WO2014183246A1 - 一种医学影像处理方法与系统 - Google Patents

一种医学影像处理方法与系统 Download PDF

Info

Publication number
WO2014183246A1
WO2014183246A1 PCT/CN2013/075527 CN2013075527W WO2014183246A1 WO 2014183246 A1 WO2014183246 A1 WO 2014183246A1 CN 2013075527 W CN2013075527 W CN 2013075527W WO 2014183246 A1 WO2014183246 A1 WO 2014183246A1
Authority
WO
WIPO (PCT)
Prior art keywords
pixel
medical image
pixels
spectral intensity
stack
Prior art date
Application number
PCT/CN2013/075527
Other languages
English (en)
French (fr)
Inventor
黄勃
Original Assignee
Huang Bo
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 Huang Bo filed Critical Huang Bo
Priority to CN201380003067.4A priority Critical patent/CN103946868A/zh
Priority to PCT/CN2013/075527 priority patent/WO2014183246A1/zh
Publication of WO2014183246A1 publication Critical patent/WO2014183246A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • 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
    • G06T2207/20092—Interactive image processing based on input by user
    • G06T2207/20101—Interactive definition of point of interest, landmark or seed
    • 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
    • G06T2207/30004—Biomedical image processing

Definitions

  • the object of the present invention is to provide a medical image processing method which can solve the problem of extracting the tongue portion of the same spectral intensity value in the digital image of the tongue in computer tongue imaging.
  • the object of the present invention is to provide a medical image processing method which can solve the problem of extracting the tongue portion of the same spectral intensity value in the digital image of the tongue in computer tongue imaging.
  • a medical image processing method comprising the steps of:
  • one or more pixels in the area required for the digital image of the tongue are randomly selected manually;
  • a common area after each pixel is subjected to pixel expansion processing is extracted as the selected area.
  • the pixel expansion process comprises the following steps: each artificially randomly selected pixel is pushed onto the stack,
  • the top pixel of the stack is popped
  • the flag of the popped pixel is identified as "selected"
  • the neighboring pixels of the popped pixel are inspected in order. If one of the pixels satisfies the boundary condition and the flag is not identified as "selected”, the pixel is pushed onto the stack and the pixel is identified as " “Selected”, otherwise continue to find the next neighborhood pixel until the neighboring pixels of the popped pixel have been checked;
  • step 1 if the stack is not empty, repeat step 1, if the stack is empty, all the pixels whose flag is marked as "selected" are the areas after pixel expansion processing.
  • the boundary condition is that a sum of squares of the difference between the spectral intensity value of the popped pixel and the spectral intensity value of the neighboring pixel is less than a boundary threshold, that is,
  • P'h, Ph, and Thresh are the spectral intensity values of the pop-up pixels of the medical image in the spatial h component of the HSV spectral intensity value, the spectral intensity values of the neighboring pixels, and the corresponding boundary thresholds, among which P's, Ps, and Thress
  • the boundary threshold Thresx (x is the h component or the s component or the v component of the medical image) is calculated as follows:
  • Step a obtaining a gray level Lv of a maximum value of the relative entropy of the full image of the medical image component map, and a value Hv of the histogram of the medical image component map corresponding to the Lv, wherein the relative entropy is defined as
  • P1, p2, ..., pn and q1, q2, ..., qn are probability distributions of tongue coating and tongue pixels
  • Step b respectively calculating the gray level La, Lb at the left and right peaks of the histogram of the medical image component map, and the values Ha, Hb of the histogram of the medical image component map corresponding to La, Lb;
  • Step c if (Ha-Hv)/(Lv-La)>(Hb-Hv)/(Lb-Lv),
  • Thresx E*(Hb-Hv)/(Lb-Lv),
  • Thresx E*(Ha-Hv)/(Lv-La).
  • Another object of the embodiments of the present invention is to provide a medical image processing system, including:
  • a multi-pixel selection module for manually selecting one or more pixels in a region required for a digital image of a tongue
  • a pixel expansion processing module configured to perform pixel expansion processing on each selected pixel
  • the common area extraction module is configured to extract a common area after each pixel is subjected to pixel expansion processing as the selected area.
  • the pixel expansion processing module includes:
  • a top-of-stack pixel pop-up module for popping the top pixel of the stack
  • a pop-up pixel identification module configured to identify a flag of the pop-up pixel as “selected”
  • a neighboring pixel processing module configured to sequentially check the neighboring pixels of the popped pixel, if one of the pixels satisfies the boundary condition and the flag is not identified as “selected”, the pixel is pushed onto the stack, and the pixel is The pixel identifier is "selected”, otherwise the next neighborhood pixel is continued to be searched until the neighboring pixels of the popped pixel have been checked;
  • the stack non-empty judgment module is configured to determine whether the stack is empty. If the stack is not empty, jump to the top pixel pop-up block module, and if the stack is empty, jump to the end module;
  • the end module is used to identify all the flag bits as "selected" pixels as the area after pixel expansion processing.
  • the boundary condition is that a sum of squares of the difference between the spectral intensity value of the popped pixel and the spectral intensity value of the neighboring pixel is less than a boundary threshold, that is,
  • P'h, Ph, and Thresh are the spectral intensity values of the pop-up pixels of the medical image in the spatial h component of the HSV spectral intensity value, the spectral intensity values of the neighboring pixels, and the corresponding boundary thresholds, among which P's, Ps, and Thress
  • the boundary threshold Thresx (x is the h component or the s component or the v component of the medical image) is calculated as follows:
  • Step a obtaining a gray level Lv of a maximum value of the relative entropy of the full image of the medical image component map, and a value Hv of the histogram of the medical image component map corresponding to the Lv, wherein the relative entropy is defined as
  • P1, p2, ..., pn and q1, q2, ..., qn are probability distributions of tongue coating and tongue pixels
  • Step b respectively calculating the gray level La, Lb at the left and right peaks of the histogram of the medical image component map, and the values Ha, Hb of the histogram of the medical image component map corresponding to La, Lb;
  • Step c if (Ha-Hv)/(Lv-La)>(Hb-Hv)/(Lb-Lv),
  • Thresx E*(Hb-Hv)/(Lb-Lv),
  • Thresx E*(Ha-Hv)/(Lv-La).
  • the medical image processing method can accurately extract the tongue coating of the same spectral intensity value in the computer tongue imaging, which is beneficial for subsequent processing.
  • FIG. 1 is a schematic block diagram of a medical image processing method of the present invention
  • FIG. 2 is a schematic block diagram of a pixel expansion processing step of the medical image processing method of the present invention.
  • FIG. 3 is a schematic block diagram of a boundary threshold calculation method of a medical image processing method according to the present invention.
  • FIG. 4 is a schematic structural view of a medical image processing system of the present invention.
  • FIG. 5 is a schematic structural diagram of a pixel expansion processing module of a medical image processing system according to the present invention.
  • the method of extracting the tongue portion of the same spectral intensity value is actually equivalent to performing a binary image segmentation.
  • the so-called binary image means that the gray value of all points on the image uses only two possibilities. If it is not “0", it is "255", that is, the whole image shows a clear black and white effect.
  • the threshold segmentation technique is generally adopted, which is particularly effective for segmenting images with strong contrast between the object and the background. It is simple to calculate and can always define non-overlapping regions with closed, connected boundaries. All pixels whose gray level is greater than or equal to the threshold are judged to belong to the object, and the gray value is represented by "255", otherwise these pixels are excluded from the object area, and the gray value is "0", indicating the background.
  • the boundary of the object becomes a collection of such internal points, at least one of which has no belonging to the object.
  • the threshold method can be used to get better results. If the difference between the object and the background is not in the gray value (such as the texture), this property can be converted to the difference in grayscale, and then the image is segmented using a thresholding technique. In order to make the segmentation more robust and more applicable, the system should be able to automatically select the threshold.
  • Image segmentation algorithms based on knowledge of objects, environments, and application domains are more versatile and adaptable than algorithms based on fixed thresholds. This knowledge includes: the grayscale characteristics of the image corresponding to the object, the size of the object, the proportion of the object in the image, and the number of different types of objects in the image.
  • the tongue coating is often distributed with a variety of spectral intensity values of the tongue coating, and the boundaries between different tongue coatings are very blurred and difficult to distinguish clearly. Therefore, we consider the results of multiple binary image segmentation. Taking the common area as the extraction result, the result is more accurate, avoiding introducing unnecessary noise to the subsequent steps.
  • FIG. 1 an implementation process of a medical image processing method according to an embodiment of the present invention is described in detail as follows:
  • the first step one or more pixels in the required area of the digital image of the tongue are randomly selected by hand; the tongue coating is often distributed with a plurality of tongues of spectral intensity values, and the boundaries between different tongue coatings are very blurred, and it is difficult to clearly distinguish them.
  • the subsequent regions extracted by the second and third steps are different, but the common part is basically stable and is also the area to be extracted. Therefore, the first step needs to manually select the digital image of the tongue. Multiple pixels in the area are required for later steps.
  • the pixel expansion processing comprises the following steps: each artificially randomly selected pixel is pushed onto the stack,
  • the top pixel of the stack is popped
  • the flag of the popped pixel is identified as "selected"
  • the neighboring pixels of the popped pixel are inspected in order. If one of the pixels satisfies the boundary condition and the flag is not identified as "selected”, the pixel is pushed onto the stack and the pixel is identified as " “Selected”, otherwise continue to find the next neighborhood pixel until the neighboring pixels of the popped pixel have been checked;
  • step 1 if the stack is not empty, repeat step 1, if the stack is empty, all the pixels whose flag is marked as "selected" are the areas after pixel expansion processing.
  • Such a method reduces the stacking operation of the traditional method and improves the efficiency of the method.
  • the boundary condition is that the sum of the squares of the spectral intensity values of the popped pixels and the spectral intensity values of the neighboring pixels is smaller than the boundary threshold, that is,
  • P'h, Ph, and Thresh are the spectral intensity values of the pop-up pixels of the medical image in the spatial h component of the HSV spectral intensity value, the spectral intensity values of the neighboring pixels, and the corresponding boundary thresholds, among which P's, Ps, and Thress
  • the HSV model was founded in 1978 by Elvi Ray Smith, a nonlinear transformation of the three primary color modes.
  • Hue (H) is the basic property of color, which is commonly referred to as the color name, such as red, yellow, and so on.
  • Saturation (S) refers to the purity of the color. The higher the color, the softer the color, the lower the color is gradually grayed out, taking a value of 0-100%.
  • Brightness (V), take 0-100%.
  • HSV The color is described in a point in the cylindrical coordinate system.
  • the central axis of the cylinder takes the value from the black at the bottom to the white at the top and the gray at the middle.
  • the angle around this axis corresponds to the "hue” to the axis.
  • the distance corresponds to "saturation” and the height along this axis corresponds to "brightness", "hue” or "lightness”.
  • HSV Models are commonly used in computer graphics applications. HSV is often used in a variety of applications where the user must select a color to apply to a particular graphic element. Color wheel. In it, the hue is represented as a circle; a separate triangle can be used to represent saturation and lightness. Typically, the vertical axis of this triangle indicates saturation and the horizontal axis indicates brightness. In this way, selecting a color can first select the hue in the circle, and select the desired saturation and lightness from the triangle.
  • HSV Another visual method of the model is the cone.
  • the hue is expressed as the angle around the central axis of the cone
  • the saturation is expressed as the distance from the center of the cross section of the cone to this point
  • the brightness is expressed as the distance from the center of the cross section of the cone to the apex .
  • the color space can also be expressed as a cylinder similar to the cone described above, the hue varies along the outer circumference of the cylinder, the saturation varies along the distance from the center of the cross section, and the distance of the brightness along the cross section to the bottom surface and the top surface And change.
  • This representation may be considered A more accurate mathematical model of the HSV color space; however, the number of levels of saturation and hue that can be distinguished in practice decreases as the brightness approaches black.
  • computers typically store RGB with a limited precision range. Value; this constrains the precision, plus the limitations of human color perception, making the cone representation more practical in most cases.
  • HSV color space conforms to the human eye's perception of color.
  • the sensitivity of human vision to brightness is much stronger than the sensitivity of color shade. Therefore, HSV color space is more suitable for human visual characteristics than RGB color space, which is convenient for color processing and recognition. ; ⁇
  • the three coordinates in the HSV color space are independent, that is, people can independently perceive changes in each color component
  • the HSV color space is a uniform color model with a linear scale, the perceived distance between the colors and the HSV The Euclidean distance of the point on the color space coordinates is proportional.
  • the threshold segmentation method is divided into global threshold method and local threshold segmentation method.
  • the so-called local threshold segmentation method divides the original image into smaller images and selects corresponding thresholds for each sub-image. After threshold segmentation, grayscale discontinuities may occur at the boundary between adjacent sub-images, so smoothing techniques are needed to eliminate them.
  • Common methods of local threshold method are gray-scale difference histogram method and differential histogram method.
  • each image is arbitrary. If a sub-image falls on the target area or background area and is segmented according to statistical results, it may produce worse results.
  • the local threshold method should be used for statistics on each sub-image, which is slow and difficult to adapt to real-time requirements.
  • the global threshold segmentation method is applied more in image processing, and it uses a fixed threshold to segment images within the entire image.
  • the classic threshold selection is handled by the gray histogram.
  • different threshold selection methods it can be divided into modal methods, iterative threshold selection and other methods. These methods use the histogram of the image as the research object to determine the threshold of the segmentation.
  • inter-class variance threshold segmentation method two-dimensional maximum entropy segmentation method, fuzzy threshold segmentation method, co-occurrence matrix segmentation method, region growing method and so on.
  • the segmentation threshold may select a gradation value corresponding to a valley between two peaks of the histogram as a segmentation threshold.
  • the image is divided into two parts: the tongue coating area and the tongue area.
  • the sum of the relative entropies of the tongue and the tongue area is obtained for each gray level in the image.
  • the relative entropy and the maximum gray level of the tongue coating area and the tongue area are used as the threshold values of the segmentation image.
  • the gray level of the pixel and the gray level of the neighborhood mean are not different.
  • the corresponding two-dimensional histogram It is mainly concentrated near the diagonal of the i, j plane, and generally presents a state of bimodal and valley, and the two peaks correspond to the target and the background, respectively.
  • the disadvantage of the conventional method is that only the gray scale information of the pixel is considered, and the spatial information of the pixel is not taken into consideration, so the segmentation effect is not ideal when the signal to noise ratio of the image is lowered.
  • Thresx x is the h component or s component or v component of the medical image
  • Step a obtaining a gray level Lv of a maximum value of the relative entropy of the full image of the medical image component map, and a value Hv of the histogram of the medical image component map corresponding to the Lv, wherein the relative entropy is defined as
  • P1, p2, ..., pn and q1, q2, ..., qn are probability distributions of tongue coating and tongue pixels
  • Step b respectively calculating the gray level La, Lb at the left and right peaks of the histogram of the medical image component map, and the values Ha, Hb of the histogram of the medical image component map corresponding to La, Lb;
  • the left and right peaks of the histogram correspond to the tongue and tongue types, respectively.
  • Step c if (Ha-Hv)/(Lv-La)>(Hb-Hv)/(Lb-Lv),
  • Thresx E*(Hb-Hv)/(Lb-Lv),
  • Thresx E*(Ha-Hv)/(Lv-La).
  • E is an empirical constant.
  • a common area after each pixel is subjected to pixel expansion processing is extracted as the selected area.
  • the tongue coating of the same spectral intensity value in the computer tongue imaging can be accurately extracted, which is beneficial for subsequent processing.
  • the architecture principle of a medical image processing system is as follows:
  • a multi-pixel selection module for manually selecting one or more pixels in a region required for a digital image of a tongue
  • a pixel expansion processing module configured to perform pixel expansion processing on each selected pixel
  • the common area extraction module is configured to extract a common area after each pixel is subjected to pixel expansion processing as the selected area.
  • the architecture principle of a pixel expansion processing module of a medical image processing system is as follows:
  • a top-of-stack pixel pop-up module for popping the top pixel of the stack
  • a pop-up pixel identification module configured to identify a flag of the pop-up pixel as “selected”
  • a neighboring pixel processing module configured to sequentially check the neighboring pixels of the popped pixel, if one of the pixels satisfies the boundary condition and the flag is not identified as “selected”, the pixel is pushed onto the stack, and the pixel is The pixel identifier is "selected”, otherwise the next neighborhood pixel is continued to be searched until the neighboring pixels of the popped pixel have been checked;
  • the stack non-empty judgment module is configured to determine whether the stack is empty. If the stack is not empty, jump to the top pixel pop-up block module, and if the stack is empty, jump to the end module;
  • the end module is used to identify all the flag bits as "selected" pixels as the area after pixel expansion processing.
  • the boundary condition is that a sum of squares of the difference between the spectral intensity value of the popped pixel and the spectral intensity value of the neighboring pixel is less than a boundary threshold, that is, (P’h-Ph)2 ⁇ Thresh && (P’s-Ps)2 ⁇ Thress && (P'v-Pv)2 ⁇ Thresv, where P'h, Ph, and Thresh are the spectral intensity values of the pop-up pixels of the medical image in the spatial h component of the HSV spectral intensity value, the spectral intensity values of the neighboring pixels, and the corresponding The boundary threshold, where P's, Ps, and Thress are the spectral intensity values of the pop-up pixels of the medical image in the HSV spectral intensity value space s component, the spectral intensity values of the neighboring pixels, and the corresponding boundary threshold, where P'v , Pv, and Thresv are the spectral intensity values of the pop-up pixels
  • the boundary threshold Thresx (x is the h component or the s component or the v component of the medical image) is calculated as follows:
  • Step a obtaining a gray level Lv of a maximum value of the relative entropy of the full image of the medical image component map, and a value Hv of the histogram of the medical image component map corresponding to the Lv, wherein the relative entropy is defined as
  • P1, p2, ..., pn and q1, q2, ..., qn are probability distributions of tongue coating and tongue pixels
  • Step b respectively calculating the gray level La, Lb at the left and right peaks of the histogram of the medical image component map, and the values Ha, Hb of the histogram of the medical image component map corresponding to La, Lb;
  • Step c if (Ha-Hv)/(Lv-La)>(Hb-Hv)/(Lb-Lv),
  • Thresx E*(Hb-Hv)/(Lb-Lv),
  • Thresx E*(Ha-Hv)/(Lv-La).

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)
  • Image Processing (AREA)
  • Apparatus For Radiation Diagnosis (AREA)

Abstract

本发明涉及一种医学影像处理方法，其包括包括以下步骤：第一步，人工随机选取舌数字影像所需区域中1个以上的像素；第二步，对所选取的每个像素进行像素扩充处理；第三步，提取每个像素经过像素扩充处理后的共同区域作为所选区域，解决了现阶段只有将舌苔与舌质部分区分开来的方法，却没有半自动将同种舌谱强度值的舌苔部分提取出来的方法的问题。

Description

一种医学影像处理方法与系统 技术领域
在计算机舌影像学中，为了教学与科研的需要，往往需要将舌数字影像中的同种舌谱强度值的舌苔部分提取出来，以供后续处理使用，但是舌影像中，舌苔部分往往分布有多种谱强度值的舌苔，而不同舌苔之间的边界十分模糊，难以清楚的区分开来，现阶段只有将舌苔与舌质部分区分开来的方法，却没有半自动将同种谱强度值的舌苔部分提取出来的方法。
 背景技术
本发明的目的在于提出一种医学影像处理方法，其能解决在计算机舌影像学中，需要将舌数字影像中的同种谱强度值的舌苔部分提取出来的问题。
 技术问题
本发明的目的在于提出一种医学影像处理方法，其能解决在计算机舌影像学中，需要将舌数字影像中的同种谱强度值的舌苔部分提取出来的问题。
 技术解决方案
　　　为了达到上述目的，本发明所采用的技术方案如下：
一种医学影像处理方法，其特征在于包括以下步骤：
第一步，人工随机选取舌数字影像所需区域中1个以上的像素；
第二步，对所选取的每个像素进行像素扩充处理；
　　　第三步，提取每个像素经过像素扩充处理后的共同区域作为所选区域。
优选的，所述像素扩充处理，包括以下步骤：每个人工随机选取的像素入栈，
第1步，栈顶像素出栈；
第2步，将出栈像素的标志位标识为“已选取”；
第3步，按顺序检查与出栈像素的邻域像素，若其中某个像素满足边界条件且其标志位未标识为“已选取”，则把该像素入栈，并把该像素标识为“已选取”，否则继续查找下一个邻域像素，直到该出栈像素的邻域像素都已检查；
第4步，如果栈非空，重复执行第1步，如果栈为空，所有标志位标识为“已选取”的像素为像素扩充处理后的区域。
优选的，所述边界条件为出栈像素的谱强度值与其邻域像素的谱强度值的差值平方和分别小于边界阈值,即
Figure PCTCN2013075527-appb-M000001
，
其中的P’h、Ph、Thresh为医学影像在HSV谱强度值空间h分量的出栈像素的谱强度值、其邻域像素的谱强度值和对应的边界阈值，其中的P’s、Ps、Thress为医学影像在HSV谱强度值空间s分量的出栈像素的谱强度值、其邻域像素的谱强度值和对应的边界阈值，其中的P’v、Pv、Thresv为医学影像在HSV谱强度值空间v分量的出栈像素的谱强度值、其邻域像素的谱强度值和对应的边界阈值。
优选的，所述边界阈值Thresx（x为医学影像的h分量或s分量或v分量）的计算方法如下：
第a步，取得该医学影像分量图全图的相对熵的最大值的灰度级Lv，以及Lv处对应的该医学影像分量图的直方图的值Hv，其中相对熵的定义为
Figure PCTCN2013075527-appb-M000002
 ,
p1，p2,…，pn和q1，q2,…，qn是舌苔和舌质像素的概率分布；
第b步，分别计算该医学影像分量图的直方图的左峰与右峰处的灰度级La、Lb，以及La、Lb处对应的该医学影像分量图的直方图的值Ha、Hb；
第c步，如果（Ha-Hv）/(Lv-La)>（Hb-Hv）/(Lb-Lv),
Thresx=E*（Hb-Hv）/(Lb-Lv),
　　　否则，
　　　Thresx=E*（Ha-Hv）/(Lv-La)。
本发明实施例的另一目的在于提供一种医学影像处理系统，其特征在于包括：
多点像素选取模块，用于人工随机选取舌数字影像所需区域中1个以上的像素；
像素扩充处理模块，用于对所选取的每个像素进行像素扩充处理；
共同区域提取模块，用于提取每个像素经过像素扩充处理后的共同区域作为所选区域。
优选的，所述像素扩充处理模块，包括：
选取像素入栈模块，用于每个人工随机选取的像素入栈；
栈顶像素出栈模块，用于将栈顶像素出栈；
出栈像素标识模块，用于将出栈像素的标志位标识为“已选取”；
邻域像素处理模块，用于按顺序检查与出栈像素的邻域像素，若其中某个像素满足边界条件且其标志位未标识为“已选取”，则把该像素入栈，并把该像素标识为“已选取”，否则继续查找下一个邻域像素，直到该出栈像素的邻域像素都已检查；
栈非空判断模块，用于判断栈是否为空，如果栈非空，跳转到栈顶像素出栈模块，如果栈为空，跳转到结束模块；
结束模块，用于将所有标志位标识为“已选取”的像素为像素扩充处理后的区域。
优选的，所述边界条件为出栈像素的谱强度值与其邻域像素的谱强度值的差值平方和分别小于边界阈值,即
Figure PCTCN2013075527-appb-M000003
，
其中的P’h、Ph、Thresh为医学影像在HSV谱强度值空间h分量的出栈像素的谱强度值、其邻域像素的谱强度值和对应的边界阈值，其中的P’s、Ps、Thress为医学影像在HSV谱强度值空间s分量的出栈像素的谱强度值、其邻域像素的谱强度值和对应的边界阈值，其中的P’v、Pv、Thresv为医学影像在HSV谱强度值空间v分量的出栈像素的谱强度值、其邻域像素的谱强度值和对应的边界阈值。
优选的，所述边界阈值Thresx（x为医学影像的h分量或s分量或v分量）的计算方法如下：
第a步，取得该医学影像分量图全图的相对熵的最大值的灰度级Lv，以及Lv处对应的该医学影像分量图的直方图的值Hv，其中相对熵的定义为
Figure PCTCN2013075527-appb-M000004
 ,
p1，p2,…，pn和q1，q2,…，qn是舌苔和舌质像素的概率分布；
第b步，分别计算该医学影像分量图的直方图的左峰与右峰处的灰度级La、Lb，以及La、Lb处对应的该医学影像分量图的直方图的值Ha、Hb；
第c步，如果（Ha-Hv）/(Lv-La)>（Hb-Hv）/(Lb-Lv),
Thresx=E*（Hb-Hv）/(Lb-Lv),
　　　否则，
　　　Thresx=E*（Ha-Hv）/(Lv-La)。
 有益效果
　　　本发明具有如下有益效果：
该医学影像处理方法可以准确的将计算机舌影像学中的同种谱强度值的舌苔提取出来，有利于后续处理使用。
 附图说明
　　　图1为本发明的医学影像处理方法的原理方框图；
　　　图2为本发明的医学影像处理方法的像素扩充处理步骤的原理方框图；
　　　图3为本发明的医学影像处理方法的边界阈值计算方法的原理方框图；
　　　图4为本发明的医学影像处理系统的结构示意图；
　　　图5为本发明的医学影像处理系统的像素扩充处理模块的结构示意图。
 本发明的最佳实施方式
　　　下面，结合附图以及具体实施方式，对本发明做进一步描述。
　　　在计算机舌影像学中，为了教学与科研的需要，往往需要将舌数字影像中的同种舌谱强度值的舌苔部分提取出来，以供后续处理使用，但是舌影像中，舌苔部分往往分布有多种谱强度值的舌苔，而不同舌苔之间的边界十分模糊，难以清楚的区分开来，现阶段只有将舌苔与舌质部分区分开来的方法，却没有半自动将同种谱强度值的舌苔部分提取出来的方法。
　　　将同种谱强度值的舌苔部分提取出来的方法，实际上相当于进行了一次二值影像分割。所谓二值影像，就是指影像上的所有点的灰度值只用两种可能，不为"0"就为"255"，也就是整个影像呈现出明显的黑白效果。为了得到理想的二值影像，一般采用阈值分割技术，它对物体与背景有较强对比的影像的分割特别有效，它计算简单而且总能用封闭、连通的边界定义不交叠的区域。所有灰度大于或等于阈值的像素被判决为属于物体，灰度值用"255"表示，否则这些像素点被排除在物体区域以外，灰度值为"0"，表示背景。这样一来物体的边界就成为这样一些内部的点的集合，这些点都至少有一个邻点不属于该物体。如果感兴趣的物体在内部有均匀一致的灰度值，并且其处在一个具有另外一个灰度值的均匀背景下，使用阈值法可以得到比较好的效果。如果物体同背景的差别不在灰度值上（比如纹理不同），可以将这个性质转换为灰度的差别，然后利用阈值化技术来分割该影像。为了使分割更加鲁棒，适用性更强，系统应该可以自动选择阈值。基于物体、环境和应用域等知识的影像分割算法比基于固定阈值的算法更具有普遍性和适应性。这些知识包括：对应于物体的影像灰度特性、物体的尺寸、物体在影像中所占的比例、影像中不同类型物体的数量等。
　　　但是舌影像中，舌苔部分往往分布有多种谱强度值的舌苔，而不同舌苔之间的边界十分模糊，难以清楚的区分开来，因此我们考虑将多次二值影像分割的结果综合起来，取其共同区域，作为提取结果，其结果更为准确，避免给后续步骤引入不必要的噪声。
　　　因此，如图1所示，为本发明实施例提供的一种医学影像处理方法实现流程，详述如下：
　　　第一步，人工随机选取舌数字影像所需区域中1个以上的像素；舌苔部分往往分布有多种谱强度值的舌苔，而不同舌苔之间的边界十分模糊，难以清楚的区分开来，对于每个像素，后续的第二步和第三步处理提取的区域各有不同，但其公共部分是基本稳定的，也是需要提取的区域，所以第一步中需要人工随机选取舌数字影像所需区域中多个像素，以备后续步骤应用。
　　　第二步，对所选取的每个像素进行像素扩充处理；
　　　其中，像素扩充处理，包括以下步骤：每个人工随机选取的像素入栈，
　　　第1步，栈顶像素出栈；
　　　第2步，将出栈像素的标志位标识为“已选取”；
　　　第3步，按顺序检查与出栈像素的邻域像素，若其中某个像素满足边界条件且其标志位未标识为“已选取”，则把该像素入栈，并把该像素标识为“已选取”，否则继续查找下一个邻域像素，直到该出栈像素的邻域像素都已检查；
　　　第4步，如果栈非空，重复执行第1步，如果栈为空，所有标志位标识为“已选取”的像素为像素扩充处理后的区域。
　　　这样的方法，减少了传统方法的出栈入栈操作，提高了方法的效率。
　　　这里边界条件为出栈像素的谱强度值与其邻域像素的谱强度值的差值平方和分别小于边界阈值,即
Figure PCTCN2013075527-appb-M000005
，
其中的P’h、Ph、Thresh为医学影像在HSV谱强度值空间h分量的出栈像素的谱强度值、其邻域像素的谱强度值和对应的边界阈值，其中的P’s、Ps、Thress为医学影像在HSV谱强度值空间s分量的出栈像素的谱强度值、其邻域像素的谱强度值和对应的边界阈值，其中的P’v、Pv、Thresv为医学影像在HSV谱强度值空间v分量的出栈像素的谱强度值、其邻域像素的谱强度值和对应的边界阈值。
　　　HSV 模型在 1978 年由埃尔维·雷·史密斯创立，它是三原色光模式的一种非线性变换。
　　　色相（H）是色彩的基本属性，就是平常所说的色彩名称，如红色、黄色等。饱和度（S）是指色彩的纯度，越高色彩越纯，低则逐渐变灰，取0-100%的数值。明度（V），取0-100%。
　　　HSV 把色彩描述在圆柱坐标系内的点，这个圆柱的中心轴取值为自底部的黑色到顶部的白色而在它们中间是的灰色，绕这个轴的角度对应于“色相”，到这个轴的距离对应于“饱和度”，而沿着这个轴的高度对应于“亮度”，“色调”或“明度”。
　　　HSV 模型通常用于计算机图形应用中。在用户必须选择一个色彩应用于特定图形元素各种应用环境中，经常使用 HSV 色轮。在其中，色相表示为圆环；可以使用一个独立的三角形来表示饱和度和明度。典型的，这个三角形的垂直轴指示饱和度，而水平轴表示明度。在这种方式下，选择色彩可以首先在圆环中选择色相，在从三角形中选择想要的饱和度和明度。
　　　HSV 模型的另一种可视方法是圆锥体。在这种表示中，色相被表示为绕圆锥中心轴的角度，饱和度被表示为从圆锥的横截面的圆心到这个点的距离，明度被表示为从圆锥的横截面的圆心到顶点的距离。某些表示使用了六棱锥体。这种方法更适合在一个单一物体中展示这个 HSV 色彩空间；但是由于它的三维本质，它不适合在二维计算机界面中选择色彩。
　　　HSV 色彩空间还可以表示为类似于上述圆锥体的圆柱体，色相沿着圆柱体的外圆周变化，饱和度沿着从横截面的圆心的距离变化，明度沿着横截面到底面和顶面的距离而变化。这种表示可能被认为是 HSV 色彩空间的更精确的数学模型；但是在实际中可区分出的饱和度和色相的级别数目随着明度接近黑色而减少。此外计算机典型的用有限精度范围来存储 RGB 值；这约束了精度，再加上人类色彩感知的限制，使圆锥体表示在多数情况下更实用。
　　　HSV色彩空间符合人眼对颜色的感觉，人的视觉对亮度的敏感程度远强于对颜色浓淡的敏感程度，所以HSV色彩空间比RGB色彩空间更符合人的视觉特性，便于色彩的处理和识别； 
　　　HSV 颜色空间中 3 个坐标是独立的，即人可以独立感知各颜色分量的变化；
　　　HSV 颜色空间构成的是一个均匀的颜色模型，采用线性的标尺，颜色之间感觉上的距离与 HSV 颜色空间坐标上点的欧几里德距离成正比。
　　　因此我们采用HSV色彩空间的边界阈值条件来限制像素扩充处理的边界。
　　　又因为提取同色舌苔区域本质上还可以看做一种阈值分割法，阈值分割法分为全局阈值法和局部阈值分割法。所谓局部阈值分割法是将原始影像划分成较小的影像，并对每个子影像选取相应的阈值。在阈值分割后，相邻子影像之间的边界处可能产生灰度级的不连续性，因此需用平滑技术进行排除。局部阈值法常用的方法有灰度差直方图法、微分直方图法。局部阈值分割法虽然能改善分割效果，但存在几个缺点：
　　　（1）每幅子影像的尺寸不能太小，否则统计出的结果无意义。
　　　（2）每幅影像的分割是任意的，如果有一幅子影像正好落在目标区域或背景区域，而根据统计结果对其进行分割，也许会产生更差的结果。
　　　（3）局部阈值法对每一幅子影像都要进行统计，速度慢，难以适应实时性的要求。
　　　全局阈值分割方法在影像处理中应用比较多，它在整幅影像内采用固定的阈值分割影像。经典的阈值选取以灰度直方图为处理对象。根据阈值选择方法的不同，可以分为模态方法、迭代式阈值选择等方法。这些方法都是以影像的直方图为研究对象来确定分割的阈值的。另外还有类间方差阈值分割法、二维最大熵分割法、模糊阈值分割法、共生矩阵分割法、区域生长法等等。
　　　对于比较简单的影像，可以假定物体和背景分别处于不同的灰度级，影像被零均值高斯噪声污染，所以影像的灰度分布曲线近似认为是由两个正态分布函数叠加而成，影像的直方图将会出现两个分离的峰值。对于这样的影像，分割阈值可以选择直方图的两个波峰间的波谷所对应的灰度值作为分割的阈值。这种分割方法不可避免的会出现误分割，使一部分本属于背景的像素被判决为物体，属于物体的一部分像素同样会被误认为是背景。可以证明，当物体的尺寸和背景相等时，这样选择阈值可以使误分概率达到最小。在大多数情况下，由于影像的直方图在波谷附近的像素很稀疏，因此这种方法对影像的分割影响不大。这一方法可以推广到具有不同灰度均值的多物体影像。
　　　因此我们这里选用了取得该医学影像全图的相对熵的最大值的灰度级，其中相对熵的定义为
Figure PCTCN2013075527-appb-M000006
,
　　　这是一种迭代式阈值选择算法，将影像分割成舌苔区域、舌质区域两部分，对影像中的每一个灰度级分别求取舌苔区域、舌质区域两部分相对熵的和，选取使舌苔区域、舌质区域两部分相对熵的和最大的灰度级作为分割影像的阈值。
　　　对于给定的影像，由于大部份的像素点属于目标区域或背景，而目标和背景区域内部像素点的灰度级比较均匀，像素点的灰度和其邻域均值的灰度级相差不大，所以影像对应的二维直方图 主要集中在i,j平面的对角线附近，并且在总体上呈现双峰和一谷的状态，两个峰分别对应于目标和背景。传统方法的缺点是仅仅考虑了像素点的灰度信息，没有考虑到像素点的空间信息，所以当影像的信噪比降低时分割效果不理想。毫无疑问，像素点的灰度是最基本的特征，但它对噪声比较敏感，为此，在分割影像时可以再考虑影像的区域信息，区域灰度特征包含了影像的部分空间信息，且对噪声的敏感程度要低于点灰度特征。而本方法恰好解决了这一问题。
　　　因此我们经研究，其中所述边界阈值Thresx（x为医学影像的h分量或s分量或v分量）的计算方法如下：
第a步，取得该医学影像分量图全图的相对熵的最大值的灰度级Lv，以及Lv处对应的该医学影像分量图的直方图的值Hv，其中相对熵的定义为
Figure PCTCN2013075527-appb-M000007
,
p1，p2,…，pn和q1，q2,…，qn是舌苔和舌质像素的概率分布；
第b步，分别计算该医学影像分量图的直方图的左峰与右峰处的灰度级La、Lb，以及La、Lb处对应的该医学影像分量图的直方图的值Ha、Hb；其中直方图的左峰与右峰分别对应舌苔与舌质类别。
第c步，如果（Ha-Hv）/(Lv-La)>（Hb-Hv）/(Lb-Lv),
Thresx=E*（Hb-Hv）/(Lb-Lv),
　　　否则，
Thresx=E*（Ha-Hv）/(Lv-La)。这里E是一个经验常数。
　　　第三步，提取每个像素经过像素扩充处理后的共同区域作为所选区域。
　　　通过上述的方法，可以准确的将计算机舌影像学中的同种谱强度值的舌苔提取出来，有利于后续处理使用。我们对1210个计算机舌影像进行了测试，三名医生负责人工标定。他们都是拥有20年专业经验的中医临床专家，且都是中国相关权威学术机构的会员。因为谱强度值的感知是一种共同感受，且舌象谱强度值与疾病的关系拥有共同的医学分布规律，所以他们的意见高度一致。经过测试，分割的准确率高达95.29%。
如图4所示，为本发明实施例提供的一种医学影像处理系统的架构原理，详述如下，包括：
多点像素选取模块，用于人工随机选取舌数字影像所需区域中1个以上的像素；
像素扩充处理模块，用于对所选取的每个像素进行像素扩充处理；
共同区域提取模块，用于提取每个像素经过像素扩充处理后的共同区域作为所选区域。
如图5所示，为本发明实施例提供的一种医学影像处理系统的像素扩充处理模块的架构原理，详述如下，包括：
选取像素入栈模块，用于每个人工随机选取的像素入栈；
栈顶像素出栈模块，用于将栈顶像素出栈；
出栈像素标识模块，用于将出栈像素的标志位标识为“已选取”；
邻域像素处理模块，用于按顺序检查与出栈像素的邻域像素，若其中某个像素满足边界条件且其标志位未标识为“已选取”，则把该像素入栈，并把该像素标识为“已选取”，否则继续查找下一个邻域像素，直到该出栈像素的邻域像素都已检查；
栈非空判断模块，用于判断栈是否为空，如果栈非空，跳转到栈顶像素出栈模块，如果栈为空，跳转到结束模块；
　　　结束模块，用于将所有标志位标识为“已选取”的像素为像素扩充处理后的区域。
优选的，所述边界条件为出栈像素的谱强度值与其邻域像素的谱强度值的差值平方和分别小于边界阈值,即(P’h-Ph)2<Thresh && (P’s-Ps)2<Thress && (P’v-Pv)2<Thresv，其中的P’h、Ph、Thresh为医学影像在HSV谱强度值空间h分量的出栈像素的谱强度值、其邻域像素的谱强度值和对应的边界阈值，其中的P’s、Ps、Thress为医学影像在HSV谱强度值空间s分量的出栈像素的谱强度值、其邻域像素的谱强度值和对应的边界阈值，其中的P’v、Pv、Thresv为医学影像在HSV谱强度值空间v分量的出栈像素的谱强度值、其邻域像素的谱强度值和对应的边界阈值。
优选的，所述边界阈值Thresx（x为医学影像的h分量或s分量或v分量）的计算方法如下：
第a步，取得该医学影像分量图全图的相对熵的最大值的灰度级Lv，以及Lv处对应的该医学影像分量图的直方图的值Hv，其中相对熵的定义为
Figure PCTCN2013075527-appb-M000008
 ,
p1，p2,…，pn和q1，q2,…，qn是舌苔和舌质像素的概率分布；
第b步，分别计算该医学影像分量图的直方图的左峰与右峰处的灰度级La、Lb，以及La、Lb处对应的该医学影像分量图的直方图的值Ha、Hb；
第c步，如果（Ha-Hv）/(Lv-La)>（Hb-Hv）/(Lb-Lv),
Thresx=E*（Hb-Hv）/(Lb-Lv),
　　　否则，
　　　Thresx=E*（Ha-Hv）/(Lv-La)。
对于本领域的技术人员来说，可根据以上描述的技术方案以及构思，做出其它各种相应的改变以及变形，而所有的这些改变以及变形都应该属于本发明权利要求的保护范围之内。
 本发明的实施方式
工业实用性
序列表自由内容

Claims (1)

  1. 1.一种医学影像处理方法，其特征在于包括以下步骤：
    第一步，人工随机选取舌数字影像所需区域中1个以上的像素；
    第二步，对所选取的每个像素进行像素扩充处理；
    第三步，提取每个像素经过像素扩充处理后的共同区域作为所选区域。
    2.如权利要求1所述的医学影像处理方法，其特征在于，所述像素扩充处理，包括以下步骤：
    第1步，每个人工随机选取的像素入栈；
    第2步，栈顶像素出栈；
    第3步，将出栈像素的标志位标识为“已选取”；
    第4步，按顺序检查与出栈像素的邻域像素，若其中某个像素满足边界条件且其标志位未标识为“已选取”，则把该像素入栈，并把该像素标识为“已选取”，否则继续查找下一个邻域像素，直到该出栈像素的邻域像素都已检查；
    第5步，如果栈非空，重复执行第2步，如果栈为空，所有标志位标识为“已选取”的像素为像素扩充处理后的区域。
    3.如权利要求1所述的医学影像处理方法，其特征在于，所述边界条件为出栈像素的谱强度值与其邻域像素的谱强度值的差值平方和分别小于边界阈值,即
    Figure PCTCN2013075527-appb-M000009
    ，
    其中的P’h、Ph、Thresh为医学影像在HSV谱强度值空间h分量的出栈像素的谱强度值、其邻域像素的谱强度值和对应的边界阈值，其中的P’s、Ps、Thress为医学影像在HSV谱强度值空间s分量的出栈像素的谱强度值、其邻域像素的谱强度值和对应的边界阈值，其中的P’v、Pv、Thresv为医学影像在HSV谱强度值空间v分量的出栈像素的谱强度值、其邻域像素的谱强度值和对应的边界阈值。
    4.如权利要求1所述的医学影像处理方法，其特征在于，所述边界阈值Thresx（x为医学影像的h分量或s分量或v分量）的计算方法如下：
    第a步，取得该医学影像分量图全图的相对熵的最大值的灰度级Lv，以及Lv处对应的该医学影像分量图的直方图的值Hv，其中相对熵的定义为
    Figure PCTCN2013075527-appb-M000010
    ,
    p1，p2,…，pn和q1，q2,…，qn是舌苔和舌质像素的概率分布；
    第b步，分别计算该医学影像分量图的直方图的左峰与右峰处的灰度级La、Lb，以及La、Lb处对应的该医学影像分量图的直方图的值Ha、Hb；
    第c步，如果（Ha-Hv）/(Lv-La)>（Hb-Hv）/(Lb-Lv),
    Thresx=E*（Hb-Hv）/(Lb-Lv),
    　　　否则，
    　　　Thresx=E*（Ha-Hv）/(Lv-La)。
    5.一种医学影像处理系统，其特征在于包括：
    多点像素选取模块，用于人工随机选取舌数字影像所需区域中1个以上的像素；
    像素扩充处理模块，用于对所选取的每个像素进行像素扩充处理；
    共同区域提取模块，用于提取每个像素经过像素扩充处理后的共同区域作为所选区域。
    6.如权利要求5所述的医学影像处理系统，其特征在于，所述像素扩充处理模块，包括：
    选取像素入栈模块，用于每个人工随机选取的像素入栈；
    栈顶像素出栈模块，用于将栈顶像素出栈；
    出栈像素标识模块，用于将出栈像素的标志位标识为“已选取”；
    邻域像素处理模块，用于按顺序检查与出栈像素的邻域像素，若其中某个像素满足边界条件且其标志位未标识为“已选取”，则把该像素入栈，并把该像素标识为“已选取”，否则继续查找下一个邻域像素，直到该出栈像素的邻域像素都已检查；
    栈非空判断模块，用于判断栈是否为空，如果栈非空，跳转到栈顶像素出栈模块，如果栈为空，跳转到结束模块；
    结束模块，用于将所有标志位标识为“已选取”的像素为像素扩充处理后的区域。
PCT/CN2013/075527 2013-05-13 2013-05-13 一种医学影像处理方法与系统 WO2014183246A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN201380003067.4A CN103946868A (zh) 2013-05-13 2013-05-13 一种医学影像处理方法与系统
PCT/CN2013/075527 WO2014183246A1 (zh) 2013-05-13 2013-05-13 一种医学影像处理方法与系统

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2013/075527 WO2014183246A1 (zh) 2013-05-13 2013-05-13 一种医学影像处理方法与系统

Publications (1)

Publication Number Publication Date
WO2014183246A1 true WO2014183246A1 (zh) 2014-11-20

Family

ID=51193112

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2013/075527 WO2014183246A1 (zh) 2013-05-13 2013-05-13 一种医学影像处理方法与系统

Country Status (2)

Country Link
CN (1) CN103946868A (zh)
WO (1) WO2014183246A1 (zh)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105232050A (zh) * 2015-09-18 2016-01-13 天津中医药大学 一种人体舌苔厚度的无创检测装置
CN107610087A (zh) * 2017-05-15 2018-01-19 华南理工大学 一种基于深度学习的舌苔自动分割方法
CN110570515A (zh) * 2019-09-03 2019-12-13 天津工业大学 一种利用ct图像进行人体骨骼三维建模的方法
CN113724248A (zh) * 2021-09-16 2021-11-30 北京航空航天大学 医学影像针道检测方法及装置
CN114913422A (zh) * 2022-05-23 2022-08-16 桂林理工大学 一种岩溶石山山区水土流失野外监测装置
CN115953392A (zh) * 2023-03-09 2023-04-11 四川博瑞客信息技术有限公司 基于人工智能的舌体苔质评估方法

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105869151B (zh) * 2016-03-24 2018-12-25 天津大学 舌分割及舌苔舌质分离方法
CN109978895B (zh) * 2019-03-29 2021-08-13 北京峰云视觉技术有限公司 一种舌体图像分割方法和装置

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101082980A (zh) * 2006-06-01 2007-12-05 上海吉量软件科技有限公司 一种新型的交点配对填充算法
CN101101668A (zh) * 2007-07-13 2008-01-09 杭州电子科技大学 一种基于振荡网络的医学图像分割方法
WO2009038376A1 (en) * 2007-09-21 2009-03-26 Korea Institute Of Oriental Medicine Extraction method of tongue region using graph-based approach and geometric properties
CN102509312A (zh) * 2011-09-20 2012-06-20 哈尔滨工业大学 人体数字舌图像颜色色域空间及其提取方法
CN102737376A (zh) * 2012-03-31 2012-10-17 常熟市支塘镇新盛技术咨询服务有限公司 一种应用在冠状动脉造影图像分割中的改进的区域生长方法

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102982784B (zh) * 2011-09-06 2015-03-04 联咏科技股份有限公司 影像调整方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101082980A (zh) * 2006-06-01 2007-12-05 上海吉量软件科技有限公司 一种新型的交点配对填充算法
CN101101668A (zh) * 2007-07-13 2008-01-09 杭州电子科技大学 一种基于振荡网络的医学图像分割方法
WO2009038376A1 (en) * 2007-09-21 2009-03-26 Korea Institute Of Oriental Medicine Extraction method of tongue region using graph-based approach and geometric properties
CN102509312A (zh) * 2011-09-20 2012-06-20 哈尔滨工业大学 人体数字舌图像颜色色域空间及其提取方法
CN102737376A (zh) * 2012-03-31 2012-10-17 常熟市支塘镇新盛技术咨询服务有限公司 一种应用在冠状动脉造影图像分割中的改进的区域生长方法

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105232050A (zh) * 2015-09-18 2016-01-13 天津中医药大学 一种人体舌苔厚度的无创检测装置
CN107610087A (zh) * 2017-05-15 2018-01-19 华南理工大学 一种基于深度学习的舌苔自动分割方法
CN107610087B (zh) * 2017-05-15 2020-04-28 华南理工大学 一种基于深度学习的舌苔自动分割方法
CN110570515A (zh) * 2019-09-03 2019-12-13 天津工业大学 一种利用ct图像进行人体骨骼三维建模的方法
CN110570515B (zh) * 2019-09-03 2024-05-10 天津工业大学 一种利用ct图像进行人体骨骼三维建模的方法
CN113724248A (zh) * 2021-09-16 2021-11-30 北京航空航天大学 医学影像针道检测方法及装置
CN113724248B (zh) * 2021-09-16 2024-03-26 北京航空航天大学 医学影像针道检测方法及装置
CN114913422A (zh) * 2022-05-23 2022-08-16 桂林理工大学 一种岩溶石山山区水土流失野外监测装置
CN114913422B (zh) * 2022-05-23 2024-03-15 桂林理工大学 一种岩溶石山山区水土流失野外监测装置
CN115953392A (zh) * 2023-03-09 2023-04-11 四川博瑞客信息技术有限公司 基于人工智能的舌体苔质评估方法

Also Published As

Publication number Publication date
CN103946868A (zh) 2014-07-23

Similar Documents

Publication Publication Date Title
WO2014183246A1 (zh) 一种医学影像处理方法与系统
CN106023151B (zh) 一种开放环境下中医舌象目标检测方法
WO2019062092A1 (zh) 基于超像素-多色彩空间的人体轮廓提取方法
Ma et al. Shadow segmentation and compensation in high resolution satellite images
CN110268420B (zh) 在图像中检测背景物上的外来物的计算机实现的方法、在图像中检测背景物上的外来物的设备以及计算机程序产品
CN112785579A (zh) 基于图像处理技术的混凝土裂缝识别方法
CN103914699A (zh) 一种基于色彩空间的自动唇彩的图像增强的方法
CN104504722B (zh) 一种利用灰色点校正图像颜色的方法
CN110335233B (zh) 基于图像处理技术的高速公路护栏板缺陷检测系统和方法
CN110751635B (zh) 一种基于帧间差分和hsv颜色空间的口腔检测方法
WO2020038312A1 (zh) 多通道舌体边缘检测装置、方法及存储介质
CN110268442B (zh) 在图像中检测背景物上的外来物的计算机实现的方法、在图像中检测背景物上的外来物的设备以及计算机程序产品
CN113343976B (zh) 基于颜色-边缘融合特征生长的抗高光干扰工程测量标志提取方法
CN111461036B (zh) 一种利用背景建模增强数据的实时行人检测方法
CN111539980B (zh) 一种基于可见光的多目标追踪方法
Huang et al. Adaptive automatic segmentation of HEp-2 cells in indirect immunofluorescence images
KR100903816B1 (ko) 정지영상에서의 얼굴추출시스템과 그 방법
CN113643281B (zh) 一种舌体图像分割方法
CN109064490B (zh) 一种基于MeanShift的运动目标跟踪方法
Kang et al. Dental plaque quantification using mean-shift-based image segmentation
Islami Implementation of HSV-based Thresholding Method for Iris Detection
CN117274981A (zh) 一种基于图像识别的辣椒病理推测方法和装置
Barata et al. 1 Toward a Robust Analysis of Dermoscopy Images Acquired under Different
CN108629780B (zh) 基于颜色分解和阈值技术的舌图像分割方法
CN109658382B (zh) 基于图像聚类和灰度投影的舌体定位方法

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: 13884861

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: 13884861

Country of ref document: EP

Kind code of ref document: A1