WO2022252442A1 - 一种针对皮革纤维mct断层扫描图像的智能去噪方法及应用 - Google Patents

一种针对皮革纤维mct断层扫描图像的智能去噪方法及应用 Download PDF

Info

Publication number
WO2022252442A1
WO2022252442A1 PCT/CN2021/120005 CN2021120005W WO2022252442A1 WO 2022252442 A1 WO2022252442 A1 WO 2022252442A1 CN 2021120005 W CN2021120005 W CN 2021120005W WO 2022252442 A1 WO2022252442 A1 WO 2022252442A1
Authority
WO
WIPO (PCT)
Prior art keywords
mct
gray value
denoising
image
frequency
Prior art date
Application number
PCT/CN2021/120005
Other languages
English (en)
French (fr)
Inventor
华玉爱
李天铎
芦建梅
张华勇
许静
Original Assignee
齐鲁工业大学
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 齐鲁工业大学 filed Critical 齐鲁工业大学
Publication of WO2022252442A1 publication Critical patent/WO2022252442A1/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
    • G06T5/70—Denoising; Smoothing
    • 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
    • G06T2207/10072—Tomographic images
    • G06T2207/10101—Optical tomography; Optical coherence tomography [OCT]
    • 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
    • G06T2207/20—Special algorithmic details
    • G06T2207/20024—Filtering details
    • G06T2207/20032—Median filtering
    • 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
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30124—Fabrics; Textile; Paper

Definitions

  • the invention discloses an intelligent denoising method, device, electronic equipment, readable storage medium and application for leather fiber MCT tomographic scanning images, belonging to the technical field of leather image denoising processing.
  • the weaving structure of leather fibers has always been a concern in the industry. So far, people's understanding of the weaving structure of leather fibers is still very superficial. Studying the weaving structure of leather fibers has important theoretical and practical values.
  • the acquisition methods of leather fiber tomographic images mainly include: non-destructive methods such as X-ray tomography (CT) method and nuclear magnetic resonance (MRI) method, and destructive methods such as physical slicing method and layer removal method.
  • CT X-ray tomography
  • MRI nuclear magnetic resonance
  • the cost of nuclear magnetic resonance equipment and maintenance is high, and the destructive sampling method is to physically slice or polish the embedded leather fibers, which requires high process requirements and is difficult. serial tomographic images.
  • the micro X-ray tomography equipment and maintenance costs are not high, and the operation is convenient, and the micro X-ray tomography (MCT) technology can be used without destroying the sample.
  • MCT micro X-ray tomography
  • the 3D reconstruction of leather fibers can be carried out by using the MCT tomographic images of leather fibers, so as to display the weaving structure of leather fibers, so as to study its regularity. But MCT tomographic images of leather fibers are loaded with electronic and optical noise, which usually have low gray values in the images. Denoising and segmentation of MCT tomographic images of leather fibers is the basis for 3D reconstruction. Currently, there are many methods for image denoising, segmentation and reconstruction, but there is no one general method that can handle all types of images. There are many image processing methods in the medical field, and 3D reconstruction software has also been developed. But so far, there are still many unsolved problems in the denoising and segmentation of medical images, and no method or software can perfectly complete the task of medical image processing.
  • Zhang Huayong and others used micro-X-ray tomography (MCT) technology to collect tomographic images of leather fibers and applied 3D-Doctor software in the medical field, supplemented by self-developed computer image processing programs for image processing.
  • MCT micro-X-ray tomography
  • 3D-Doctor software in the medical field, supplemented by self-developed computer image processing programs for image processing.
  • denoising and 3D reconstruction the 3D reconstruction effect of leather fibers is obtained.
  • the 3D reconstructed image obtained by the medical image processing technology cannot reveal the fine texture structure of the leather fiber bundle surface.
  • the invention discloses an intelligent denoising method for in-situ MCT tomographic images of leather fibers.
  • the invention also discloses a device loaded with the method.
  • the invention also discloses an electronic device loaded with the method.
  • the invention also discloses an application method utilizing the method.
  • the invention mainly includes the median filtering of the leather fiber MCT tomographic image, the smoothing of the frequency sequence of the gray value and the exploration of the minimum value of the frequency of the gray value.
  • the optimal window size and optimal filtering times of the image median filter through trial and error, and then calculate the gray value frequency and determine it according to the maximum and minimum values of the gray value frequency Denoising grayscale threshold. Since the gray value histogram is usually not smooth, but presents zigzag fluctuations, and its minimum values are numerous and not unique, it is necessary to smooth the gray value histogram multiple times to eliminate such random disturbances.
  • the present invention designs a computer to automatically search for the frequency smoothing times of the optimal gray value, thereby determining the gray threshold for image denoising, so that the image denoising operation is fast and accurate, and the denoising effect is remarkable.
  • the method of the present invention automatically explores the boundary point between the gray value of the object (ie leather fiber) and the noise gray value of the MCT image smoothed by the median filter, and uses the boundary point as the denoising threshold to remove the gray threshold of the image. Noise processing to complete the denoising work.
  • An intelligent denoising method for leather fiber MCT tomographic images characterized in that it comprises:
  • the gray value with the smallest frequency is used as the boundary point between the object gray value and the noise gray value, that is, the denoising threshold, and the pixels whose gray value is lower than the denoising threshold are regarded as background noise points , and the remaining points are used as object points.
  • the object gray value frequency histogram and the noise gray value frequency histogram each have a single peak, that is, the gray value corresponding to the trough of the gray value frequency histogram in the form of a double peak has the minimum frequency gray value as the object gray value
  • the denoising threshold is the denoising threshold with the denoising threshold of the object noise gray value. Due to the inherent characteristics of the leather fiber MCT tomographic image, the noise gray value frequency peak is always greater than the object gray value frequency peak. Therefore, the denoising involved in the present invention The algorithm takes the minimum value point of the gray value frequency between the mode of the noise gray value and the mode of the object gray value (that is, the gray value with the smallest gray value frequency) as the denoising threshold.
  • a computer when determining the denoising threshold, a computer is used to automatically detect the gray-scale frequency minimum point of the smooth leather fiber MCT tomographic image obtained through said step 1) as said denoising Threshold, that is, the minimum value of the gray value frequency sequence between two gray value frequency peaks is used as the denoising threshold.
  • the gray value of the leather fiber MCT tomographic image is between 0 and 255, and most of the pixels with a gray value of 0 are outside the field of view of the MCT device.
  • the boundary point of the gray value of the leather fiber and the noise is usually the valley of its frequency histogram, therefore, the work of exploring the denoising threshold is equivalent to exploring the valley of the gray value frequency distribution, that is, the gray value The minimum value of the frequency series between two gray value frequency peaks.
  • step 2) smoothing is performed on the frequency sequence of gray values for multiple times. Since the gray value histogram is usually not smooth, that is, it often presents zigzag fluctuations, and its minimum value is usually not a trough, but a local phenomenon of random disturbance, so it is necessary to smooth the gray value frequency sequence multiple times. Eliminating such random disturbances can avoid or reduce the calculation error of the minimum point of the gray value.
  • the commonly used image smoothing algorithms include Gaussian filtering, median filtering and arithmetic mean filtering and many other methods.
  • the smoothing algorithm is an arithmetic mean filtering method, specifically:
  • the gray value frequency of some leather fiber MCT tomographic images may have a minimum value before the noise mode.
  • the reference value of p is 7. If p is too large, the calculation speed will be affected.
  • the selection window width is 5, which can be adjusted for the smoothing effect of the video number histogram in practical applications.
  • the characteristics of this technical solution are: if the number of times of filtering p is too small, the random disturbance in the frequency sequence of gray values cannot be effectively eliminated, which will lead to a large deviation of the obtained minimum point; if the number of times of filtering is too large, It will make the trough of the gray value frequency distribution disappear, resulting in the absence of the minimum value point of the gray value frequency sequence, and the boundary point between the object and the noise gray value cannot be determined. Therefore, if the minimum value of the frequency sequence of gray values is not found, the computer automatically reduces the number of filters and performs the arithmetic mean filtering operation again, and so on until the minimum value of the frequency sequence of gray values is obtained, and this value is used as threshold. The gray value of the pixels whose gray value is less than this threshold is set to 0, so as to complete the image denoising task.
  • the algorithm is invalid, which indicates that the gray value distribution of the image has no valley.
  • the present invention designs a computer to automatically explore the optimal gray-value frequency smoothing times, so that each frame MCT
  • the tomographic image automatically determines the gray threshold for image denoising, which makes the image denoising operation fast and accurate.
  • Common methods for smoothing the frequency histogram of gray values include Gaussian filtering, median filtering, and mean filtering. Experiments show that the arithmetic mean filtering method has the best effect.
  • the algorithm of the present invention does not calculate the mode of the object, but only calculates the minimum frequency value after the mode of the noise gray value.
  • the gray value frequency histogram is smoothed too few times, its random disturbance cannot be effectively eliminated; but if the number of smoothing is too many, its troughs will also disappear. Therefore, choosing an appropriate gray value frequency histogram is the key to determine the gray value denoising threshold.
  • the gray value used in the present invention may be different for each image frame.
  • the invention also discloses a device loaded with the method, characterized in that the device includes:
  • Data collection module used to collect leather fiber MCT tomographic images or leather fiber embedded MCT tomographic images
  • Data processing module used to perform a median filter module on the collected scanned image and distinguish background noise points and object points according to the denoising threshold, thereby removing background noise;
  • Data output module used to output the denoised scanned image.
  • the invention also discloses an electronic device loaded with the method, including a memory, a processor, and a computer program stored in the memory and operable on the processor, wherein the processor implements the program when executing the program.
  • the present invention also discloses a computer-readable storage medium, on which a computer program is stored, and is characterized in that, when the computer program is executed by a processor, the intelligent denoising method for leather fiber MCT tomographic images is realized.
  • the invention also discloses that the method is applied to the denoising process of the leather fiber in situ MCT tomographic image and the leather fiber embedded MCT tomographic image.
  • the main purpose of the present invention is to explore the boundary between the leather fiber image and the background noise in the leather fiber MCT tomographic image, and remove the noise. Since the gray value difference between the leather fiber image and the background noise in the leather fiber MCT image is relatively large, the median filter can make the boundary between the leather fiber image and the background noise clearer, so the median filter is selected for the leather fiber MCT. The original image is filtered.
  • the window size and filtering times of median filtering have a great influence on the filtering effect, it is very important to choose an appropriate window size and filtering times of median filtering.
  • the present invention suggests that the size of the reference window for median filtering is 11 ⁇ 11, and the number of filtering times is 5-11. In practical applications, the window size and filtering times of median filtering can be adjusted appropriately.
  • the present invention uses the valley value between the peak value of the gray scale frequency of the object image and the peak value of the noise gray scale frequency as the denoising threshold.
  • the present invention adopts intelligent filtering algorithm to automatically search for the optimal frequency smoothing number of gray values, thereby automatically determining the gray threshold of image denoising. This algorithm is simple, fast, accurate, and completely intelligent, and the denoising effect is remarkable.
  • the method of the present invention is suitable for denoising scenes of leather fiber in situ MCT tomographic images and leather fiber embedded MCT tomographic images.
  • Figure 1a Original MCT tomographic image
  • Figure 1b Smoothed MCT tomographic image
  • Figure 1c denoised MCT tomographic image
  • Figure 2a Grayscale histogram of the original MCT tomographic image
  • Figure 2b The grayscale histogram after smoothing the original MCT tomographic image
  • Figure 2c Curve after smoothing the grayscale histogram of the smoothed MCT tomographic image
  • Figure 2d Grayscale histogram of the denoised MCT tomographic image
  • Figure 3a Original MCT tomographic image
  • Figure 3c denoised MCT tomographic image
  • Figure 4a Grayscale histogram of the original MCT tomographic image
  • Fig. 4b Gray histogram after smoothing the original MCT tomographic image
  • Figure 4c Curves after smoothing the grayscale histogram of the smoothed MCT tomographic image
  • Figure 4d Grayscale histogram of the denoised MCT tomographic image
  • Figure 5a Original MCT tomographic image
  • Figure 5c denoised MCT tomographic image
  • Figure 6a Grayscale histogram of the original MCT tomographic image
  • Fig. 6b Gray histogram after smoothing the original MCT tomographic image
  • Figure 6c Curve after smoothing the gray histogram of the smoothed MCT tomographic image
  • Figure 6d Grayscale histogram of the denoised MCT tomographic image
  • FIG. 7 is a schematic diagram of the modules of the device of the present invention, wherein, 1. Image scanning step; 2. Data collection module; 3. Data processing module; 4. Data output module; 5. Data display terminal.
  • An intelligent denoising method for MCT tomographic images of leather fibers comprising:
  • the gray value with the smallest frequency is used as the boundary point between the object gray value and the noise gray value, that is, the denoising threshold, and the pixels whose gray value is lower than the denoising threshold are regarded as background noise points , and the remaining points are used as object points;
  • the object gray value frequency histogram and the noise gray value frequency histogram each have a single peak, that is, the gray value with the smallest frequency corresponding to the trough of the gray value frequency histogram in the shape of a double peak is used as the object gray value and the noise gray value
  • the denoising threshold is the denoising threshold.
  • the frequency calculation of the gray value refers to, for each frame of the leather fiber MCT tomographic image, record the maximum value of the gray value as n (n ⁇ 255), and the minimum value of the non-zero gray value is usually 1. Since image denoising does not need to consider pixels with a gray value of 0, it is only necessary to calculate the gray value frequency of the gray value between 1 and n to generate the image gray value frequency sequence x 1 , x 2 ,L, x n .
  • the minimum value point of the gray frequency of the leather fiber MCT tomographic image is used as the denoising threshold, that is, the minimum value of the gray value frequency sequence between the object gray value frequency peak and the noise gray value frequency peak point as the denoising threshold.
  • Described smoothing algorithm has Gaussian filtering method, median filtering method and arithmetic mean filtering method etc.
  • the smoothing algorithm adopted in the present embodiment is arithmetic mean filtering method, specifically:
  • the gray value frequency of some leather fiber MCT tomographic images may have a minimum value before the noise mode. Therefore, in order to explore the boundary point between the noise gray value and the object gray value (i.e. noise gray threshold), only need to explore the minimum value point of the gray value frequency sequence after the noise gray value mode; wherein the reference value of p is 7. If p is too large, the calculation speed will be affected.
  • the selection window width is 5, which can be adjusted for the smoothing effect of the video number histogram in practical applications.
  • the characteristics of this technical solution are: if the number of filtering times p is too small, the random disturbance in the frequency sequence of gray values cannot be effectively eliminated, which will lead to a large deviation of the obtained minimum point; if the number of filtering times is too large, It will make the trough of the gray value frequency distribution disappear, resulting in the absence of the minimum point of the gray value frequency sequence, and the boundary point between the object and the noise gray value cannot be determined.
  • the computer automatically reduces the number of filters and performs the arithmetic mean filtering operation again, and so on until the minimum value of the frequency sequence of gray values is obtained, and this value is used as threshold.
  • the gray value of the pixels whose gray value is less than this threshold is set to 0, so as to complete the image denoising task.
  • a device loaded with the method described in Embodiments 1 and 2 the device includes:
  • Data collection module used to collect leather fiber MCT tomographic images or leather fiber embedded MCT tomographic images
  • Data processing module used to perform a median filter module on the collected scanned image and distinguish background noise points and object points according to the denoising threshold, thereby removing background noise;
  • Data output module used to output the denoised scanned image.
  • An electronic device loaded with the method described in Embodiments 1 and 2, including a memory, a processor, and a computer program stored on the memory and operable on the processor, when the processor executes the program, it realizes the aiming at Intelligent denoising method for MCT tomographic images of leather fibers.
  • a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the intelligent denoising method for leather fiber MCT tomographic images is realized.
  • the early stage image scanning step comprises:
  • MCT imaging equipment MCT tomography scanner: SkyScan2211; camera lens: MX11002;
  • Exposure time 1000ms
  • Image size 4032 pixels ⁇ 4032 pixels
  • Figure 1a Original MCT tomographic image
  • Figure 1b Smoothed MCT tomographic image
  • Figure 1c denoised MCT tomographic image
  • Figure 2a Grayscale histogram of the original MCT tomographic image
  • Figure 2b The grayscale histogram after smoothing the original MCT tomographic image
  • Figure 2c Curve after smoothing the grayscale histogram of the smoothed MCT tomographic image
  • Figure 2d Grayscale histogram of the denoised MCT tomographic image
  • Figure 2c is the gray histogram of the smoothed MCT tomographic image, and the curve is relatively smooth.
  • the early stage image scanning step comprises:
  • MCT tomography scanner SkyScan2211; camera lens: MX11002;
  • Exposure time 2300ms
  • Image size 4032 pixels ⁇ 4032 pixels
  • Figure 3a Original MCT tomographic image
  • Figure 3c denoised MCT tomographic image
  • Figure 4a Grayscale histogram of the original MCT tomographic image
  • Fig. 4b Gray histogram after smoothing the original MCT tomographic image
  • Figure 4c Curves after smoothing the grayscale histogram of the smoothed MCT tomographic image
  • Figure 4d Grayscale histogram of the denoised MCT tomographic image
  • the gray histogram of the original MCT tomographic image shown in Fig. 4a is in the form of a single peak, which cannot distinguish between the object gray value range and the noise gray value range.
  • the gray histogram of the smoothed MCT tomographic image shows a bimodal shape, as shown in Figure 4b. This makes objects and noise easily distinguishable.
  • Figure 4c is the grayscale histogram of the denoised MCT tomographic image with a smooth curve.
  • Figure 3c is the MCT tomographic image after denoising. The denoising effect is ideal, but the suspicious noise has not been removed.
  • the exposure time of the two is different: the exposure time of application example 1 is 1000ms, and the exposure time of application example 2 is 2300ms. Since the exposure time of Application Example 2 is 2.3 times that of Application Example 1, its noise intensity is larger than that of Application Example 1. This directly affects the denoising effect. Of course, other denoising measures can be further taken to improve the denoising effect, but this does not belong to the content to be protected by the present invention.
  • the early stage image scanning step comprises:
  • MCT tomography scanner SkyScan2211; camera lens: MX11002:
  • Exposure time 1000ms
  • Image size 2016 pixels ⁇ 2016 pixels
  • Figure 5a Original MCT tomographic image
  • Figure 5c denoised MCT tomographic image
  • Figure 6a Grayscale histogram of the original MCT tomographic image
  • Fig. 6b Gray histogram after smoothing the original MCT tomographic image
  • Figure 6c Curve after smoothing the gray histogram of the smoothed MCT tomographic image
  • Figure 6d Grayscale histogram of the denoised MCT tomographic image
  • Figure 5b shows that the MCT tomographic images of embedded leather fibers have a large noise intensity, which brings certain difficulties to denoising.
  • the gray histogram of the original MCT tomographic image shown in Figure 6a shows a single-peak shape, which shows that the object gray value range and the noise gray value range are not easy to distinguish.
  • the gray histogram of the smoothed MCT tomographic image shows a bimodal shape, as shown in Figure 6b. This makes objects and noise easily distinguishable.
  • Figure 6c is the grayscale histogram of the denoised MCT tomographic image with a smooth curve.
  • Figure 5c is the MCT tomographic image after denoising, and the denoising effect is ideal.

Landscapes

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

Abstract

一种针对皮革纤维原位MCT断层扫描图像的智能去噪方法，为了更清晰地突出皮革纤维束的边缘，首先通过反复试验确定了图像中值滤波的最佳窗口尺寸和最佳滤波次数，然后计算灰度值频数并根据灰度值频数的极大极小值确定去噪灰度阈值，进而对灰度值直方图进行多次平滑，消除随机扰动。由于对每一帧图像的灰度值频数平滑次数进行人工试验的工作量巨大。本发明设计了计算机自动探索最佳灰度值频数平滑次数，从而自动探测图像去噪的灰度阈值，使得图像去噪运算快速而准确，去噪效果显著。

Description

一种针对皮革纤维MCT断层扫描图像的智能去噪方法及应用 技术领域
本发明公开一种针对皮革纤维MCT断层扫描图像的智能去噪方法、装置、电子设备、可读存储介质及应用，属于皮革图像去噪处理的技术领域。
背景技术
皮革纤维的编织结构一直是业界关注的问题。到目前为止，人们对皮革纤维编织结构的认识还很肤浅。研究皮革纤维的编织结构具有重要的理论价值和应用价值。
采集和分析皮革纤维断层扫描图像并进行三维重构是研究皮革纤维编织结构的有效途径。皮革纤维断层扫描图像的采集方法主要有：X射线断层扫描(CT)法和核磁共振(MRI)法等非破坏性方法以及物理切片法和层去法等破坏性方法。
核磁共振仪器设备造价和维护成本高昂，而破坏性采样方法是对包埋皮革纤维进行物理切片或打磨，工艺要求高、难度大，所采集的图像难以保持皮革纤维的结构，不易采集到高质量的连续断层扫描图像。相比于核磁共振仪器设备，所述显微X射线断层扫描设备造价和维护成本都不高，操作便利，而且，所述显微X射线断层扫描(MCT)技术可在不破坏样本的情况下，获取样本内部结构的断层扫描图像，其图像能够显示尺寸在几十到几百微米的纤维编织结构，可用于采集皮革纤维图像。
利用皮革纤维的MCT断层扫描图像可以进行皮革纤维的三维重构，从而展示皮革纤维的编织结构，以便研究其规律。但是皮革纤维的MCT断层扫描图像带有电子和光学噪声，这些噪声在图像中通常具有低灰度值。皮革纤维的MCT断层扫描图像的去噪和分割是进行三维重构的基础。目前，图像去噪、分割和重构方法很多，但没有一种通用的方法可以处理所有类型的图像。医学领域有众多图像处理方法，也有三维重构软件被开发出来。但到目前为止，医学图像的去噪与分割仍然有很多尚未解决的问题，没有一种方法和一款软件能够很完美地完成医学图像处理任务。
在本技术领域中，张华勇等人采用显微X射线断层扫描(MCT)技术采集皮革纤维断层扫描序列图像并应用医学领域的3D-Doctor软件并辅之以自主研发的计算机图像处理程序进行图像的去噪和三维重构，获得了皮革纤维的三维重构效果。但由于皮革纤维MCT断层扫描图像与医学图像的较大差异，其采用的医学图像处理技术所获得的三维重构图像不能够显示皮革纤维束表面精细的纹理结构。尽管图像去噪的方法有很多，但在应用这些方法进行不同类型的图像处理时，需要对一些参数进行不同的设置，譬如，对图像进行滤波运算所用的窗口宽度和滤波次数、用于去噪的灰度阈值等。不同的参数设置对图像去噪效果的影响是显著的。而参数的选定有时需要进行大量的试验。由于皮革纤维MCT断层扫描图像往往有两千多帧，针对所有图像采用固定统一的参数设置有时效果不好，因而在处理每一帧图像时需要设置不同的参数，这一工作靠人工完成 存在很大困难。
发明内容
针对现有技术的不足，本发明公开一种针对皮革纤维原位MCT断层扫描图像的智能去噪方法。
本发明还公开一种加载有该方法的装置。
本发明还公开一种加载有该方法的电子设备。
本发明还公开利用该方法的应用方法。
发明概述：
本发明主要包括皮革纤维MCT断层扫描图像的中值滤波、灰度值频数序列平滑及灰度值频数极小值探索。为了更清晰地突出皮革纤维束的边缘，首先通过反复试验确定图像中值滤波的最佳窗口尺寸和最佳滤波次数，然后计算灰度值频数并根据灰度值频数的极大极小值确定去噪灰度阈值。由于灰度值直方图通常并不光滑，而是呈现锯齿形波动，其极小值众多而不唯一，因而需要对灰度值直方图进行多次平滑，消除这类随机扰动。如果对灰度值频数直方图进行平滑的次数太少，则不能够有效消除其随机扰动，但如果平滑次数过多，则其波谷也会消失。因而，选取恰当的灰度值频数直方图平滑次数是探索灰度值去噪阈值的关键。由于皮革纤维MCT断层扫描图像的数量很多，并且每一帧图像的灰度值频数分布都不相同，对灰度值频数直方图进行平滑所需要的次数也会不同。对每一帧图像的灰度值频数平滑次数进行人工试验的工作量巨大。为此，本发明设计了计算机自动探索最佳灰度值频数平滑次数，从而确定图像去噪的灰度阈值，使得图像去噪运算快速而准确，去噪效果显著。
本发明所述方法自动探索中值滤波平滑后的MCT图像的对象(即皮革纤维)灰度值与噪声灰度值的分界点，并将该分界点作为去噪阈值进行图像的灰度阈值去噪处理，完成去噪工作。
本发明的详细技术方案如下：
一种针对皮革纤维MCT断层扫描图像的智能去噪方法，其特征在于，包括：
1)对皮革纤维原位显微MCT断层扫描图像进行多次中值滤波；与其它图像平滑方法相比较，中值滤波具有很好的锐化图像边缘的功效，这正是进行皮革纤维MCT断层扫描图像分割所需要的；优选的，所述中值滤波的滤波窗口尺寸为11×11，滤波次数为5～11，本方案中，所述中值滤波能够将噪声强度较大的皮革纤维MCT灰度图像的纤维束和噪声有效地区分开来，使得皮革纤维MCT灰度图像的灰度直方图呈双峰形态：一个波峰代表纤维灰度值众数，一个波峰代表噪声灰度值众数。
2)确定去噪阈值：频数最小的灰度值作为对象灰度值与噪声灰度值的分界点即去噪阈值，灰度值低于所述去噪阈值的像素点被视为背景噪声点，其余点作为对象点。
优选的，对象灰度值频数直方图和噪声灰度值频数直方图各自具有单一波峰，即呈双峰形态的灰度值频数直方图的波谷对应的频数最小的灰度值作为对象灰度值与对象噪声灰度值的分界点即去噪阈值，由于皮革纤维MCT断层扫描图像的固有特点，噪声灰度值频数峰值总是大 于对象灰度值频数峰值，因此，本发明所涉及的去噪算法将噪声灰度值众数和对象灰度值众数之间的灰度值频数的最小值点(即灰度值频数最小的灰度值)作为去噪阈值。
优选的，所述步骤2)中，确定去噪阈值时，利用计算机自动检测经过所述步骤1)所得到的平滑的皮革纤维MCT断层扫描图像的灰度频数极小值点作为所述去噪阈值，即将灰度值频数序列的介于两个灰度值频数波峰之间的极小值作为所述去噪阈值。皮革纤维MCT断层扫描图像的灰度值在0到255之间，灰度值为0的像素点大部分处于MCT设备的视域之外。在技术方案中，皮革纤维与噪声的灰度值的分界点通常是其频数直方图的波谷，因此，探索去噪阈值的工作就等价于探索灰度值频数分布的波谷，即灰度值频数序列的介于两个灰度值频数波峰之间的极小值。
根据本发明优选的，在步骤2)中，要对灰度值频数序列进行多次平滑。由于灰度值直方图通常并不光滑，即常常呈现锯齿形波动，其极小值通常并非波谷，只是随机扰动的局部现象而已，因此需要对灰度值频数序列进行多次平滑。消除这类随机扰动，可避免或减少灰度值最小值点的计算误差。
在数字图像处理领域，常用的图像平滑算法有高斯滤波法、中值滤波法和算术平均值滤波法等众多方法。根据本发明优选的，所述平滑算法为算术平均值滤波法，具体为：
首先选定一个较大的滤波次数p，选择窗宽，运用算术平均值滤波法对灰度值频数序列x 1,x 2,L,x n进行滤波。
然后计算滤波后的灰度值频数序列的最大值点。经验表明，某些皮革纤维MCT断层扫描图像的灰度值频数在噪声众数之前可能存在极小值，因此，为了探索噪声灰度值与对象灰度值的分界点(即去噪灰度阈值)，只需要探索噪声灰度值众数之后的灰度值频数序列的极小值点；其中所述p的参考值为7。p太大的话，影响运算速度。所述选择窗宽为5，在实际应用中，可视频数直方图的平滑效果而调整。本技术方案的特点为：如果滤波次数p太小，则不能够有效消除灰度值频数序列中的随机扰动，这会导致所求得的极小值点偏差较大；如果滤波次数过大，则会使灰度值频数分布的波谷消失，导致灰度值频数序列的极小值点不存在，不能确定对象与噪声灰度值的分界点。因此，如果没有发现灰度值频数序列的极小值，则计算机自动减少滤波次数，重新进行算术平均值滤波运算，如此下去，直到求得灰度值频数序列的极小值，将此值作为阈值。灰度值小于此阈值的像素点的灰度值设置为0，从而完成图像去噪任务。
根据本发明优选的，如果最终未求得灰度值频数序列的极小值，则判定该算法失效，这表明图像的灰度值分布没有波谷。
在本发明中，对每一帧图像的灰度值频数平滑次数进行人工试验的工作量巨大，为此，本发明设计了计算机自动探索最佳灰度值频数平滑次数，从而对每一帧MCT断层扫描图像自动确定图像去噪的灰度阈值，使得图像去噪运算快速而准确。对灰度值频数直方图进行平滑的常用方法有高斯滤波法、中值滤波法和平均值滤波法等。实验表明，算术平均值滤波法效果最佳。
基于理论和实践，在噪声灰度值频数众数和对象灰度值频数众数之间可能存在不止一个 灰度值频数极小值，也可能不存在灰度值频数极小值。因此，本发明所述算法不计算对象众数，只计算噪声灰度值众数之后的频数极小值。然而，经试验表明，如果对灰度值频数直方图进行平滑的次数太少，则不能够有效消除其随机扰动；但如果平滑次数过多，则其波谷也会消失。因而，选取恰当的灰度值频数直方图是确定灰度值去噪阈值的关键。由于皮革纤维MCT断层扫描图像的数量很多，并且每一帧图像的灰度值频数分布都不相同，对灰度值频数直方图进行平滑的次数也会不同，为此，本发明所采用的灰度值频数直方图平滑次数对每一帧图像可能是不同的。
本发明还公开一种加载有该方法的装置，其特征在于，所述装置包括：
数据采集模块：用于采集皮革纤维MCT断层扫描图像或皮革纤维包埋MCT断层扫描图像；
数据处理模块：用于对采集到所述扫描图像进行中值滤波模块和根据去噪阈值区分背景噪声点和对象点，进而去除背景噪声；
数据输出模块：用于将去噪后的扫描图像进行输出。
本发明还公开一种加载有该方法的电子设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，其特征在于，所述处理器执行所述程序时实现所述针对皮革纤维MCT断层扫描图像的智能去噪方法。
本发明还公开一种计算机可读存储介质，其上存储有计算机程序，其特征在于，该计算机程序被处理器执行时实现所述针对皮革纤维MCT断层扫描图像的智能去噪方法。
本发明还公开利用该方法应用于皮革纤维原位MCT断层扫描图像和皮革纤维包埋MCT断层扫描图像的去噪过程。
本发明的技术效果：
1、本发明的主要目的是探索皮革纤维MCT断层扫描图像中皮革纤维影像和背景噪声之间的边界、去除噪声。由于皮革纤维MCT图像中皮革纤维影像和背景噪声之间灰度值差异相对较大，采用中值滤波能够使皮革纤维影像和背景噪声之间的边界更加清晰，因而选择中值滤波对皮革纤维MCT原始图像进行滤波。
2、由于中值滤波的窗口尺寸和滤波次数对滤波效果影响很大，因而选择适当的中值滤波的窗口尺寸和滤波次数至关重要。经过反复测试，本发明建议中值滤波的参考窗口尺寸为11×11，滤波次数为5～11。在实际应用中可对中值滤波的窗口尺寸和滤波次数进行适当调整。
3、为设定去噪阈值，需要计算对象灰度频数与噪声灰度频数的分界值。本发明采用对象图像灰度频数峰值与噪声灰度频数峰值之间的谷值作为去噪阈值。
4、在确定去噪阈值时，由于灰度值直方图通常并不光滑，即常常呈现锯齿形波动，因此需要对灰度值频数序列进行几次平滑，消除这类随机扰动。然而，如果对灰度值频数直方图进行平滑的次数太少，则不能够有效消除其随机扰动，但如果平滑次数过多，则其波谷也会消失。由于皮革纤维MCT断层扫描图像的数量很多，并且每一帧图像的灰度值频数分布都不相同，对灰度值频数直方图进行平滑的次数也会不同。因此，对每一帧图像的灰度值频数平滑次数进行人工试验的工作量巨大。为此，本发明采用智能滤波算法自动探索最佳灰度值频数平滑次数、从而自动 确定图像去噪的灰度阈值。这一算法简单、快速、准确，且完全智能化，去噪效果显著。
5、本发明所述方法适用于皮革纤维原位MCT断层扫描图像和皮革纤维包埋MCT断层扫描图像的去噪场景。
附图说明
图1a：原始MCT断层扫描图像；
图1b：平滑的MCT断层扫描图像；
图1c：去噪的MCT断层扫描图像；
图2a：原始MCT断层扫描图像的灰度直方图；
图2b：对原始MCT断层扫描图像进行平滑后的灰度直方图；
图2c：对平滑MCT断层扫描图像的灰度直方图进行平滑后的曲线；
图2d：去噪的MCT断层扫描图像的灰度直方图；
图3a：原始MCT断层扫描图像；
图3b：平滑的MCT断层扫描图像；
图3c：去噪的MCT断层扫描图像；
图4a：原始MCT断层扫描图像的灰度直方图；
图4b：对原始MCT断层扫描图像进行平滑后的灰度直方图；
图4c：对平滑MCT断层扫描图像灰度直方图进行平滑后的曲线；
图4d：去噪的MCT断层扫描图像的灰度直方图；
图5a：原始MCT断层扫描图像；
图5b：平滑的MCT断层扫描图像；
图5c：去噪的MCT断层扫描图像；
图6a：原始MCT断层扫描图像的灰度直方图；
图6b：对原始MCT断层扫描图像进行平滑后的灰度直方图；
图6c：对平滑MCT断层扫描图像灰度直方图进行平滑后的曲线；
图6d：去噪的MCT断层扫描图像的灰度直方图；
图7是本发明所述装置的模块示意图，其中，1、图像扫描步骤；2、数据采集模块；3、数据处理模块；4、数据输出模块；5、数据展示终端。
具体实施方式
下面结合实施例和说明书附图对本发明做详细的说明，但不限于此。
实施例1、
一种针对皮革纤维MCT断层扫描图像的智能去噪方法，包括：
1)对皮革纤维原位MCT断层图像进行多次中值滤波；所述中值滤波的滤波窗口尺寸为11×11，滤波次数为5～11；
2)确定去噪阈值：频数最小的灰度值作为对象灰度值与噪声灰度值的分界点即去噪阈值， 灰度值低于所述去噪阈值的像素点被视为背景噪声点，其余点作为对象点；
对象灰度值频数直方图和噪声灰度值频数直方图各自具有单一波峰，即呈双峰形态的灰度值频数直方图的波谷对应的频数最小的灰度值作为对象灰度值与噪声灰度值的分界点即去噪阈值。
其中，灰度值频数计算是指，对每一帧皮革纤维MCT断层扫描图像，记其灰度值的最大值为n(n≤255)，而非零灰度值的最小值通常为1。由于图像去噪无需考虑灰度值为0的像素点，所以只需计算灰度值在1～n之间的灰度值频数，生成图像的灰度值频数序列x 1,x 2,L,x n。
实施例2、
如实施例1所述的一种针对皮革纤维MCT断层扫描图像的智能去噪方法，其区别在于，所述步骤2)中，确定去噪阈值时，还可以利用计算机自动检测所述步骤1)中皮革纤维MCT断层扫描图像的灰度频数极小值点作为所述去噪阈值，即将灰度值频数序列介于对象灰度值频数波峰和噪声灰度值频数波峰之间的的极小值点作为所述去噪阈值。
对灰度值频数序列进行多次平滑。
所述平滑算法有高斯滤波法、中值滤波法和算术平均值滤波法等，本实施例中所采用的平滑算法为算术平均值滤波法，具体为：
首先选定一个较大的滤波次数p，选择窗宽，运用算术平均值滤波法对灰度值频数序列x 1,x 2,L,x n进行滤波。然后计算滤波后的灰度值频数序列的最大值点。本实施例中试验表明，某些皮革纤维MCT断层扫描图像的灰度值频数在噪声众数之前可能存在极小值，因此，为了探索噪声灰度值与对象灰度值的分界点(即去噪灰度阈值)，只需要探索噪声灰度值众数之后的灰度值频数序列的极小值点；其中所述p的参考值为7。p太大的话，影响运算速度。所述选择窗宽为5，在实际应用中，可视频数直方图的平滑效果而调整。本技术方案的特点为：如果滤波次数p太小，则不能够有效消除灰度值频数序列中的随机扰动，这会导致所求得的极小值点偏差较大；如果滤波次数过大，则会使灰度值频数分布的波谷消失，导致灰度值频数序列的极小值点不存在，不能确定对象与噪声灰度值的分界点。因此，如果没有发现灰度值频数序列的极小值，则计算机自动减少滤波次数，重新进行算术平均值滤波运算，如此下去，直到求得灰度值频数序列的极小值，将此值作为阈值。灰度值小于此阈值的像素点的灰度值设置为0，从而完成图像去噪任务。
如果最终未求得灰度值频数序列的极小值，则判定该算法失效，这表明图像的灰度值分布没有波谷。
本实施例中，经试验表明，如果对灰度值频数直方图进行平滑的次数太少，则不能够有效消除其随机扰动，但如果平滑次数过多，则其波谷也会消失。因而，选取恰当的灰度值频数直方图是确定灰度值去噪阈值的关键。由于皮革纤维MCT断层扫描图像的数量很多，并且每一帧图像的灰度值频数分布都不相同，对灰度值频数直方图进行平滑的次数也会不同，为此，本发明采用特殊的滤波算法进行处理。
实施例3、
如图7所示，一种加载有如实施例1、2所述方法的装置，所述装置包括：
数据采集模块：用于采集皮革纤维MCT断层扫描图像或皮革纤维包埋MCT断层扫描图像；
数据处理模块：用于对采集到所述扫描图像进行中值滤波模块和根据去噪阈值区分背景噪声点和对象点，进而去除背景噪声；
数据输出模块：用于将去噪后的扫描图像进行输出。
实施例4、
一种加载有如实施例1、2所述方法的电子设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行所述程序时实现所述针对皮革纤维MCT断层扫描图像的智能去噪方法。
实施例5、
一种计算机可读存储介质，其上存储有计算机程序，该计算机程序被处理器执行时实现所述针对皮革纤维MCT断层扫描图像的智能去噪方法。
实施例6、
利用如实施例1、2所述方法应用于皮革纤维原位MCT断层扫描图像和皮革纤维包埋MCT断层扫描图像的去噪过程。
应用例1、
利用实施例2所述方法对以下皮样材料进行图像处理，前期图像扫描步骤包括：
1)皮样材料：从美国牛臀干蓝皮上剥离出来的一部分纤维组织；
2)MCT取像设备：MCT断层扫描仪：SkyScan2211；摄像镜头：MX11002；
3)MCT断层扫描图像参数：
曝光时间：1000ms；
图像尺寸：4032像素×4032像素；
层厚：0.31μm；
帧数：2357；
4)原始MCT断层扫描图像平均值滤波参数设置：窗口尺寸11×11，滤波重复次数7；
5)灰度频数序列平滑窗宽：5。
经扫描、处理、输出后产生的图像数据如图1a-1c。
图1a：原始MCT断层扫描图像；
图1b：平滑的MCT断层扫描图像；
图1c：去噪的MCT断层扫描图像；
图2a：原始MCT断层扫描图像的灰度直方图；
图2b：对原始MCT断层扫描图像进行平滑后的灰度直方图；
图2c：对平滑MCT断层扫描图像的灰度直方图进行平滑后的曲线；
图2d：去噪的MCT断层扫描图像的灰度直方图；
由图2a和图2b可以看出，平滑的MCT断层扫描图像的噪声灰度值范围变窄，这使得对象和噪声更容易区分。图2c是平滑后的MCT断层扫描图像的灰度直方图，曲线比较平滑。
由图2a和图2b还可以看到，平滑后的MCT断层扫描图像中的噪声灰度值变小，这有利于后续确定去噪阈值并进行图像去噪。
由图1c可以看到，去噪效果比较理想。图像存在一些小面积区域，这些区域有可能是纤维束截面，也有可能是噪声，这在进行三维重构时可以进行鉴别。
应用例2、
利用实施例2所述方法对以下皮样材料进行图像处理，前期图像扫描步骤包括：
1)皮样材料：美国牛臀干蓝皮；
2)MCT取像设备：
MCT断层扫描仪：SkyScan2211；摄像镜头：MX11002；
3)MCT断层扫描图像参数：
曝光时间：2300ms；
图像尺寸：4032像素×4032像素；
层厚：0.20μm；
帧数：2357；
4)原始MCT断层扫描图像平均值滤波参数设置：窗口尺寸11×11，滤波次数7；
5)灰度频数序列平滑窗宽：5；
经扫描、处理、输出后产生的图像数据如图3a-3c。
图3a：原始MCT断层扫描图像；
图3b：平滑的MCT断层扫描图像；
图3c：去噪的MCT断层扫描图像；
图4a：原始MCT断层扫描图像的灰度直方图；
图4b：对原始MCT断层扫描图像进行平滑后的灰度直方图；
图4c：对平滑MCT断层扫描图像灰度直方图进行平滑后的曲线；
图4d：去噪的MCT断层扫描图像的灰度直方图；
图4a所示的原始MCT断层扫描图像的灰度直方图呈单峰形态，不能区分对象灰度值范围和噪声灰度值范围。而经过平滑的MCT断层扫描图像灰度直方图呈双峰形态，如图4b所示。这使得对象和噪声容易区分。图4c是去噪后的MCT断层扫描图像的灰度直方图，曲线平滑。图3c是去噪后的MCT断层扫描图像，去噪效果比较理想，但存在可疑的噪声没有去除。
将应用例1、2进行比较：两者的曝光时间不同：应用例1的曝光时间为1000ms，应用例2的曝光时间为2300ms。由于应用例2的曝光时间是应用例1的曝光时间的2.3 倍，导致其噪声强度比应用例1的大。这直接影响了去噪效果。当然，还可以进一步采取其它去噪措施改进去噪效果，但不属于本发明所要保护的内容。
应用例3、
利用实施例2、6所述方法及应用对以下皮样材料进行图像处理，前期图像扫描步骤包括：
1)皮样材料：环氧树脂包埋牛臀干蓝皮；
2)MCT取像设备：
MCT断层扫描仪：SkyScan2211；摄像镜头：MX11002：
3)MCT断层扫描图像参数：
曝光时间：1000ms；
图像尺寸：2016像素×2016像素；
层厚：0.80μm；
帧数：1178；
4)原始MCT断层扫描图像平均值滤波参数设置：窗口尺寸11×11，滤波次数7；
5)灰度频数序列平滑窗宽：5；
经扫描后、处理、输出后产生的图像数据如图5a-5c。
图5a：原始MCT断层扫描图像；
图5b：平滑的MCT断层扫描图像；
图5c：去噪的MCT断层扫描图像；
图6a：原始MCT断层扫描图像的灰度直方图；
图6b：对原始MCT断层扫描图像进行平滑后的灰度直方图；
图6c：对平滑MCT断层扫描图像灰度直方图进行平滑后的曲线；
图6d：去噪的MCT断层扫描图像的灰度直方图；
图5b显示，包埋皮革纤维MCT断层扫描图像的噪声强度很大，这给去噪带来了一定的困难。图6a所示的原始MCT断层扫描图像的灰度直方图呈单峰形态，这说明对象灰度值范围和噪声灰度值范围不易区分。而经过平滑的MCT断层扫描图像灰度直方图呈双峰形态，如图6b所示。这使得对象和噪声容易区分。图6c是去噪后的MCT断层扫描图像的灰度直方图，曲线平滑。图5c是去噪后的MCT断层扫描图像，去噪效果比较理想。

Claims (10)

  1. 一种针对皮革纤维MCT断层扫描图像的智能去噪方法，其特征在于，包括：
    1)对皮革纤维原位显微MCT断层扫描图像进行多次中值滤波；与其它图像平滑方法相比较，中值滤波具有很好的锐化图像边缘的功效；
    2)确定去噪阈值：频数最小的灰度值作为对象灰度值与噪声灰度值的分界点即去噪阈值，灰度值低于所述去噪阈值的像素点被视为背景噪声点，其余点作为对象点。
  2. 根据权利要求1所述的一种针对皮革纤维MCT断层扫描图像的智能去噪方法，其特征在于，对象灰度值频数直方图和噪声灰度值频数直方图各自具有单一波峰，即呈双峰形态的灰度值频数直方图的波谷对应的频数最小的灰度值作为对象灰度值与对象噪声灰度值的分界点即去噪阈值；优选的，所述中值滤波的滤波窗口尺寸为11×11，滤波次数为5～11。
  3. 根据权利要求1所述的一种针对皮革纤维MCT断层扫描图像的智能去噪方法，其特征在于，所述步骤2)中，确定去噪阈值时，利用计算机自动检测经过所述步骤1)所得到的平滑的皮革纤维MCT断层扫描图像的灰度频数极小值点作为所述去噪阈值，即将灰度值频数序列的介于两个灰度值频数波峰之间的极小值作为所述去噪阈值。
  4. 根据权利要求3所述的一种针对皮革纤维MCT断层扫描图像的智能去噪方法，其特征在于，在步骤2)中，要对灰度值频数序列进行多次平滑。
  5. 根据权利要求4所述的一种针对皮革纤维MCT断层扫描图像的智能去噪方法，其特征在于，所述平滑算法为算术平均值滤波法，具体为：
    首先选定一个较大的滤波次数p，选择窗宽，运用算术平均值滤波法对灰度值频数序列x 1,x 2,L,x n进行滤波；
    然后计算滤波后的灰度值频数序列的最大值点；优选的，选择窗宽为5。
  6. 根据权利要求5所述的一种针对皮革纤维MCT断层扫描图像的智能去噪方法，其特征在于，如果最终未求得灰度值频数序列的极小值，则判定该算法失效。
  7. 一种加载有如权利要求1-5任意一项所述方法的装置，其特征在于，所述装置包括：
    数据采集模块：用于采集皮革纤维MCT断层扫描图像或皮革纤维包埋MCT断层扫描图像；
    数据处理模块：用于对采集到所述扫描图像进行中值滤波模块和根据去噪阈值区分背景噪声点和对象点，进而去除背景噪声；
    数据输出模块：用于将去噪后的扫描图像进行输出。
  8. 一种加载有如权利要求1-6任意一项所述方法的电子设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，其特征在于，所述处理器执行所述程序时实现所述针对皮革纤维MCT断层扫描图像的智能去噪方法。
  9. 一种计算机可读存储介质，其上存储有如权利要求1-6任意一项所述计算机程序，其特征在于，该计算机程序被处理器执行时实现所述针对皮革纤维MCT断层扫描图像的智能去噪方法。
  10. 利用如权利要求1-6任意一项所述方法应用：用于皮革纤维原位MCT断层扫描图像 和皮革纤维包埋MCT断层扫描图像的去噪过程。
PCT/CN2021/120005 2021-05-31 2021-09-23 一种针对皮革纤维mct断层扫描图像的智能去噪方法及应用 WO2022252442A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110602207.6 2021-05-31
CN202110602207.6A CN113313647A (zh) 2021-05-31 2021-05-31 一种针对皮革纤维mct断层扫描图像的智能去噪方法及应用

Publications (1)

Publication Number Publication Date
WO2022252442A1 true WO2022252442A1 (zh) 2022-12-08

Family

ID=77376673

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/120005 WO2022252442A1 (zh) 2021-05-31 2021-09-23 一种针对皮革纤维mct断层扫描图像的智能去噪方法及应用

Country Status (2)

Country Link
CN (1) CN113313647A (zh)
WO (1) WO2022252442A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116342610A (zh) * 2023-05-31 2023-06-27 山东恒海钢结构有限公司 一种钢结构装配式建筑焊接异常检测方法
CN116385438A (zh) * 2023-06-05 2023-07-04 济南科汛智能科技有限公司 一种核磁共振肿瘤区域提取方法
CN117455800A (zh) * 2023-12-22 2024-01-26 肥城新查庄地质勘查有限公司 基于计算机视觉的煤矿钻探定位方法

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113313647A (zh) * 2021-05-31 2021-08-27 齐鲁工业大学 一种针对皮革纤维mct断层扫描图像的智能去噪方法及应用

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060093194A1 (en) * 2004-11-02 2006-05-04 Leo Grady System and method for filtering noise from a medical image
CN101403610A (zh) * 2008-11-06 2009-04-08 陕西科技大学 一种基于数字图像法测量皮革面积的系统及其方法
CN106780464A (zh) * 2016-12-15 2017-05-31 东华大学 一种基于改进阈值分割的织物疵点检测方法
CN107169953A (zh) * 2017-04-07 2017-09-15 西安电子科技大学 基于hog特征的桥梁混凝土表面裂缝检测方法
CN113313647A (zh) * 2021-05-31 2021-08-27 齐鲁工业大学 一种针对皮革纤维mct断层扫描图像的智能去噪方法及应用

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107967690B (zh) * 2016-10-19 2020-06-09 中国石油天然气股份有限公司 一种自适应铁谱磨粒图像二值化处理方法
CN111260616A (zh) * 2020-01-13 2020-06-09 三峡大学 一种基于Canny算子二维阈值分割优化的绝缘子裂纹检测方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060093194A1 (en) * 2004-11-02 2006-05-04 Leo Grady System and method for filtering noise from a medical image
CN101403610A (zh) * 2008-11-06 2009-04-08 陕西科技大学 一种基于数字图像法测量皮革面积的系统及其方法
CN106780464A (zh) * 2016-12-15 2017-05-31 东华大学 一种基于改进阈值分割的织物疵点检测方法
CN107169953A (zh) * 2017-04-07 2017-09-15 西安电子科技大学 基于hog特征的桥梁混凝土表面裂缝检测方法
CN113313647A (zh) * 2021-05-31 2021-08-27 齐鲁工业大学 一种针对皮革纤维mct断层扫描图像的智能去噪方法及应用

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116342610A (zh) * 2023-05-31 2023-06-27 山东恒海钢结构有限公司 一种钢结构装配式建筑焊接异常检测方法
CN116342610B (zh) * 2023-05-31 2023-08-15 山东恒海钢结构有限公司 一种钢结构装配式建筑焊接异常检测方法
CN116385438A (zh) * 2023-06-05 2023-07-04 济南科汛智能科技有限公司 一种核磁共振肿瘤区域提取方法
CN116385438B (zh) * 2023-06-05 2023-08-11 济南科汛智能科技有限公司 一种核磁共振肿瘤区域提取方法
CN117455800A (zh) * 2023-12-22 2024-01-26 肥城新查庄地质勘查有限公司 基于计算机视觉的煤矿钻探定位方法
CN117455800B (zh) * 2023-12-22 2024-03-15 肥城新查庄地质勘查有限公司 基于计算机视觉的煤矿钻探定位方法

Also Published As

Publication number Publication date
CN113313647A (zh) 2021-08-27

Similar Documents

Publication Publication Date Title
WO2022252442A1 (zh) 一种针对皮革纤维mct断层扫描图像的智能去噪方法及应用
Nikolic et al. Edge detection in medical ultrasound images using adjusted Canny edge detection algorithm
EP2238742B1 (en) Noise reduction of images
Kanwal et al. Region based adaptive contrast enhancement of medical X-ray images
Muduli et al. A novel technique for wall crack detection using image fusion
WO2017080847A1 (en) Method for automatic optimization of quantitative map generation in functional medical imaging
CN112819739B (zh) 一种扫描电子显微镜图像处理方法和系统
Patel et al. A review on image contrast enhancement techniques
CN116342444A (zh) 一种双通道多模态图像融合方法及融合成像终端机
Potočnik et al. Automated computer-assisted detection of follicles in ultrasound images of ovary
Renuka et al. Adaptive shrinkage on dual-tree complex wavelet transform for denoising real-time MR images
CN107478656B (zh) 基于机器视觉的纸浆搅拌效果检测评价方法、装置、系统
Picard et al. Automated crack detection method applied to CT images of baked carbon anode
CN114066881A (zh) 一种基于非线性变换检测方法、计算机设备及存储介质
Gupta et al. A new computational approach for edge-preserving image decomposition
Hamed et al. A proposed model for denoising breast mammogram images
Maurya et al. Morphology based brain tumor identification and segmentation in MR images
Zhang et al. Detection of foreign fibers in cotton on the basis of wavelet
Yahaghi Using total variation denoising for detecting defects in industrial radiography
Kadir et al. Automatic left ventricle segmentation in T2 weighted CMR images
Martinez-Garcia et al. Image Processing Approach for Detection of Three-Dimensional Tree-Rings from X-Ray Computed Tomography
Muthuswamy New Edge Preserving Hybrid Method for Better Enhancement of Liver CT Images
Zhang et al. A fast learning-based super-resolution method for copper strip defect image
Yin et al. CBCT image denoising based on multi-scale wavelet transform
Aguilera et al. Simulated annealing: A novel application of image processing in the wood area

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: 21943796

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: 21943796

Country of ref document: EP

Kind code of ref document: A1