像修复系统及平板探测器，具有如下有益效果：
1、本发明的方法对第一图像，例如探测器中的缺陷模板，进行缺陷像素的统计和分析，对缺陷像素进行排序，按照缺陷像素由小到大的顺序对第二图像进行修复，并且对不同的缺陷像素采用不同的方法进行修复。该修复方法考虑到了邻近像素的相似性，符合图像灰度的分布特性，不仅可以有效修复缺陷像素，也能很好地保留原有图像的细节。
2、本发明的方法对探测器中的缺陷模板进行像素统计和分类，根据该分类对探测器实时输出的图像进行修复，可以实现对图像缺陷像素的快速修复。
3、本发明所述的图像修复方法只与缺陷像素的类型大小有关，例如坏点缺陷、团簇缺陷、坏线缺陷。对于平板探测器均能适用，因此能显著提高工作效率，降低人力成本。
4、本发明所述的图像修复系统能够集成在平板探测器中，不需要对硬件设计做改动，可以降低产品成本，提高出货良率。并且在后续的升级和维护中，也只需对软件进行优化升级即可。
附图说明
通过参考附图会更加清楚的理解本发明的特征和优点，附图是示意性的而不应理解为对本发明进行任何限制，在附图中：
图1显示为本发明实施例一提供的图像修复方法的流程图。
图2显示为本发明实施例三所示方法中的水平集方法的流程图。
图3显示为本发明实施例三所示方法中的模板匹配方法的流程图。
图4显示为本发明实施例三的优选实施例的修复方法的流程图。
图5显示为本发明实施例三的另一优选实施例的修复方法的流程图。
图6a和6b分别显示为存在团簇缺陷的缺陷图像及根据实施例三的方法修复后的图像。
图7a和7b分别显示为存在坏线缺陷的缺陷图像及根据实施例四的方法修复后的图像。
图8显示为本发明实施例五提供的图像处理系统的示意图。
图9显示为本发明实施例六提供的平板探测器的示意图。
具体实施方式
为使本发明实施例的目的、技术方案和优点更加清楚，下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
实施例一
本实施例提供一种图像修复方法，如图1所示，该方法包括如下步骤：
获取第一图像；
对所述第一图像进行连通域分析，提取所述第一图像中所有的缺陷像素；
根据所述连通域的大小和形状将所述缺陷像素分为三类：孤立点缺陷、团簇缺陷及坏线缺陷；
输出第二图像并根据所述缺陷像素的类型对所述第二图像进行修复。
在本实施例中，上述第一图像优选为探测器等仪器中的缺陷模板。
在本实施例的进一步优选实施例中，可以按照从左到右、由上至下的顺序对第一图像，例如上述缺陷模板，进行连通域分析。
在本实施例的一优选实施例中，根据对缺陷像素的分类，按照缺陷像素的大小，从小到大对第二图像依次进行孤立点缺陷修复、团簇缺陷修复及坏线缺陷修复。按照该从小到大的顺序进行图像修复，能够达到良好的修复效果。
在本实施例的一优选实施例中，在对第二图像进行修复之前，首先对所述第二图像进行本底校正和增益校正。
本实施例的方法对第一图像，例如探测器等中的缺陷模板进行像素统计和分类，根据该分类对例如探测器实时输出的第二图像进行修复，可以实现对图像缺陷像素的快速修复。
实施例二
在实施例一对缺陷像素进行分类并按照缺陷像素大小，按照从小到大的顺序进行图像修复的基础上，本实施例提供了一种修复孤立点缺陷的方法，在本实施例中，采用领域加权平均的方法进行孤立点修复。
由于单个像素并不会包含太多的图像细节，因此在本优选实施例中采用了上述邻域加权平均的方法实现对孤立点缺陷的修复。采用此方法修复后的孤立点的灰度如下公式(1)所示：
Figure PCTCN2019075643-appb-000002
其中，F为孤立点缺陷3×3像素邻域内的原图数据，N为孤立点缺陷3×3像素邻域区域，K为孤立点缺陷3×3像素邻域内正常像素点对应的加权平均系数。
在本实施例的进一步优选实施例中，上述加权平均数的来源主要包括以下两个：
3×3像素邻域内所有正常像素点的灰度分布，计算出所述正常像素点的灰度梯度，其中梯度值大的所述正常像素点对应的加权系数比例高，梯度值较小的所述正常像素点对应的加权系数相应较小；
3×3像素邻域内所有正常像素点的位置分布，计算出所述正常像素点到所述孤立缺陷点的距离，距离小的所述正常像素点的加权系数比例高，距离较大的所述正常像素点对应的加权系数相应较小。
如上所示，依据本实施例的方法对孤立点缺陷进行修复时，充分考虑了该缺陷像素点邻域内的正常像素点的灰度分布及位置分布，并且利用正常像素点对应的加权平均数对孤立点进行修复。该方法能够有效修复孤立点缺陷，并且能够很好地还原原有图像的细节。
实施例三
与实施例二同理，在实施例一的基础上，本实施例提供一种团簇缺陷的修复方法，本实施例采用水平集和模板匹配相结合的方法对所述团簇缺陷进行修复，其中，水平集方法确定所述团簇缺陷的坏点的修复顺序；模板匹配方法对确定需要修复的当前坏点进行修复。
在本实施例的一优选实施例中，如图2所示，采用水平集方法确定所述团簇缺陷的坏点的修复顺序包括如下步骤：提取所述团簇缺陷并确定所述团簇缺陷的缺陷边界，初始化 缺陷标记F和到达时间T，标记出所述缺陷边界的正常像素点、边缘点及缺陷点；其中到达时间T是基于水平集原理得来的，用来确定先对团簇缺陷的哪一个坏点进行修复。
确定所述缺陷边界的所述边缘点中T值最小的所述边缘点；确定T值最小的边缘点3×3像素邻域内的坏点，T值最小的所述边缘点3×3像素邻域内的坏点即为需要修复的所述当前坏点。
确定了所述当前坏点之后，采用模板匹配方法对该当前坏点进行修复。如图3所示，在缺陷数据中提取要修复的该当前坏点；然后，如图3所示，以确定的当前坏点为中心选取一个目标模板，并且以当前坏点为中心设置搜索区域，例如选取以该当前坏点为中心的一定范围作为搜索区域。在该搜索区域内依次选取与目标模板同等大小的搜索模板，计算目标模板与该搜索模板的相似性，并对计算结果的数据进行存储。搜索计算结果中相似性最大的数据；最后利用相似性最大的该数据对应的搜索模板的中心像素值替代该当前坏点的像素值，完成当前坏点的修复。如此循环进行不同坏点的修复，直至整个团簇缺陷被完全修复。
在本实施例的另一优选实施例中，如图4所示，对团簇缺陷进行修复的方法还包括如下步骤：确定该团簇缺陷的缺陷边界之后，确定缺陷边界是否为空，如果是，则表示该缺陷边界无需修复；如果否，则执行修复步骤，该修复步骤同样包括图2所示的确定团簇缺陷的坏点修复顺序及图3所示的修复确定的坏点的过程。重复上述判定步骤和修复步骤，直至缺陷边界全部为空。
在本实施例的另一优选实施例中，如图5所示，修复步骤还包括：修复完确定的坏点之后，更新修复的坏点的T值和F值，然后将修复后的坏点加入缺陷边界，并删除缺陷边界中该T值最小的边缘点。
如图6a所示，其中的黑色区域即为图像中的团簇缺陷，图6b为采用本实施例的上述方法对图6a中的团簇缺陷进行修复后的图像。由此可见，本实施例的方法很好地修复了图像中的团簇缺陷，并且图像的边缘是非常清晰的。
实施例四
同样在实施例一的基础上，本实施例提供一种坏线缺陷的修复方法。该方法同样采用实施例三所提供的模板匹配方法，直接按照坏线缺陷中的坏点在连通域分析中所存放的位置进行修复。
在本实施例的一优选实施例中，对坏线的修复顺序一般为从左到右、由上至下。在本实施例中，针对坏线缺陷的搜索范围要尽量避开坏线所在的区域。
在本实施例的另一优选实施例中，根据图像中坏线的宽度，搜索模板需要自适应地进行变化。
如图7a所示，图像中的黑色水平线即为图像中的坏线缺陷，图7b为利用模板匹配方法对图7a中的坏线缺陷进行修复后的图像。由此可见，本实施例的方法很好地修复了图像中的坏线缺陷，坏线被有效修复，并且图像的细节被很好地保留。
实施例五
本实施例提供一种图像修复系统，如图8所示，该系统包括图像获取模块，用于获取第一图像；图像分析模块，用于对所述图像获取模块获取的所述第一图像进行连通域分析，提取所述第一图像中所有的缺陷像素，并根据所述连通域的大小和形状将所述缺陷像素分为三类：孤立点缺陷、团簇缺陷及坏线缺陷；图像修复模块，用于根据所述图像分析模块确定的所述缺陷像素的类型对第二图像进行修复。
在本实施例的一优选实施例中，如图8所示，该图像修复系统还包括图像采集模块，其对第二图像信号进行读取、放大及模数转换，并将经放大、转换后的图像信号传送至图像修复系统的图像处理模块，有图像处理模块进行后续的修复。
在本实施例的另一优选实施例中，如图8所示，该图像修复系统还包括图像预处理模块，该图像预处理模块包括本底校正模块和增益校正模块，分别对第二图像进行本底校正和增益校正。
实施例六
本实施例提供一种平板探测器，如图9所示，该平板探测器包括实施例五所述的图像修复系统，该图像修复系统集成在该平板探测器中。
如图9所示，该平板探测器还包括图像显示模块，由该图像显示模块对被修复的图像进行输出显示。
在本实施例的一优选实施例中，该平板探测器的图像采集模块包括闪烁体、薄膜晶体管(Thin Film Transistor，TFT)和光电二极管阵列，闪烁体可以包括碘化铯等本领域常用和公知的闪烁晶体，该闪烁体将X射线转化为可见光，然后TFT和光电二极管阵列将该可 见光转换成电信号，该电信号随后被读取并经放大、模数转换后传输至图像处理模块，有图像处理模块对形成的图像进行修复。
综上看出，本实施例的平板探测器集成了图像修复系统，其硬件设计并无改动，由此可以降低产品的设计生产成本，提高出货良率。在后续的升级和维护中，也只需对软件进行优化升级即可。
如上所述，上述实施例提供的图像修复方法、系统、探测器、服务器及计算机可读存储介质，具有如下有益效果：
1、本发明的方法对第一图像，例如探测器中的缺陷模板，进行缺陷像素的统计和分析，对缺陷像素进行排序，按照缺陷像素由小到大的顺序对第二图像进行修复，并且对不同的缺陷像素采用不同的方法进行修复。该修复方法考虑到了邻近像素的相似性，符合图像灰度的分布特性，不仅可以有效修复缺陷像素，也能很好地保留原有图像的细节。
2、本发明的方法对探测器中的缺陷模板进行像素统计和分类，根据该分类对探测器实时输出的图像进行修复，可以实现对图像缺陷像素的快速修复。
3、本发明所述的图像修复方法只与缺陷像素的类型大小有关，例如坏点缺陷、团簇缺陷、坏线缺陷。对于平板探测器均能适用，因此能显著提高工作效率，降低人力成本。
4、本发明所述的图像修复系统能够集成在平板探测仪中，不需要对硬件设计做改动，可以降低产品成本，提高出货良率。并且在后续的升级和维护中，也只需对软件进行优化升级即可。
上述实施例仅例示性说明本发明的原理及其功效，而非用于限制本发明，本领域技术人员可以在不脱离本发明的精神和范围的情况下作出各种修改和变型，比如变更邻域加权模板的尺寸和系数，变更模板匹配的尺寸，变更模板匹配中使用的数据相关参数，变更水平集控制修复顺序的方式，或者利用本发明形同的思路进行邻域加权平均修复缺陷像素、水平集控制+模板匹配修复缺陷像素等。这样的修改和变型均落入由所附权利要求所限定的范围之内。

Claims (16)

  1. 一种图像修复方法，其特征在于，包括如下步骤：
    获取第一图像；
    对所述第一图像进行连通域分析，提取所述第一图像中所有的缺陷像素；
    根据所述连通域的大小和形状将所述缺陷像素分为三类：孤立点缺陷、团簇缺陷及坏线缺陷；
    输出第二图像并根据所述缺陷像素的类型对所述第二图像进行修复。
  2. 根据权利要求1所述的图像修复方法，其特征在于，按照所述缺陷像素从小到大的顺序对所述第二图像依次进行孤立点缺陷修复、团簇缺陷修复及坏线缺陷修复。
  3. 根据权利要求2所述的图像修复方法，其特征在于，采用邻域加权平均的方法进行所述孤立点缺陷修复，修复后的像素灰度如下式所示：
    Figure PCTCN2019075643-appb-100001
    其中，F为孤立点缺陷3×3像素邻域内的原图数据，N为孤立点缺陷3×3像素邻域区域，K为孤立点缺陷3×3像素邻域内正常像素点对应的加权平均系数。
  4. 根据权利要求3所述的图像修复方法，其特征在于，所述加权平均数的来源包括：
    所述3×3像素邻域内所有正常像素点的灰度分布，计算出所述正常像素点的灰度梯度，其中梯度值大的所述正常像素点对应的加权系数比例高，梯度值较小的所述正常像素点对应的加权系数相应较小；
    所述3×3像素邻域内所有正常像素点的位置分布，计算出所述正常像素点到所述孤立缺陷点的距离，距离小的所述正常像素点的加权系数比例高，距离较大的所述正常像素点对应的加权系数相应较小。
  5. 根据权利要求2所述的图像修复方法，其特征在于，采用水平集和模板匹配相结合的方法对所述团簇缺陷进行修复，其中，所述水平集方法确定所述团簇缺陷的坏点的修复顺序；所述模板匹配方法对需要修复的当前坏点进行修复。
  6. 根据权利要求5所述的图像修复方法，其特征在于，确定所述团簇缺陷的坏点的修复顺序包括以下步骤：
    提取所述团簇缺陷并确定所述团簇缺陷的缺陷边界，初始化缺陷标记F和到达时间T，标记出所述缺陷边界的正常像素点、边缘点及缺陷点；
    确定所述缺陷边界的所述边缘点中T值最小的所述边缘点；
    确定T值最小的所述边缘点3×3像素邻域内的坏点，T值最小的所述边缘点的所述3×3像素邻域内的坏点即为需要修复的所述当前坏点。
  7. 根据权利要求6所述的图像修复方法，其特征在于，对所述当前坏点进行修复包括以下步骤：
    以所述当前坏点为中心选取目标模板；
    以所述当前坏点为中心选取搜索区域；
    在所述搜索区域内依次选取与目标模板同等大小的搜索模板，计算所述目标模板与所述搜索模板的相似性，并对计算结果进行存储；
    搜索所述计算结果中相似性最大的数据，用所述相似性最大的数据所对应的所述搜索模板的中心像素值替代所述当前坏点的像素值。
  8. 根据权利要求7所述的图像修复方法，其特征在于，还包括如下步骤：
    判断所述缺陷边界是否为空，如果是，则无需修复；
    如果否，则执行修复步骤，所述修复步骤包括确定所述团簇缺陷的坏点的修复顺序及对确定的所述坏点进行修复的步骤；
    重复执行上述判定步骤和修复步骤，直至所述缺陷边界全部为空。
  9. 根据权利要求8所述的图像修复方法，其特征在于，所述修复步骤还包括：修复完确定的所述坏点后，更新修复的所述坏点的T值和F值；
    将修复后的所述坏点加入所述缺陷边界，并删除所述缺陷边界中T值最小的所述边缘点。
  10. 根据权利要求2所述的图像修复方法，其特征在于，采用模板匹配的方法对所述坏线缺陷进行修复，按照所述坏线缺陷中的坏点在连通域分析中所存放的位置顺序进行修复。
  11. 根据权利要求10所述的图像修复方法，其特征在于，对所述坏线缺陷进行修复时搜索范围避开所述坏线缺陷所在的区域，并且搜索模板根据所述坏线缺陷的宽度自适应变化。
  12. 根据权利要求1-11所述的图像修复方法，其特征在于，还包括在对输出的所述第二图像进行修复之前，对所述第二图像进行本底校正和增益校正的步骤。
  13. 一种图像修复系统，其特征在于，包括：
    图像获取模块，用于获取第一图像；
    图像分析模块，用于对所述图像获取模块获取的所述第一图像进行连通域分析，提取所述第一图像中所有的缺陷像素，并根据所述连通域的大小和形状将所述缺陷像素分为三类：孤立点缺陷、团簇缺陷及坏线缺陷；
    图像修复模块，用于根据所述图像分析模块确定的所述缺陷像素的类型对第二图像进行修复。
  14. 根据权利要求13所述的图像修复系统，其特征在于，还包括图像采集模块，对所述第二图像进行信号读取、放大及模数转换，并将经放大、转换的图像信号传送至所述图像修复模块。
  15. 根据权利要求14所述的图像修复系统，其特征在于，还包括图像预处理模块，所述图像预处理模块包括本底校正模块和增益校正模块，所述本底校正模块对所述第二图像进行本底校正，所述增益校正模块用于对所述第二图像进行增益校正的。
  16. 一种平板探测器，其特征在于，包括图像显示模块以及权利要求13-15中任一项所述的图像修复系统，所述图像修复系统集成在所述平板探测器中，所述图像显示模块对由所述图像修复系统修复过的图像进行输出显示。
PCT/CN2019/075643 2018-12-10 2019-02-21 一种图像修复方法、图像修复系统及平板探测器 WO2020118896A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US16/757,775 US11461877B2 (en) 2018-12-10 2019-02-21 Image inpainting method, image inpainting system and flat panel detector thereof

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811502255.2 2018-12-10
CN201811502255.2A CN109712084B (zh) 2018-12-10 2018-12-10 一种图像修复方法、图像修复系统及平板探测器

Publications (1)

Publication Number Publication Date
WO2020118896A1 true WO2020118896A1 (zh) 2020-06-18

Family

ID=66255558

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/075643 WO2020118896A1 (zh) 2018-12-10 2019-02-21 一种图像修复方法、图像修复系统及平板探测器

Country Status (3)

Country Link
US (1) US11461877B2 (zh)
CN (1) CN109712084B (zh)
WO (1) WO2020118896A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112904841A (zh) * 2021-01-12 2021-06-04 北京布科思科技有限公司 非水平朝向的单线定位避障方法和装置、设备及存储介质

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111353946B (zh) * 2018-12-21 2023-04-11 腾讯科技（深圳）有限公司 图像修复方法、装置、设备及存储介质
CN110738613B (zh) * 2019-09-29 2022-02-18 上海奕瑞光电子科技股份有限公司 线阵探测器图像拼接实时校正方法、装置、设备和介质
CN116109523B (zh) * 2023-04-11 2023-06-30 深圳奥雅设计股份有限公司 一种智能设计的图像缺陷点自动修复方法及系统
CN117934469B (zh) * 2024-03-22 2024-06-07 张家港市升恒机械设备制造有限公司 一种矿用高压变频器外壳生产质量视觉检测方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2005149399A (ja) * 2003-11-19 2005-06-09 Sumitomo Bakelite Co Ltd 画像処理方法とそれを応用した金属腐食検査方法
CN103543168A (zh) * 2013-10-12 2014-01-29 华南理工大学 一种多层封装基板缺陷的x射线检测方法及系统
CN104063850A (zh) * 2014-06-24 2014-09-24 广东互维科技有限公司 红眼修复方法
CN106780497A (zh) * 2016-11-28 2017-05-31 重庆大学 一种基于统计信息的器官血管树自动提取方法
CN108399632A (zh) * 2018-03-02 2018-08-14 重庆邮电大学 一种联合彩色图像的rgb-d相机深度图像修复方法

Family Cites Families (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5608816A (en) * 1993-12-24 1997-03-04 Matsushita Electric Industrial Co., Ltd. Apparatus for inspecting a wiring pattern according to a micro-inspection and a macro-inspection performed in parallel
US6233364B1 (en) * 1998-09-18 2001-05-15 Dainippon Screen Engineering Of America Incorporated Method and system for detecting and tagging dust and scratches in a digital image
US6950547B2 (en) * 2001-02-12 2005-09-27 3M Innovative Properties Company Web inspection method and device
US7551769B2 (en) * 2003-02-18 2009-06-23 Marena Systems Corporation Data structures and algorithms for precise defect location by analyzing artifacts
KR100617781B1 (ko) * 2004-06-29 2006-08-28 삼성전자주식회사 이미지 센서의 화질 개선장치 및 방법
CN101126724B (zh) * 2007-09-30 2011-01-05 西北工业大学 锥束ct系统中平板探测器图像的抗干扰校正方法
US8798373B2 (en) * 2009-11-20 2014-08-05 Nec Corporation Image restoration system, image restoration method, and image restoration program
CN102426376B (zh) * 2011-08-11 2013-05-08 西北工业大学 一种平板探测器的监控与校正方法
CN104835125B (zh) * 2015-05-08 2018-04-03 上海联影医疗科技有限公司 平板探测器的坏点校正方法
CN105787905B (zh) * 2016-03-24 2019-03-26 中国人民解放军信息工程大学 一种基于动态电流的锥束ct环状伪影校正方法
CN106097282B (zh) * 2016-07-01 2019-02-26 上海奕瑞光电子科技股份有限公司 消除平板探测器图像残影的方法及平板探测器
CN108171765B (zh) * 2017-12-08 2021-06-18 上海奕瑞光电子科技股份有限公司 平板探测器残影校正方法及其校正装置
CN109934802B (zh) * 2019-02-02 2021-06-22 浙江工业大学 一种基于傅里叶变换和图像形态学的布匹疵点检测方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2005149399A (ja) * 2003-11-19 2005-06-09 Sumitomo Bakelite Co Ltd 画像処理方法とそれを応用した金属腐食検査方法
CN103543168A (zh) * 2013-10-12 2014-01-29 华南理工大学 一种多层封装基板缺陷的x射线检测方法及系统
CN104063850A (zh) * 2014-06-24 2014-09-24 广东互维科技有限公司 红眼修复方法
CN106780497A (zh) * 2016-11-28 2017-05-31 重庆大学 一种基于统计信息的器官血管树自动提取方法
CN108399632A (zh) * 2018-03-02 2018-08-14 重庆邮电大学 一种联合彩色图像的rgb-d相机深度图像修复方法

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112904841A (zh) * 2021-01-12 2021-06-04 北京布科思科技有限公司 非水平朝向的单线定位避障方法和装置、设备及存储介质
CN112904841B (zh) * 2021-01-12 2023-11-03 北京布科思科技有限公司 非水平朝向的单线定位避障方法和装置、设备及存储介质

Also Published As

Publication number Publication date
CN109712084B (zh) 2021-01-19
US20210295477A1 (en) 2021-09-23
US11461877B2 (en) 2022-10-04
CN109712084A (zh) 2019-05-03

Similar Documents

Publication Publication Date Title
WO2020118896A1 (zh) 一种图像修复方法、图像修复系统及平板探测器
CN103048331B (zh) 基于柔性模板配准的印刷缺陷检测方法
JP7292979B2 (ja) 画像処理装置及び画像処理方法
TW202029071A (zh) 用於判定空白光罩上之缺陷之類型及尺寸之系統及方法
JP2001175857A (ja) 参照画像作成方法、パターン検査装置及び参照画像作成プログラムを記録した記録媒体
CN111308448A (zh) 图像采集设备与雷达的外参确定方法及装置
JP2012032370A (ja) 欠陥検出方法、欠陥検出装置、学習方法、プログラム、及び記録媒体
CN105865723A (zh) 气体泄漏检测非均匀校正方法及气体泄漏检测装置
US10116863B2 (en) Scanning with frame averaging
CN116228780B (zh) 基于计算机视觉的硅片缺陷检测方法及系统
CN103400343A (zh) 一种补偿夜间红外下视图像亮度不均匀的方法
US11852465B2 (en) Wafer inspection method and apparatus thereof
Cao et al. Vignetting correction based on a two-dimensional Gaussian filter with harmony for area array sensors
WO2021189490A1 (zh) X射线平板探测器及其图像校正方法
CN115082504B (zh) 一种太阳能光伏板光斑识别方法
JP2021117155A (ja) 検査装置、検査方法、及び検査プログラム
WO2019176614A1 (ja) 画像処理装置、画像処理方法、およびコンピュータプログラム
CN115988334A (zh) 一种自校正的数字化相机移动遥感系统及方法
US11972562B2 (en) Method for determining plant growth curve and electronic device
JP3788586B2 (ja) パターン検査装置および方法
CN111983710B (zh) 一种用于扫描型红外搜索系统的非均匀性校正方法
CN114298995A (zh) 一种图像缺陷元抑制方法、设备及计算机可读存储介质
TW202223836A (zh) 校正異常點雲資料之方法
CN112651925A (zh) 光伏组件的诊断方法和装置
JP2723174B2 (ja) 異種センサ画像間レジストレーション補正方法

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: 19896146

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: 19896146

Country of ref document: EP

Kind code of ref document: A1