Improved two-dimensional mcmc reconstruction method and system for rock thin section
The improved MCMC reconstruction method for rock thin sections addresses anisotropy by binarizing images, calculating porosity, and using direction-matched conditional probability scanning, achieving better image reconstruction and fitting of variation functions.
Patent Information
- Application Number
- GB2022004882
- Authority / Receiving Office
- GB · GB
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2020-05-28
- Filing Date
- 2021-03-15
- Publication Date
- 2024-12-04
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing two-dimensional MCMC reconstruction methods for rock thin sections fail to accurately account for the anisotropy of rock images, leading to imbalanced conditional probabilities and poor fitting of variation functions.
An improved two-dimensional MCMC reconstruction method that binarizes the original image, calculates porosity, determines neighborhood systems and conditional probabilities, and uses a scanning direction that matches the reconstruction path, allowing for balanced conditional probability scanning and Monte Carlo-based image reconstruction, with stopping conditions based on porosity matching.
The method achieves a higher degree of fitting between the reconstructed and original images by considering anisotropy, resulting in improved image reconstruction effects that characterize rock image anisotropy effectively.
Smart Images

Figure 00000001_0000 
Figure 00000001_0001 
Figure 00000001_0002
Abstract
Description
CROSS REFERENCE TO RELATED APPLICATION
[01] This patent application claims the benefit and priority of Chinese Patent Application No. 202010468814.3 filed with the China National Intellectual Property Administration (CNIPA) on May 28, 2020 and entitled "IMPROVED TWO-DIMENSIONAL MCMC RECONSTRUCTION METHOD AND SYSTEM FOR ROCK THIN SECTION", the disclosure of which is incorporated by reference herein in its entirety as part of the present application. TECHNICAL FIELD
[02] The present disclosure belongs to the technical field of computer image processing and relates to an improved two-dimensional Markov chain Monte Carlo (MCMC) reconstruction method and system for a rock thin section. BACKGROUND ART
[03] Due to high convenience and low cost, image reconstruction has been widely used in various fields such as medicine and geology for performing characteristic prediction and structural analysis on an unknown region. To make the characteristics of a reconstructed image closer to the reality, an appropriate imaging method is a direct influencing factor for reconstruction effect.
[04] In the prior art, Kejian Wu’s article “An Efficient Markov Chain Model for the Simulation of Heterogeneous Soil Structure” systematically explains a MCMC approach. In general, using the idea of Markov chain, after an appropriate neighborhood system is selected, an electron microscopic image of a rock thin section is traversed and scanned in a fixed direction to statistically analyze the regularity of distribution thereof, thereby obtaining conditional probabilities. Monte Carlo method is then used to obtain random numbers for solving calculation problems. Finally, whether the reconstruction effect is good or not is determined by means of fitting of a variation function. However, path direction will change constantly during reconstruction. Moreover, to make the result closer to the original image, weights will be set for the conditional probabilities, which directly results in changes in statistical information. Furthermore, the fitting effect of the variation function cannot be determined quantitatively. Therefore, in view of common anisotropy of rock images, the improved method provided in the present disclosure may be more reasonable. SUMMARY
[05] With respect to the defects and shortcomings in the prior art, the present disclosure provides an improved two-dimensional reconstruction method and system for a rock thin section, and an objective of the present disclosure is to solve the existing problem that conditional probabilities are also anisotropic since the common characteristic of anisotropy of rock images is not considered in the prior art.
[06] To achieve the above-mentioned objective, the present disclosure adopts the following technical solutions:
[07] An improved two-dimensional MCMC reconstruction method for a rock thin section includes the following steps:
[08] step 1, binarizing an original image and calculating a porosity:
[09] for an original image having a size of m*n, calculating a threshold of the image by maximizing a between-cluster variance, binarizing the original image based on the threshold, and counting a total number of pixel points having different pixel values; and calculating the porosity by the following Formula: _ num^
[10] «“mi + «"mo(1)
[11] where pixels having the value of 0 in the binarized image represent a skeleton, and pixels having the value of 1 represent pores; in Formula (1), numi is the total number of pixel points having the pixel value of 1, and numo is the total number of pixel points having the pixel value of 0;
[12] step 2, determining neighborhood systems and conditional probability calculation formulas:
[13] determining neighborhood systems to be used on the basis of the resulting binarized image from step 1, listing all potential combinations for the neighborhood systems, and calculating a probability that a current pixel point has the value of 0 or 1 under the condition of each combination by the following conditional probability calculation formulas:
[14] calculating the probability that a pixel point has the value of 0 under a current neighborhood condition by the following Formula: =®\Xr’f numr^Q
[15] numr^+numr^ (2)
[16] calculating the probability that the pixel point has the value of 1 under the current neighborhood condition by the following Formula: V) numr^ num. + num,,. z x
[17] r^° (3)
[18] where N represents a combination in a neighborhood system, while xr a pixel point in the combination, X(ij) the current pixel point, nurm^o a count of xqj) being 0 under this combination condition, and numr,i a count of xqj) being 1;
[19] step 3, determining a reconstruction path and a conditional probability scanning direction:
[20] providing a blank image having the same size with the original image, calculating values for even-numbered rows in the blank image by iteration from right to left and calculating values for odd-numbered rows by iteration from left to right, and obtaining a first pixel point and a last pixel point in each row from pixel points of same columns in a previous row by a vertical two-neighborhood method, thereby obtaining a reconstruction path; and scanning the resulting binarized image from step 1 according to the reconstruction path to obtain the conditional probabilities in different directions;
[21] step 4, calculating assigned values by using Monte Carlo method:
[22] assigning the value of a first pixel point in the binarized image to a first pixel point in the blank image, followed by calculating assigned values for a first row from left to right by using a transverse two-neighborhood method and calculating the value of the last pixel in a second row by the vertical two-neighborhood method using the last pixel point of the first row as a condition, then performing matching calculation from the (n-l)th pixel point of the second row by using a leftward four-neighborhood method, performing iteration from left to right in the first row, and repeating the previous operations for subsequent rows alternately until the last pixel point is determined; and
[23] step 5, stopping looping until a porosity condition is met, thereby obtaining a reconstructed image:
[24] using the porosity as a condition for stopping looping, stopping looping when the porosity Pkxd of the reconstructed image and the porosity pkxd of the original image meet the following Formula, thus obtaining the reconstructed image;
[25] Pkxd-0.05<PKXD<Pkxd+0.05 (4)
[26] otherwise, providing a new blank image and repeating step 4 and step 5 until the Formula (4) is met, and finally obtaining the reconstructed image.
[27] The present disclosure further has the following technical features:
[28] Specifically, the resulting reconstructed image from step 5 is evaluated by a method including: calculating variation values of the binarized image and the reconstructed image by the following Formula, respectively.
[29] Formula for variation value calculation:
[31] where a is a step size, namely, a distance between two points, which ranges from 1 to 50; is the number of pairs of two points between which the distance is ; x represents the pixel value, and i represents pixel point coordinates;
[32] calculating an average distance value between variation function curves of the binarized image and the reconstructed image by the following Formula: 50 pjz = —--------
[33] 50 (6)
[34] where is the variation value of the reconstructed image, and r(a) is the variation value of the binarized image; a degree of fitting of the variation curves of the binarized image and the reconstructed image is characterized with data, and a difference between the variation values of the binarized image and the reconstructed image under a step size and an average value of the differences under different step sizes are calculated, with a smaller average value indicating a higher degree of fitting.
[35] The present disclosure further provides a MCMC based reconstruction system for a rock thin section, including:
[36] a module for calculating a porosity, the module being specifically configured to calculate a threshold of an original image by maximizing a between-cluster variance, binarize the original image based on the threshold, count a total number of pixel points having pixel values of 0 and 1, and calculate the porosity, where pixels having the value of 0 in the binarized image represent a skeleton, and pixels having the value of 1 represent pores;
[37] a module for determining neighborhood systems and conditional probabilities, the module being specifically configured to determine neighborhood systems on the basis of the binarized image, list all potential combinations for the neighborhood systems, and calculate a probability that a pixel has the value of 0 or 1 under the condition of each combination;
[38] a module for determining a reconstruction path and conditional probabilities in each direction, the module being specifically configured to calculate values for even-numbered rows in a blank image by iteration from right to left and calculate values for odd-numbered rows by iteration from left to right, obtain a first pixel point and a last pixel point in each row from pixel points of same columns in a previous row by a vertical two-neighborhood method, thereby obtaining a reconstruction path, and scan the resulting binarized image according to the reconstruction path to obtain the conditional probabilities in different directions;
[39] a module for calculating assigned values, the module being specifically configured to calculate each pixel value in the blank image by sequential iteration using Monte Carlo method;
[40] a module for acquiring a reconstructed image, the module being specifically configured to obtain the reconstructed image by using the porosity as a condition for stopping looping;
[41] a module for evaluating a resulting degree of fitting, the module being specifically configured to characterize a degree of fitting of variation curves of the binarized image and the reconstructed image with data, and calculate a difference between variation values of the binarized image and the reconstructed image under a step size and an average value of the differences under different step sizes, with a smaller average value indicating a higher degree of fitting.
[42] Compared with the prior art, the present disclosure has the following beneficial technical effects:
[43] According to the MCMC reconstruction method provided in the present disclosure, based on the characteristic of anisotropy that a rock image shows and considering that conditional probabilities are also anisotropic, the conditional probabilities in each direction are scanned. During iterative calculation of values, a conditional probability scanning direction is allowed to match a reconstruction path, and random numbers are selected to reconstruct an image so that effects in all directions are balanced. Thus, a degree of fitting of variation functions of a reconstructed image and an original image is increased while anisotropy is characterized. DETAILED DESCRIPTION OF THE EMBODIMENTS
[44] FIG. 1 shows an original image A of rock obtained by using a scanning electron microscope and a binarized image B.
[45] FIG. 2 shows transverse and vertical two neighborhoods (gray represents pixel points of neighborhood systems).
[46] FIG. 3 shows leftward and rightward four neighborhoods (gray represents pixel points of neighborhood systems).
[47] FIG. 4 is a schematic diagram of a serpentine reconstructed path.
[48] FIG. 5 shows reconstructed images by methods of the prior art and the present disclosure, respectively.
[49] FIG. 6 is a schematic diagram illustrating a fitted curve of variation functions.
[50] FIG. 7 is a flowchart according to the present disclosure.
[51] The specific embodiments of the present disclosure will be described in detail below. It should be understood that the specific embodiments described herein are merely intended to illustrate and explain the present disclosure rather than to limit the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS
[52] Based on the characteristic of anisotropy that a scanning electron microscopic image of a rock thin section shows, the present disclosure uses the idea of Markov chain and Monte Carlo sampling method, considers that conditional probabilities in different directions are anisotropic due to the characteristic information of the rock image, and reconstructs and displays an image by means of iteration calculation.
[53] Example 1:
[54] As shown in FIG. 1 to FIG. 7, this example provides an improved two-dimensional MCMC reconstruction method for a rock thin section. FIG. 7 is a flowchart according to the present disclosure. The method takes into account that conditional probabilities are anisotropic when determining a scanning mode for them. Accordingly, a conditional probability scanning direction is allowed to match a reconstruction path, and random numbers are selected to reconstruct an image so that effects in all directions are balanced. Thus, a degree of fitting of variation functions of a reconstructed image and an original image is increased while anisotropy is characterized. The method includes the following steps:
[55] Step 1, an original image is binarized, and a porosity is calculated.
[56] FIG. 1A shows an original image. For an original image having a size of m*n, a threshold of the image is calculated by maximizing a between-cluster variance, and the original image is binarized based on the threshold. FIG. IB shows a binarized rock image. A total number of pixel points having different pixel values is counted, and the porosity is calculated by the following Formula:
[57]
[58] _ num^ Pkxd numi + numQ Pixels having the value of 0 in the binarized image represent a skeleton, and pixels having the value of 1 represent pores. In Formula (1), numi is the total number of pixel points having the pixel value of 1, and numo is the total number of pixel points having the pixel value of 0. Specifically, in this example, the porosity of the binarized rock image of FIG. IB is calculated to be 17.68%.
[59] Step 2, neighborhood systems and conditional probability calculation formulas are determined.
[60] Neighborhood systems to be used are determined on the basis of the resulting binarized image from step 1. All potential combinations for the neighborhood systems are listed, and a probability that a current pixel point has the value of 0 or 1 under the condition of each combination is calculated by the following conditional probability calculation formulas:
[61] the probability that a pixel point has the value of 0 under a current neighborhood condition is calculated the following Formula: p{X(!,j)=Q\Xr’r^N) _ numr^Q num. + num,,. z x
[62] r^° (2)
[63] the probability that the pixel point has the value of 1 under the current neighborhood condition is calculated by the following Formula: numr^ numr^ + numr^ z x
[64] r^° (3)
[65] where N represents a combination in a neighborhood system, while xr a pixel point in the combination, X(ij) the current pixel point, nurm^o a count of xqj) being 0 under this combination condition, and num,, i a count of xqj) being 1.
[66] In this example, FIG. 2 shows transverse and vertical two neighborhoods, respectively, and the calculation conditional probability calculation formula thereof is as follows: / 4¾) = 01 xr ,r G ) p(x(i J) =1G, ,r e N2) numr^Q _ numr^ numr^Q + numr^ numr^Q + numr^
[68] where N2 represents a combination of two neighborhoods, while xr a pixel point in the combination, X(ij) the current pixel point, nurm^o a count of xqj) being 0 under this combination condition, and numr^i a count of xqj) being 1; and the left expression is used to calculate the probability that a pixel point has the value of 0 under this neighborhood condition, while the right expression is used to calculate the probability that a pixel point has the value of 1 under this neighborhood condition.
[69] FIG. 3 shows leftward and rightward four neighborhoods, respectively, and the calculation conditional probability calculation formula thereof is as follows: 4¾) = 01 , r e V4) p(xM = 11 xr, r e V4) numr^Q numr^ num. + num,,. num,... + num.
[70] r—>0 r—>1 r—>0 r—>1
[71] where N4 represents a combination of four neighborhoods, while xr a pixel point in the combination, X(ij) the current pixel point, nurm^o a count of xqj) being 0 under this combination condition, and numr^i a count of xqj) being 1; and the left expression is used to calculate the probability that a pixel point has the value of 0 under this neighborhood condition, while the right expression is used to calculate the probability that a pixel point has the value of 1 under this neighborhood condition. The relevant explanation is similar to that for two neighborhoods.
[72] Step 3, a reconstruction path and a conditional probability scanning direction are determined.
[73] FIG. 4 is a schematic diagram of a serpentine path. A blank image having the same size with the original image is provided. In the blank image, values for even-numbered rows are calculated by iteration from right to left and values for odd-numbered rows are calculated by iteration from left to right, and a first pixel point and a last pixel point in each row are obtained from pixel points of same columns in a previous row by a vertical two-neighborhood method, thereby obtaining a reconstruction path. The resulting binarized image from step 1 is scanned according to the reconstruction path to obtain the conditional probabilities in different directions.
[74] Step 4, assigned values are calculated by using Monte Carlo method.
[75] The value of a first pixel point in the binarized image is assigned to a first pixel point in the blank image. Subsequently, assigned values for a first row are calculated from left to right by using a transverse two-neighborhood method and the value of the last pixel in a second row is calculated by the vertical two-neighborhood method using the last pixel point of the first row as a condition. Then, matching calculation is performed from the (n-l)th pixel point of the second row by using a leftward four-neighborhood method, and iteration is performed from left to right in the first row. The previous operations are repeated for subsequent rows alternately until the last pixel point is determined.
[76] Step 5, looping is stopped until a porosity condition is met, thereby obtaining a reconstructed image.
[77] Using the porosity as a condition for stopping looping, looping is stopped when the porosity Pkxd of the reconstructed image and the porosity pkxd of the original image meet the following Formula; otherwise, a new blank image is provided, and step 4 and step 5 are repeated:
[78] Pkxd-0.05<PKXD<Pkxd+0.05 (4)
[79] Specifically, FIG. 5A shows a reconstructed image with conditional probabilities being obtained by scanning a binarized image once in the prior art, and FIG. 5B is a reconstructed image with conditional probabilities being the same as a reconstruction direction in this example. The two images have the porosities of 14.21% and 19.36%, respectively, both of which meet the condition for stopping looping.
[80] Step 6, variation values of the binarized image and the reconstructed image are calculated by the following Formula, respectively.
[81] Formula for variation value calculation: ] ^(«)
[82] (5)
[83] where a is a step size, namely, a distance between two points, which ranges from 1 to 50; is the number of pairs of two points between which the distance is ; x represents the pixel value, and i represents pixel point coordinates.
[84] An average distance value between variation function curves of the binarized image and the reconstructed image is calculated by the following Formula: 50 Xl«(a)-4a)l pjz = —--------
[85] 50 (6)
[86] where is the variation value of the reconstructed image, and is the variation value of the binarized image; a degree of fitting of the variation curves of the binarized image and the reconstructed image is characterized with data, and a difference between the variation values of the binarized image and the reconstructed image under a step size and an average value of the differences under different step sizes are calculated, with a smaller average value indicating a higher degree of fitting.
[87] FIG. 6 shows variation functions of three images, with dots representing variation values of the original image, the red line relating to FIG. 5 A of the prior art and the blue line relating to FIG. 5B of this example. It can be seen that the blue line fits the dots more. The average distances of the two lines to the dots are 0.0208 (FIG. 5A) and 0.0073 (FIG. 5B), respectively. A better image reconstruction effect can be achieved when the conditional probabilities in the same direction with the reconstruction path are used during reconstruction, namely, when the method of the present disclosure is used.
[88] Example 2:
[89] This example provides a MCMC based reconstruction system for a rock thin section, including:
[90] a module for calculating a porosity, the module being specifically configured to calculate a threshold of an original image by maximizing a between-cluster variance, binarize the original image based on the threshold, count a total number of pixel points having pixel values of 0 and 1, and calculate the porosity, where pixels having the value of 0 in the binarized image represent a skeleton, and pixels having the value of 1 represent pores;
[91] a module for determining neighborhood systems and conditional probabilities, the module being specifically configured to determine neighborhood systems on the basis of the binarized image, list all potential combinations for the neighborhood systems, and calculate a probability that a pixel has the value of 0 or 1 under the condition of each combination;
[92] a module for determining a reconstruction path and conditional probabilities in each direction, the module being specifically configured to calculate values for even-numbered rows in a blank image by iteration from right to left and calculate values for odd-numbered rows by iteration from left to right, obtain a first pixel point and a last pixel point in each row from pixel points of same columns in a previous row by a vertical two-neighborhood method, thereby obtaining a reconstruction path, and scan the resulting binarized image according to the reconstruction path to obtain the conditional probabilities in different directions;
[93] a module for calculating assigned values, the module being specifically configured to calculate each pixel value of the blank image by sequential iteration using Monte Carlo method;
[94] a module for acquiring a reconstructed image, the module being specifically configured to obtain the reconstructed image by using the porosity as a condition for stopping looping;
[95] a module for evaluating a resulting degree of fitting, the module being specifically configured to characterize a degree of fitting of variation curves of the binarized image and the reconstructed image with data, and calculate a difference between variation values of the binarized image and the reconstructed image under a step size and an average value of the differences under different step sizes, with a smaller average value indicating a higher degree of fitting.
Claims
1. An improved two-dimensional Markov chain Monte Carlo (MCMC) reconstruction method for a rock thin section, specifically comprising the following steps:step 1, binarizing an original image and calculating a porosity:for an original image having a size of m*n, calculating a threshold of the image by maximizing a between-cluster variance, binarizing the original image based on the threshold, and counting a total number of pixel points having different pixel values; and calculating the porosity by the following Formula:Pkxdnum^numx + numQ(1)wherein pixels having the value of 0 in the binarized image represent a skeleton, and pixels having the value of 1 represent pores; in Formula (1), numi is the total number of pixel points having the pixel value of 1, and numo is the total number of pixel points having the pixel value of 0;step 2, determining neighborhood systems and conditional probability calculation formulas:determining neighborhood systems to be used on the basis of the resulting binarized image from step 1, listing all potential combinations for the neighborhood systems, and calculating a probability that a current pixel point has the value of 0 or 1 under the condition of each combination by the following conditional probability calculation formulas:calculating the probability that a pixel point has the value of 0 under a current neighborhood condition by the following Formula:p^( / j) = 0|^,rGV)numr^numr^ + numr^ (2)calculating the probability that the pixel point has the value of 1 under the current neighborhood condition by the following Formula:p(x(v) = l|xr,re V)num^numr^0+numr^ (3)wherein N represents a combination in a neighborhood system, while xr represents a pixel point in the combination, X(ij) represents the current pixel point, nurm^o represents a count of X(ij), being 0 under this combination condition, and numr .i represents a count of xqj), being 1;step 3, determining a reconstruction path and a conditional probability scanning direction:providing a blank image having the same size with the original image, calculating values for even-numbered rows in the blank image by iteration from right to left and calculating values for odd-numbered rows by iteration from left to right, and obtaining a first pixel point and a last pixel point in each row from pixel points of same columns in a previous row by a vertical two-neighborhood method, thereby obtaining a reconstruction path; and scanning the resulting binarized image from step 1 according to the reconstruction path to obtain the conditional probabilities in different directions;step 4, calculating assigned values by using Monte Carlo method:assigning the value of a first pixel point in the binarized image to a first pixel point in the blank image, followed by calculating assigned values for a first row from left to right by using a transverse two-neighborhood method and calculating the value of the last pixel in a second row by the vertical two-neighborhood method using the last pixel point of the first row as a condition, then performing matching calculation from the (n-l)th pixel point of the second row by using a leftward four-neighborhood method, performing iteration from left to right in the first row, and repeating the previous operations for subsequent rows alternately until the last pixel point is determined; andstep 5, stopping looping until a porosity condition is met, thereby obtaining a reconstructed image:using the porosity as a condition for stopping looping, stopping looping when the porosity Pkxd of the reconstructed image and the porosity pkxd of the original image meet the following Formula, thus obtaining the reconstructed image;Pkxd-0.05<PKXD<Pkxd+0.05 (4)otherwise, providing a new blank image and repeating step 4 and step 5 until the Formula (4) is met, and finally obtaining the reconstructed image.
2. The improved two-dimensional MCMC reconstruction method for a rock thin section according to claim 1, wherein the resulting reconstructed image from step 5 is evaluated by a method comprising: calculating variation values of the binarized image and the reconstructed image by the following Formula, respectively:। ^(«)2 * V(a) tt (5)wherein is a step size, namely, a distance between two points, which ranges from 1 to 50;is the number of pairs of two points between which the distance is ; x represents thepixel value, and i represents pixel point coordinates;calculating an average distance value between variation function curves of the binarized image and the reconstructed image by the following Formula:Pjz =50q=l________________50 (6)wherein w7 is the variation value of the reconstructed image, and r\aJ is the variation value of the binarized image; a degree of fitting of the variation curves of the binarized image and the reconstructed image is characterized with data, and a difference between the variation values of the binarized image and the reconstructed image under a step size and an average value of the differences under different step sizes are calculated, with a smaller average value indicating a higher degree of fitting.
3. An improved MCMC reconstruction system for a rock thin section, comprising:a module for calculating a porosity, the module being specifically configured to calculate a threshold of an original image by maximizing a between-cluster variance, binarize the original image based on the threshold, count a total number of pixel points having pixel values of 0 and 1, and calculate the porosity, wherein pixels having the value of 0 in the binarized image represent a skeleton, and pixels having the value of 1 represent pores;a module for determining neighborhood systems and conditional probabilities, the module being specifically configured to determine neighborhood systems on the basis of the binarized image, list all potential combinations for the neighborhood systems, and calculate a probability that a pixel has the value of 0 or 1 under the condition of each combination;a module for determining a reconstruction path and conditional probabilities in each direction, the module being specifically configured to calculate values for even-numbered rows in a blank image by iteration from right to left and calculate values for odd-numbered rows by iteration from left to right, obtain a first pixel point and a last pixel point in each row from pixelpoints of same columns in a previous row by a vertical two-neighborhood method, thereby obtaining a reconstruction path, and scan the resulting binarized image according to the reconstruction path to obtain the conditional probabilities in different directions;a module for calculating assigned values, the module being specifically configured to calculate each pixel value in the blank image by sequential iteration using Monte Carlo method;a module for acquiring a reconstructed image, the module being specifically configured to obtain the reconstructed image by using the porosity as a condition for stopping looping;a module for evaluating a resulting degree of fitting, the module being specifically configured to characterize a degree of fitting of variation curves of the binarized image and the reconstructed image with data, and calculate a difference between variation values of the binarized image and the reconstructed image under a step size and an average value of the differences under different step sizes, with a smaller average value indicating a higher degree of fitting.
Citation Information
Patent Citations
Markov Chain Monte Carlo automatic history matching method and system based on t-distribution
CN106355003A
Organ image segmentation method and device
CN110874843A
Improved Markov chain Monte Carlo two-dimensional rock slice reconstruction method and improved Markov chain Monte Carlo two-dimensional rock slice reconstruction system
CN111784724A
Numerical method of estimating physical properties of three-dimensional porous media
WO2001077993A1
