Relative total variation turbine blade CT image scattering correction method fused with wavelet transform

Abstract

The invention relates to a relative total variation turbine blade CT image scattering correction method fused with wavelet transform, and belongs to the technical field of industrial CT image scattering correction. The method comprises the following steps: obtaining a CBCT image and a label CT image; obtaining a projection sinogram corresponding to the image; performing wavelet decomposition on the projection sinogram to obtain an approximate component, a horizontal high-frequency component, a vertical high-frequency component and a diagonal high-frequency component; subtracting the approximate components of the projection sinogram to obtain an approximate component residual image; processing the approximate component residual image by adopting a relative total variation low-pass filtering algorithm to obtain a scattering component; subtracting the estimated scattering component from a projection sinogram approximate component corresponding to the CBCT image to obtain a corrected approximate component; and performing wavelet reconstruction on the corrected approximate component and a high-frequency component obtained by decomposing a projection sinogram corresponding to the CBCT image to obtain a corrected projection sinogram, and reconstructing to obtain a corrected CBCT image.

Description

Technical Field

[0001] The invention belongs to the technical field of industrial CT image scattering correction, and relates to a relative total variation turbine blade CT image scattering correction method integrated with wavelet transform. Background Art

[0002] Aircraft engines, the devices that power aircraft, are known as the "crown jewels of the manufacturing industry." Turbine blades, core components of aircraft engines, are crucial for aircraft flight safety. Turbine blades operate in the harshest environments of aircraft engines, characterized by high temperatures, high pressures, and high rotational speeds, making them susceptible to damage. Ensuring the integrity and reliability of turbine blades is crucial for ensuring aircraft engine safety and quality.

[0003] Due to the complex internal structure of turbine blades, the manufacturing process can lead to tiny defects such as holes, slag inclusions, and air holes. Operating under high temperatures, pressures, and high engine speeds, various regions of the blades are subjected to complex stresses, resulting in fatigue, overstress, creep, corrosion, and wear, which can further lead to defects such as microcracks. Whether surface or internal defects exist in turbine blades, they pose a potentially high risk to aircraft engines and further endanger flight safety.

[0004] Industrial computed tomography (CT) technology plays an important role in applications such as turbine blade defect detection and wall thickness measurement. It has replaced traditional "scribing and sectioning" and has become the primary detection technology for non-destructive testing and non-destructive evaluation of blade quality. Cone-beam CT has the advantages of fast imaging speed and high precision. When performing cone-beam CT scanning on turbine blades with high material density and complex shapes, factors such as X-ray hardening and Compton scattering can cause the blade CT reconstructed image to have problems such as uneven grayscale, low contrast, and blurred edges, which seriously affect the evaluation and processing of its internal structure.

[0005] With the maturity and application of area array detector technology, X-ray scattering has become a significant obstacle to high-quality CT imaging and a significant challenge. CT scatter artifact correction methods are primarily categorized into four types: hardware correction, software correction, combined hardware and software correction, and deep learning algorithms.

[0006] Due to the wide variety and complex structures of workpieces used in actual industrial production and applications, hardware-based scatter correction is often limited in effectiveness. While Monte Carlo simulation can accurately simulate scatter signals, it is time-consuming and requires prior knowledge of the object's structure, making it unsuitable for practical industrial production. Existing deep learning algorithms are primarily targeted at correcting scatter artifacts in medical cone-beam CT images. However, due to the wide variety and complex structures of industrial CT objects, it is difficult to directly transfer these algorithms to the task of correcting scatter artifacts in industrial CT images. Summary of the Invention

[0007] In view of this, the purpose of the present invention is to provide a scatter correction method based on label CT as prior information, combined with wavelet transform and relative total variation low-pass filtering algorithm. This method can quickly and efficiently estimate the scatter signal in the image, thereby effectively removing scatter artifacts.

[0008] In order to achieve the above object, the present invention provides the following technical solutions:

[0009] A relative total variation turbine blade CT image scatter correction method integrated with wavelet transform includes the following steps:

[0010] S1: Acquire cone beam computed tomography (CBCT) images;

[0011] S2: Get labeled CT image;

[0012] S3: Obtain the projection sinogram corresponding to the CBCT image and the labeled CT image;

[0013] S4: Perform wavelet decomposition on the projection sinusoids of the CBCT image and the label CT image to obtain the approximate component LL, the horizontal high-frequency component LH, the vertical high-frequency component HL, and the diagonal high-frequency component HH;

[0014] S5: subtracting the approximate component of the projection sinusoidal graph to be corrected from the approximate component of the projection sinusoidal graph corresponding to the labeled CT image to obtain an approximate component residual map;

[0015] S6: The approximate component residual map is processed using a relative total variation low-pass filtering algorithm to obtain the scattered component;

[0016] S7: Subtracting the estimated scattered component from the projection sinusoidal approximate component corresponding to the CBCT image to obtain the corrected approximate component;

[0017] S8: The high-frequency components obtained by decomposing the corrected approximate components and the projection sinogram corresponding to the CBCT image are subjected to wavelet reconstruction to obtain a corrected projection sinogram, and the corrected CBCT image is obtained through reconstruction.

[0018] Furthermore, in step S1, a high-energy cone-beam CT device is used to scan turbine blades of different styles to obtain CBCT images.

[0019] Furthermore, step S2 specifically includes the following steps:

[0020] The CLAHE enhancement algorithm is used to enhance CBCT images to solve the problem of uneven image grayscale;

[0021] The processed images were used to obtain the leaf contour using RSF image segmentation algorithm or labelme annotation software;

[0022] The pixel values inside the leaf contour are retained and the pixel values outside are set to 0 to obtain the labeled CT image.

[0023] Furthermore, in step S3, an equidistant fan beam projection algorithm is used to obtain projection sinograms corresponding to the CBCT image and the label CT image.

[0024] Further, step S4 specifically includes the following steps:

[0025] First, define a scale and translation basis function:

[0026]

[0027]

[0028] Where superscript i represents H, V, and D, which represent the horizontal, vertical, and diagonal components respectively; the discrete wavelet transform of an image f(x, y) of size M×N is:

[0029]

[0030]

[0031] The approximately symmetrical compact support orthogonal wavelet sym4 is selected for decomposition, and the decomposition level is set to 1 to obtain the approximate components (ie LL1), horizontal high-frequency component (i.e. LH1), vertical high-frequency component (i.e. HL1) and diagonal high frequency components (i.e. HH1).

[0032] Further, in step S5, the approximate component P of the projected sinusoidal graph to be corrected is obtained. CBCT_LL1 Subtract the approximate component P of the projection sinusoid corresponding to the label CT image label_LL1 , and obtain the approximate component residual graph ΔP LL1 :

[0033] ΔP LL1 =PCBCT_LL1 -P label_LL1 .

[0034] Furthermore, in step S6, the approximate component residual map ΔP LL1 Estimate the scattered component S using a relative total variation low-pass filtering algorithm LL1 , a quadratic penalty term is used to strengthen the similarity of low-frequency information between input and output, expressed as:

[0035]

[0036] Where I is the input, p is the two-dimensional pixel index; S is the obtained low-frequency structure image, and the data fidelity term (S p -I p ) 2 This is to ensure that the decomposed low-frequency structure is similar to the structural information in the input image; p |(▽S) p | is the TV regularization operator, denoted as:

[0037]

[0038] Where, and are the partial derivatives in two directions;

[0039] The objective function of the low-pass filtering algorithm relative to total variation is expressed as:

[0040]

[0041] Among them (S p -I p ) 2 is the data fidelity term, relative to the total variation regularization term (D x (p) / (L x (p)+ε)+D y (p) / (L y (p)+ε)) introduces the effect of removing high-frequency texture details of the image; λ is the constraint weight, and ε is a small positive number to ensure that the denominator is not 0;

[0042] D is a general pixel-by-pixel window total variation, denoted as:

[0043]

[0044]

[0045] Where: q belongs to R(p), which is the rectangular area centered on pixel p; D x (p) and D y(p) is the windowed total variation of pixel p in the x and y directions, which calculates the absolute spatial difference within the window R(p), g p,q It is a weighting function defined according to spatial correlation and is expressed as:

[0046]

[0047] Where: σ controls the spatial scale of the window;

[0048] The new windowed intrinsic transform L is used to capture the overall spatial variation and is expressed as:

[0049]

[0050]

[0051] Furthermore, in step S6, based on an objective function with a quadratic measure penalty, the nonlinear part of the objective function of the relative total variation low-pass filtering algorithm is transformed into solving a series of linear equation systems, including:

[0052] Expand the penalty term in the x direction to:

[0053]

[0054] To include Rearrange the items and group the elements to get:

[0055]

[0056] The second line in the above formula is due to the introduction of ε s The approximation is obtained by numerical stability; rearrange the x-direction penalty term to decompose it into a quadratic term and the nonlinear part u xq w xq , which is expressed as follows:

[0057]

[0058]

[0059] The above formula represents the u of each pixel x In fact, it contains the adjacent gradient information in an isotropic spatial filtering manner; G σ is a Gaussian filter with a standard deviation of σ; the division is pixel-wise, * is the convolution operator, and w x It is only related to the gradient of the pixel;

[0060] The penalty term in the y direction is similarly expressed as:

[0061]

[0062] In the formula is the square of the partial derivative of the y component, u yq w yq is the corresponding nonlinear part;

[0063]

[0064]

[0065] The objective function is expressed in matrix form:

[0066]

[0067] Where: v S and v I Represents the vector representation of S and I respectively; C x and C y are the Toeplitz matrices of the forward difference discrete gradient operator, U x 、U y 、W x and W y is a diagonal matrix whose diagonal values are U x [i,i]=u xi , U y [i,i]=u yi , W x [i,i]=w xi , W y [i,i]=w yi ;

[0068] The optimization process in the matrix form of the objective function is:

[0069] (1) From the structural information image S estimated in the previous iteration, the values of the corresponding directions u and w are calculated according to the nonlinear parts of the penalty terms in the x and y directions, thereby forming the matrices U and W in the matrix form expression of the objective function;

[0070] (2) Using U x 、U y 、W x and W y The value of , which is minimized at each iteration, boils down to solving a linear system:

[0071]

[0072] Where E is the unit matrix, Based on the structure vector The calculated weight matrix;

[0073] The matrix form of the objective function is iterated multiple times, and the low-frequency structure image S is quickly updated during the iteration.

[0074] Further, in step S7, the projection sinusoidal graph P to be corrected is CBCT_LL1 Subtract the estimated scattered component S LL1 Get the corrected approximate component P cor_LL1 :

[0075] P cor_LL1 =P CBCT_LL1 -P scatter_LL1

[0076] Furthermore, in step S8, the two-dimensional image f(x,y) is reconstructed by inverse discrete wavelet transform to obtain:

[0077]

[0078] Corrected projected sinogram P cor By P cor_LL1 、P CBCT_LH1 、P CBCT_HL1 、P CBCT_HH1 The CT image is reconstructed by wavelet and then reconstructed by equidistant fan beam back projection algorithm to obtain the corrected CT image.

[0079] The present invention utilizes labeled CT images as prior information to effectively estimate the low-frequency scatter components in CBCT images using a relative total variation low-pass filtering algorithm. Furthermore, combined with a wavelet transform, this method enables even more accurate estimation of the scatter signal, enabling rapid and efficient removal of scatter artifacts. This method significantly improves image quality when processing complex industrial CT images, such as those of turbine blades.

[0080] Other advantages, objects, and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art upon examination of the following description or may be learned from practice of the present invention. The objects and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below with reference to the accompanying drawings, in which:

[0082] Figure 1 Flowchart of the relative total variation turbine blade CT image scatter correction method integrated with wavelet transform;

[0083] Figure 2 It is wavelet decomposition and reconstruction;

[0084] Figure 3Comparison of CT images of turbine blades with scattering artifacts before and after correction, where (a) is CBCT, (b) is label, (c) is RTV-LPF, and (d) is WT-RTV. DETAILED DESCRIPTION

[0085] The following describes the embodiments of the present invention by means of specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0086] It should be noted that the illustrations provided in the following embodiments are merely schematic illustrations of the basic concept of the present invention. Therefore, the illustrations only show components related to the present invention and are not drawn according to the number, shape, and size of components in actual implementation. In actual implementation, the type, quantity, and proportion of each component may be changed arbitrarily, and the component layout may also be more complex.

[0087] In the following description, numerous details are discussed to provide a more thorough explanation of the embodiments of the present invention. However, it will be apparent to those skilled in the art that the embodiments of the present invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring the embodiments of the present invention.

[0088] Example 1:

[0089] The present invention provides a relative total variation turbine blade CT image scatter correction method integrated with wavelet transform, comprising the following steps:

[0090] 1) Acquire CBCT images;

[0091] 2) Obtaining labeled CT images;

[0092] 3) Obtaining the projection sinogram corresponding to the CBCT image and the labeled CT image;

[0093] 4) Input the projection sinogram corresponding to the CBCT and label CT images, set the projection sinogram to perform wavelet decomposition to obtain the approximate component (LL), horizontal high-frequency component (LH), vertical high-frequency component (HL), and diagonal high-frequency component (HH);

[0094] 5) Subtract the approximate component (LL) of the projected sinusoidal graph to obtain the approximate component residual graph;

[0095] 6) The approximate component residual map is processed using a relative total variation low-pass filtering algorithm to obtain the scattered component;

[0096] 7) Subtracting the estimated scattered component from the projection sinusoidal approximate component corresponding to the CBCT image to obtain the corrected approximate component;

[0097] 8) Performing wavelet reconstruction on the corrected approximate component and the high-frequency component obtained by decomposing the CBCT projection sinogram to obtain a corrected projection sinogram, and reconstructing the corrected CBCT image.

[0098] Example 2:

[0099] like Figure 1 As shown, in this embodiment, the relative total variation turbine blade CT image scatter correction method fused with wavelet transform specifically includes the following steps:

[0100] 1) A high-energy cone-beam CT scanner was used to scan turbine blades of various designs to obtain CBCT images. The acquired CBCT images were enhanced using the CLAHE enhancement algorithm to address grayscale inhomogeneity. The enhanced images were first segmented using the RSF image segmentation algorithm to obtain the blade contours. If the segmentation of weak edge regions was poor, the contours were manually extracted using the LabelMe annotation software. The blade contours obtained using the two methods were labeled by retaining the internal pixel values and setting the external pixel values to 0.

[0101] 2) Using the equidistant fan beam projection algorithm to obtain the projection sinusoidal graph P corresponding to the CBCT image and the label CT image CBCT 、P label ;

[0102] 3) Input the projection sinogram corresponding to the CBCT and label CT images, set the projection sinogram to perform wavelet decomposition to obtain the approximate component (LL), horizontal high-frequency component (LH), vertical high-frequency component (HL), and diagonal high-frequency component (HH);

[0103] First, define a scale and translation basis function:

[0104]

[0105]

[0106] Where superscript i represents H, V, and D, which represent the horizontal, vertical, and diagonal components respectively; the discrete wavelet transform of an image f(x, y) of size M×N is:

[0107]

[0108]

[0109] The wavelet function uses the approximately symmetrical compact support orthogonal wavelet sym4 to decompose, and the decomposition level is set to 1 to obtain (i.e. LL1), (i.e. LH1), (i.e. HL1), (i.e. HH1).

[0110] 4) The projection sinusoidal graph LL1 (ie P CBCT_LL1 ) minus the projection sinogram LL1 (i.e., P label_LL1 ), and obtain the approximate component residual graph ΔP LL1 :

[0111] ΔP LL1 =P CBCT_LL1 -P label_LL1

[0112] Residual graph of approximate components ΔP LL1 Estimate the scattered component S using a relative total variation low-pass filtering algorithm LL1 The high-frequency information and low-frequency information of an image, i.e., texture details and main structure, exhibit completely different properties, making it decomposable. A representative decomposition method is to use a total variation regularization term to retain the main structural information and remove high-frequency texture details. A quadratic penalty term is used to strengthen the similarity of low-frequency information between the input and output, expressed as:

[0113]

[0114] Where I is the input, p is the two-dimensional pixel index, S is the obtained low-frequency structure image, and the data fidelity term is to ensure that the decomposed low-frequency structure is similar to the structural information in the input image. p |(▽S) p | is the TV regularization operator, denoted as:

[0115]

[0116] Where, and are the partial derivatives in two directions. The objective function of the low-pass filtering algorithm relative to total variation is expressed as:

[0117]

[0118] In order to further enhance the contrast between texture details and structures, a new penalty term is introduced, which combines the new windowed intrinsic transformation L and the universal pixel-by-pixel window total variation D to form a more effective regularizer for structure-texture decomposition. p -I p ) 2 is the data fidelity term, the new regularization term (D x (p) / (L x (p)+ε)+D y (p) / (L y The (p)+ε)) introduces the effect of removing high-frequency texture details in the image, which is called the relative total variation low-pass filter algorithm (RTV-lowpass filter, RTV-LPF). λ is the constraint weight, and ε is a small positive number that ensures that the denominator is not zero. A general pixel-by-pixel window total variation D is denoted as:

[0119]

[0120]

[0121] Where: q belongs to R(p), which is the rectangular area centered on pixel p. and are the partial derivatives in two directions. x (p) and D y (p) is the windowed total variation of pixel p in the x and y directions, which calculates the absolute spatial difference within the window R(p), g p,q It is a weighting function defined according to spatial correlation and is expressed as:

[0122]

[0123] Where: σ controls the spatial scale of the window.

[0124] The new windowed intrinsic transform L is used to capture the overall spatial variation and is expressed as:

[0125]

[0126]

[0127] 5) The objective function of the total variation low-pass filtering algorithm is highly non-convex. By introducing an efficient solver, the objective function can be linearly optimized based on a quadratic penalty. The nonlinear part of the problem can then be transformed into solving a series of linear equations, similar to iterative reweighted least squares.

[0128] Expand the penalty term in the x direction to:

[0129]

[0130] To include Reorganizing the items and grouping the elements yields:

[0131]

[0132]

[0133] The second line in the above formula is due to the introduction of ε s The numerical stability of the approximation is obtained by rearranging the x-direction penalty term into a quadratic term and the nonlinear part u xq w xq The specific expressions are as follows:

[0134]

[0135]

[0136] The above formula shows that the u of each pixel is x In fact, it contains the adjacent gradient information in an isotropic spatial filtering manner. σ is a Gaussian filter with a standard deviation of σ. The division is pixel-wise, * is the convolution operator, and w x It is only related to the gradient of the pixel. Similarly, the penalty term in the y direction can be expressed as:

[0137]

[0138] Where: is the square of the partial derivative of the y component, u yq w yq is the corresponding nonlinear part.

[0139]

[0140]

[0141] The objective function can be expressed in matrix form:

[0142]

[0143] Where: v S and v I Represent the vector representation of S and I respectively. x and C y are the Toeplitz matrices of the forward difference discrete gradient operator, U x 、U y 、W x and W yIs a diagonal matrix. Its diagonal values are U x [i,i]=u xi , U y [i,i]=u yi , W x [i,i]=w xi , W y [i,i]=w yi .

[0144] The optimization process in the matrix form of the objective function is:

[0145] (1) From the structural information image S estimated in the previous iteration, the values of the corresponding directions u and w can be calculated respectively according to the nonlinear parts in the penalty terms in the x and y directions, thus forming the matrices U and W in the matrix form expression of the objective function.

[0146] (2) Using U x 、U y 、W x and W y The value of , which is minimized at each iteration, boils down to solving a linear system:

[0147]

[0148] Where E is the unit matrix, Based on the structure vector The calculated weight matrix.

[0149] The objective function matrix is iterated multiple times, and the low-frequency structure image S is quickly updated in the iteration. The number of iterations is set to 5, the constraint weight λ is set to 0.005, and the standard deviation σ is set to 1. After 5 iterations, the low-frequency structure image S, that is, the scattering component S LL1 .

[0150] 6) Based on the projection sinusoidal graph LL1 (P CBCT_LL1 ) minus the estimated scattered component S LL1 Get the corrected approximate component P cor_LL1 :

[0151] P cor_LL1 =P CBCT_LL1 -P scatter_LL1

[0152] 7) The two-dimensional image f(x,y) can be reconstructed by inverse discrete wavelet transform:

[0153]

[0154] Corrected projected sinogram P cor By P cor_LL1 、PCBCT_LH1 、P CBCT_HL1 、P CBCT_HH1 Wavelet reconstruction is obtained.

[0155] Wavelet decomposition and reconstruction process Figure 2 shown.

[0156] 8) The equidistant fan beam back-projection algorithm is used to reconstruct the corrected CT image.

[0157] Example 3:

[0158] To verify the effectiveness of the present invention, this example evaluated the correction results based on both subjective and objective evaluation indicators. The experimental computer hardware configuration included a 13th Gen Intel(R) Core(TM) i5-13600KF @ 3.50GHz processor, 256GB of RAM, and an NVIDIA GeForce RTX 4090 graphics card. The software environment included MATLAB R2022a. The experimental system was a cone-beam CT system.

[0159] The present invention selects the approximately symmetrical compact support orthogonal wavelet sym4 for decomposition, sets the decomposition layer number to 1, the iteration number to 5, the constraint weight λ=0.005, and the weighted space scale σ=1.

[0160] In order to verify the scatter correction performance of the present invention on turbine blade CT images with scatter artifacts, different styles of turbine blade CT images were selected for experiments. The results are shown in the figure below. Figure 3 As shown in (a)-(d) in the figure. The corrected images are compared with the uncorrected images, and the overall effect is improved. Specifically, there are significant scattering artifacts in the leaf basin area of the image before correction. In particular, in the uncorrected image of leaf sample 3, the cavity outline is blurred and the grayscale distribution inside the cavity is uneven. In comparison, the scattering artifacts in the leaf basin area of the corrected image are effectively suppressed, the cavity outline is more clearly visible, and the grayscale distribution inside the leaf is more uniform, significantly improving the overall image quality.

[0161] Furthermore, the correction results were quantitatively compared. Three objective metrics, namely mean absolute error (MAE), peak signal-to-noise ratio (PSNR), and structural similarity (SSIM), were used to verify the effectiveness of the scatter correction proposed in this paper. As shown in Table 1, the proposed model improved the MAE by 51.33%, 36.04%, and 17.64% for leaf samples 1 to 3 after correction; the PSNR values increased by 27.89%, 15.91%, and 17.37%, respectively; and the SSIM values increased by 0.85%, 2.01%, and 2.15%, respectively.

[0162] Table 1

[0163]

[0164]

[0165] In comparison, the model of the present invention performs better in removing the scattering effect of the overall structural contour of the blade.

[0166] In summary, this method can effectively suppress scattering artifacts in leaf CT images, particularly those in the leaf basin. After correction, the outline of the leaf cavity is clearly visible, and the grayscale distribution inside the leaf is more uniform, significantly improving the overall image quality.

[0167] In the above embodiments, references to "this embodiment" in the specification indicate that a particular feature, structure, or characteristic described in conjunction with the embodiment is included in at least some embodiments, but not necessarily all embodiments. Multiple occurrences of "this embodiment" do not necessarily refer to the same embodiment.

[0168] In the above embodiments, although the invention has been described in conjunction with specific embodiments thereof, many alternatives, modifications, and variations of these embodiments will be apparent to those skilled in the art based on the foregoing description. For example, other memory structures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed. The embodiments of the present invention are intended to encompass all such alternatives, modifications, and variations that fall within the broad scope of the appended claims.

[0169] This embodiment further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, any one of the methods in this embodiment is implemented.

[0170] This embodiment also provides an electronic terminal, including: a processor and a memory;

[0171] The memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the terminal executes any one of the methods in this embodiment.

[0172] Regarding the computer-readable storage medium in this embodiment, those skilled in the art will appreciate that all or part of the steps in the aforementioned method embodiments can be implemented using hardware associated with the computer program. The aforementioned computer program can be stored in a computer-readable storage medium. When executed, the program performs the steps in the aforementioned method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0173] The electronic terminal provided in this embodiment includes a processor, a memory, a transceiver and a communication interface. The memory and the communication interface are connected to the processor and the transceiver and complete communication with each other. The memory is used to store computer programs, the communication interface is used for communication, and the processor and the transceiver are used to run computer programs so that the electronic terminal executes the various steps of the above method.

[0174] In this embodiment, the memory may include a random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage.

[0175] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, and discrete hardware components.

[0176] The present invention can be used in a wide variety of general-purpose or special-purpose computing system environments or configurations, such as personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments that include any of the above.

[0177] The present invention may be described in the general context of computer-executable instructions, such as program modules, executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. The present invention may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media, including storage devices.

[0178] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for scatter correction of turbine blade CT images using relative total variation (RTV) wavelet transform, characterized by: The following steps are involved: S1: Acquire CBCT images; S2: Get labeled CT image; S3: Obtain the projection sinogram corresponding to the CBCT image and the labeled CT image; S4: Perform wavelet decomposition on the projection sinusoids of the CBCT image and the label CT image to obtain the approximate component LL, the horizontal high-frequency component LH, the vertical high-frequency component HL, and the diagonal high-frequency component HH; S5: subtracting the approximate component of the projection sinusoidal graph to be corrected from the approximate component of the projection sinusoidal graph corresponding to the labeled CT image to obtain an approximate component residual map; S6: The approximate component residual map is processed using a relative total variation low-pass filtering algorithm to obtain the scattered component; S7: Subtracting the estimated scattered component from the projection sinusoidal approximate component corresponding to the CBCT image to obtain the corrected approximate component; S8: The high-frequency components obtained by decomposing the corrected approximate components and the projection sinogram corresponding to the CBCT image are subjected to wavelet reconstruction to obtain a corrected projection sinogram, and the corrected CBCT image is obtained through reconstruction.

2. The method for scatter correction of turbine blade CT images based on relative total variation and wavelet transform according to claim 1, characterized in that: In step S1, a high-energy cone-beam CT device is used to scan turbine blades of different styles to obtain CBCT images.

3. The method for scatter correction of turbine blade CT images based on relative total variation and wavelet transform according to claim 1, characterized in that: Step S2 specifically includes the following steps: The CLAHE enhancement algorithm is used to enhance CBCT images to solve the problem of uneven image grayscale; The processed images were used to obtain the leaf contour using RSF image segmentation algorithm or labelme annotation software; The pixel values inside the leaf contour are retained and the pixel values outside are set to 0 to obtain the labeled CT image.

4. The method for scatter correction of turbine blade CT images based on relative total variation and wavelet transform according to claim 1, characterized in that: In step S3, an equidistant fan beam projection algorithm is used to obtain projection sinograms corresponding to the CBCT image and the label CT image.

5. The method for scatter correction of turbine blade CT images based on relative total variation and wavelet transform according to claim 1, characterized in that: Step S4 specifically includes the following steps: First, define a scale and translation basis function: Where superscript i represents H, V, and D, which represent the horizontal, vertical, and diagonal components respectively; the discrete wavelet transform of an image f(x, y) of size M×N is: The approximately symmetrical compact support orthogonal wavelet sym4 is selected for decomposition, and the decomposition level is set to 1 to obtain the approximate components Horizontal high-frequency component Vertical high-frequency component and diagonal high-frequency components 6. The method for scatter correction of turbine blade CT images using relative total variation and wavelet transform according to claim 1, characterized in that: In step S5, the approximate component P of the projected sinusoidal graph to be corrected is CBCT_LL1 Subtract the approximate component P of the projection sinusoid corresponding to the label CT image label_LL1 , and obtain the approximate component residual graph ΔP LL1 : ΔP LL1 =P CBCT_LL1 -P label_LL1 。 7. The method for scatter correction of turbine blade CT images based on relative total variation and wavelet transform according to claim 1, characterized in that: In step S6, the approximate component residual map ΔP LL1 Estimate the scattered component S using a relative total variation low-pass filtering algorithm LL1 , a quadratic penalty term is used to strengthen the similarity of low-frequency information between input and output, expressed as: Where I is the input, p is the two-dimensional pixel index; S is the obtained low-frequency structure image, and the data fidelity term (S p -I p ) 2 This is to ensure that the decomposed low-frequency structure is similar to the structural information in the input image; is the TV regularization operator, denoted as: Where, and are the partial derivatives in two directions; The objective function of the low-pass filtering algorithm relative to total variation is expressed as: Among them (S p -I p ) 2 is the data fidelity term, relative to the total variation regularization term (D x (p) / (L x (p)+ε)+D y (p) / (L y (p)+ε)) introduces the effect of removing high-frequency texture details of the image; λ is the constraint weight, and ε is a small positive number to ensure that the denominator is not 0; D is a general pixel-by-pixel window total variation, denoted as: Where: q belongs to R(p), which is the rectangular area centered on pixel p; D x (p) and D y (p) is the windowed total variation of pixel p in the x and y directions, which calculates the absolute spatial difference within the window R(p), g p,q It is a weighting function defined according to spatial correlation and is expressed as: Where: σ controls the spatial scale of the window; The new windowed intrinsic transform L is used to capture the overall spatial variation and is expressed as:

8. The method for scatter correction of turbine blade CT images using relative total variation and wavelet transform according to claim 7, characterized in that: In step S6, based on an objective function with a quadratic measure penalty, the nonlinear part of the objective function of the relative total variation low-pass filtering algorithm is transformed into solving a series of linear equation systems, including: Expand the penalty term in the x direction to: To include Rearrange the items and group the elements to get: The second line in the above formula is due to the introduction of ε s The approximation is obtained by numerical stability; rearrange the x-direction penalty term to decompose it into a quadratic term and the nonlinear part u xq w xq , which is expressed as follows: The above formula represents the u of each pixel x In fact, it contains the adjacent gradient information in an isotropic spatial filtering manner; G σ is a Gaussian filter with a standard deviation of σ; the division is pixel-wise, * is the convolution operator, and w x It is only related to the gradient of the pixel; The penalty term in the y direction is similarly expressed as: In the formula is the square of the partial derivative of the y component, u yq w yq is the corresponding nonlinear part; The objective function is expressed in matrix form: Where: v S and v I Represents the vector representation of S and I respectively; C x and C y are the Toeplitz matrices of the forward difference discrete gradient operator, U x 、U y 、W x and W y is a diagonal matrix whose diagonal values are U x [i,i]=u xi , U y [i,i]=u yi , W x [i,i]=w xi , W y [i,i]=w yi ; The optimization process in the matrix form of the objective function is: (1) From the structural information image S estimated in the previous iteration, the values of the corresponding directions u and w are calculated according to the nonlinear parts of the penalty terms in the x and y directions, thereby forming the matrices U and W in the matrix form expression of the objective function; (2) Using U x 、U y 、W x and W y The value of , which is minimized at each iteration, boils down to solving a linear system: Where E is the unit matrix, Based on the structure vector The calculated weight matrix; The matrix form of the objective function is iterated multiple times, and the low-frequency structure image S is quickly updated during the iteration.

9. The method for scatter correction of turbine blade CT images using relative total variation and wavelet transform according to claim 1, characterized in that: In step S7, the projection sinusoidal graph P to be corrected CBCT_LL1 Subtract the estimated scattered component S LL1 Get the corrected approximate component P cor_LL1 : P cor_LL1 =P CBCT_LL1 -P scatter_LL1 。 10. The method for scatter correction of turbine blade CT images based on relative total variation and wavelet transform according to claim 1, characterized in that: In step S8, the two-dimensional image f(x,y) is reconstructed by inverse discrete wavelet transform: Corrected projected sinogram P cor By P cor_LL1 、P CBCT_LH1 、P CBCT_HL1 、P CBCT_HH1 The CT image is reconstructed by wavelet and then reconstructed by equidistant fan beam back projection algorithm to obtain the corrected CT image.

Images