Focusing noise optimization method and device, storage medium and product

Abstract

The invention discloses a focusing noise optimization method and device, a storage medium and a product, and relates to the technical field of interference noise optimization, and the focusing noise optimization method comprises the steps: obtaining an initial interference image containing multiple types of focusing noise, inputting the initial interference image into a preset noise error correlation model, and obtaining the noise influence coefficient of each type of focusing noise, the noise error correlation model is obtained by training the to-be-trained model based on an interference image sample, the interference image sample comprises multiple types of focusing noise and an actual focusing error, and the focusing noise of the corresponding type is removed through a corresponding denoising algorithm based on the noise influence coefficient. According to the method and the device, the to-be-trained model learns the relationship between various types of noise and the focusing error, so that the influence coefficient of various types of focusing noise on the focusing error can be determined. Based on the noise influence coefficient, the focusing noise of the corresponding type is removed in a targeted manner, the wavefront curvature extraction precision is prevented from being influenced, and the focusing precision of the wavefront curvature interference method is further improved.

Description

Technical Field

[0001] This application relates to the field of interference noise optimization technology, and in particular to focusing noise optimization methods, equipment, storage media and products. Background Technology

[0002] Among existing focusing methods, wavefront curvature interferometry directly captures wavefront information through an interferometric system and derives focus deviation based on the magnitude and change of curvature. It has the advantages of fast response and good sensitivity to distortion, and has become a key technology direction in the field of precision detection.

[0003] In practical applications, wavefront curvature interferometry is susceptible to interference from various types of noise, leading to disordered wavefront curvature calculations. Currently, noise generated during the application process is typically suppressed using a single filter. However, this single-filter approach blurs the details of the interference fringes, affecting the accuracy of wavefront curvature extraction and consequently reducing the focusing accuracy of the wavefront curvature interferometry. Summary of the Invention

[0004] The main objective of this application is to provide a method, apparatus, storage medium, and product for optimizing focusing noise, aiming to solve the technical problem of low focusing accuracy in wavefront curvature interferometry.

[0005] To achieve the above objectives, this application proposes a focusing noise optimization method, the method comprising: Acquire an initial interferometric image containing multiple types of focusing noise; The initial interferometric image is input into a preset noise error correlation model to obtain the noise influence coefficients of various types of focusing noise on the focusing error of the initial interferometric image. The noise error correlation model is obtained by training a preset model to be trained based on preset interferometric image samples. The interferometric image samples include multiple types of focusing noise and the actual focusing error during focusing. Based on the noise influence coefficient, the corresponding type of focusing noise in the initial interferometric image is removed using a corresponding denoising algorithm to obtain the target interferometric image.

[0006] In one embodiment, the focusing noise includes speckle noise, additive noise, and phase jitter noise, and the noise influence coefficient includes speckle coefficient and additive coefficient. The step of removing the corresponding type of focusing noise from the initial interferometric image based on the noise influence coefficient using a corresponding denoising algorithm to obtain the target interferometric image includes: Based on the speckle coefficient and the preset homomorphic filtering algorithm, speckle noise in the initial interference image is removed to obtain a first interference image with the speckle noise removed; Based on a preset phase compensation algorithm and a preset standard interferogram, phase jitter noise in the first interferogram is removed to obtain a second interferogram with the phase jitter noise removed. Based on the additive coefficients and the preset sparse basis tracking algorithm, additive noise in the second interference image is removed to obtain the target interference image with the additive noise removed.

[0007] In one embodiment, the step of removing speckle noise from the initial interferometric image based on the speckle coefficient and a preset homomorphic filtering algorithm to obtain a first interferometric image with the speckle noise removed includes: The pixel values ​​of the initial interference image are logarithmically divided to convert the speckle noise in the initial interference image into additive speckle noise, thus obtaining the converted interference image. Based on the speckle coefficient, determine the filter cutoff frequency when filtering the speckle noise; Based on the filter cutoff frequency, the low-frequency additive speckle noise and the high-frequency interference fringes in the converted interference image are separated by a preset Gaussian filtering algorithm to obtain the first interference image.

[0008] In one embodiment, the step of removing phase jitter noise from the first interferogram based on a preset phase compensation algorithm and a preset standard interferogram to obtain a second interferogram with the phase jitter noise removed includes: Standard fringes are extracted from the standard interferogram along the fringe direction, and the standard fringes are rotated to the horizontal direction to obtain a standard fringe template; The first interference image is rotated and translated, and the interference fringes in the first interference image are stitched together with the standard fringes in the standard fringes template to obtain the stitched interference image. Based on a preset linear interpolation algorithm, the grayscale values ​​at the junction of the interference fringes and the standard fringes in the stitched interference image are interpolated to obtain the second interference image.

[0009] In one embodiment, the step of removing additive noise from the second interferometric image based on the additive coefficients and a preset sparse basis tracking algorithm to obtain the target interferometric image with the additive noise removed includes: Based on the additive coefficients, the Lagrange coefficients of the sparse basis pursuit algorithm are determined; Based on the preset observation matrix and the Lagrange coefficients, a target optimization function is determined when removing the additive noise. The target optimization function is the sum of the spatial norm and the numerical fidelity of the denoised second interferometric image. The spatial norm is the sum of the pixel values ​​of the denoised second interferometric image, and the numerical fidelity is the pixel difference between the denoised second interferometric image and the second interferometric image. Based on the target optimization function, additive noise in the second interference image is removed to obtain the target interference image with the additive noise removed.

[0010] In one embodiment, the step prior to inputting the initial interferometric image into a preset noise error correlation model to obtain the noise influence coefficients of various types of focusing noise on the focusing error of the initial interferometric image further includes: The interference image samples are obtained, wherein the interference image samples further include quantization parameters of various types of focusing noise; Based on the quantization parameters, the actual focusing error of the interference sample image, and the model to be trained, the predicted focusing error during focusing is determined. Based on the predicted focus error and the actual focus error, the model to be trained is trained to obtain the noise error correlation model.

[0011] In one embodiment, the model to be trained is a random forest model, and the step of determining the predicted focus error during focusing based on the quantization parameters, the actual focus error of the interference sample image, and the model to be trained includes: Based on a preset multiple linear regression algorithm, the quantization parameters are used as independent variables and the actual focusing error is used as the dependent variable to determine the initial influence coefficients of various types of focusing noise on the actual focusing error. Based on the initial influence coefficient, the quantization parameter, and the actual focusing error, the predicted focusing error during focusing is determined using the random forest model.

[0012] In addition, to achieve the above objectives, this application also proposes a focus noise optimization device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the focus noise optimization method as described above.

[0013] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the focus noise optimization method described above.

[0014] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the focus noise optimization method described above.

[0015] One or more technical solutions proposed in this application have at least the following technical effects: This application obtains an initial interferometric image containing multiple types of focusing noise, inputs the initial interferometric image into a preset noise error association model, and obtains the noise influence coefficients of each type of focusing noise on the focusing error of the initial interferometric image. The noise error association model is obtained by training a preset model to be trained based on preset interferometric image samples. The interferometric image samples include multiple types of focusing noise and the actual focusing error during focusing. Based on the noise influence coefficients, the corresponding type of focusing noise in the initial interferometric image is removed through a corresponding denoising algorithm.

[0016] To address the issue that single-filtering methods can blur the details of interference fringes, thereby affecting the accuracy of wavefront curvature extraction and reducing the focusing accuracy of wavefront curvature interferometry, this application uses a noise error correlation model to determine the noise influence coefficient and performs targeted denoising based on this coefficient. Since the interference image samples used for training include multiple types of focusing noise and actual focusing errors, training the model using these samples allows it to learn the relationship between various types of noise and focusing errors, thus determining the influence coefficients of different types of focusing noise on the focusing errors. Furthermore, this application does not use a uniform single filter for denoising the interference images; instead, based on the noise influence coefficients, it employs appropriate denoising algorithms to specifically remove corresponding types of focusing noise, thereby avoiding impact on wavefront curvature extraction accuracy and improving the focusing accuracy of wavefront curvature interferometry. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating an embodiment of the focusing noise optimization method of this application. Figure 2A diagram of an autofocus device with wavefront curvature interference provided in Embodiment 1 of the focusing noise optimization method of this application; Figure 3 A physical image of the wavefront curvature interferometric autofocus device provided in Embodiment 1 of the focusing noise optimization method of this application; Figure 4 The wavefront curvature interferogram provided in Embodiment 1 of the focusing noise optimization method of this application; Figure 5 This is a technical roadmap provided for Embodiment 1 of the focusing noise optimization method of this application; Figure 6 This is a flowchart illustrating Embodiment 2 of the focusing noise optimization method of this application; Figure 7 This is a schematic diagram of the device structure of the hardware operating environment involved in the focusing noise optimization method in this application embodiment; Figure 8 This is a schematic diagram illustrating the data acquisition consent process involved in the focus noise optimization method in this application embodiment.

[0020] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0022] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0023] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device or a focus noise optimization device capable of performing the above functions. The following description uses a focus noise optimization device as an example to illustrate this embodiment and the subsequent embodiments.

[0024] Autofocus technology is a core support for yield control in semiconductor manufacturing, and it needs to meet the "nanoscale" precision imaging requirements such as 12-inch wafer lithography alignment, defect screening of advanced processes at 3nm and below, and thin film thickness detection of power semiconductors.

[0025] Traditional image-domain autofocus methods, including gray-scale variance methods, template matching methods, and Sobel operator methods, rely on indirect image features. These methods are prone to getting stuck in local extrema and experiencing response lag in low-contrast, dynamic scenes, or noisy environments, leading to a sharp decrease in focusing accuracy and failing to meet the "nanoscale" inspection requirements of semiconductor manufacturing. While existing wavefront curvature interferometry autofocus methods can directly capture wavefront information, they are affected by three types of noise in the imaging link, causing wavefront curvature calculation disorder. The inherent characteristics of the interferometric system exacerbate focusing uncertainty. Existing noise suppression algorithms mostly use single filtering, which blurs the details of interference fringes, affecting the accuracy of wavefront curvature extraction and thus reducing the focusing accuracy of wavefront curvature interferometry.

[0026] Furthermore, existing methods cannot clearly define the influence weights of the three types of noise, resulting in a lack of specificity in noise reduction schemes. Also, phase jitter noise lacks a dedicated correction module, which can easily lead to fringe shift and cause curvature calculation errors. Most studies are based on verification in conventional noise scenarios and have not designed experimental and optimization schemes for extreme scenarios that may occur in semiconductor manufacturing, thus limiting their engineering applications.

[0027] Based on this, embodiments of this application provide a method for optimizing focus noise, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the focusing noise optimization method of this application.

[0028] In this embodiment, the focus noise optimization method includes steps S10~S30: Step S10: Obtain an initial interferometric image containing multiple types of focusing noise; It should be noted that focusing noise refers to various types of noise that interfere with the quality of the interferometric image and affect the focusing accuracy during wavefront curvature interferometry autofocus. The initial interferometric image refers to the raw interferometric image obtained directly by the interferometric system without denoising processing, containing various types of focusing noise. The interferometric image is an image that reflects the wavefront information of light waves, obtained through wavefront curvature interferometry, and its fringe structure carries key focusing information such as focus deviation.

[0029] It is understood that this embodiment directly acquires the raw interferometric image from the wavefront curvature interferometry system. This image has not undergone any denoising processing and retains the various types of focusing noise introduced during the actual imaging process. This provides a real and complete input data foundation for subsequent targeted denoising based on noise type and degree of influence.

[0030] Step S20: Input the initial interferometric image into a preset noise error association model to obtain the noise influence coefficients of various types of focusing noise on the focusing error of the initial interferometric image. The noise error association model is obtained by training a preset model to be trained based on preset interferometric image samples. The interferometric image samples include multiple types of focusing noise and the actual focusing error during focusing. It should be noted that the noise error correlation model refers to a model trained using machine learning methods, used to establish a quantitative relationship between various types of focusing noise and focusing error. It can output noise influence coefficients of various types of focusing noise on focusing error based on the input initial interferometric image. The noise influence coefficient represents a quantitative parameter indicating the degree of influence of a specific type of focusing noise on focusing error, used to guide the selection and parameter configuration of subsequent targeted denoising algorithms. Interferometric image samples refer to a set of known data used to train the noise error correlation model; each sample contains information on multiple types of focusing noise and their actual focusing errors generated during the actual focusing process. The model to be trained refers to a machine learning model that does not yet possess the ability to map noise to error before training; it needs to be trained using interferometric image samples to generate the noise error correlation model. A schematic diagram of the automatic focusing device using wavefront curvature interferometry in this embodiment can be found in [reference needed]. Figure 2 The physical prototype of the wavefront curvature interferometry autofocus device can be referenced. Figure 3 The wavefront curvature interferogram can be referenced. Figure 4 .

[0031] It is understood that in this embodiment, the initial interferometric image is input into a preset noise error correlation model. This model is obtained by training a preset training model with a large number of interferometric image samples containing multiple types of focusing noise and corresponding actual focusing errors. Based on this, the model analyzes the noise components in the initial interferometric image and outputs the noise influence coefficients of various types of focusing noise on the focusing error of the current image. This achieves a quantitative assessment of the influence of different noise types, providing key parameter basis for subsequent implementation of differentiated denoising strategies according to noise type.

[0032] Since this embodiment uses a noise error correlation model trained based on real interference image samples, and the interference image samples clearly contain multiple types of focusing noise and their corresponding actual focusing errors, the model can learn and reflect the inherent correlation between various types of noise and focusing errors, thereby outputting an accurate noise influence coefficient.

[0033] Furthermore, since the noise impact coefficient is a quantitative representation of the degree of influence of various types of focusing noise, this step enables the system to distinguish the contribution of different noises to focusing accuracy, thereby avoiding the problem of blurred stripe details caused by using a uniform filtering method, and laying the foundation for achieving high-precision, targeted noise reduction.

[0034] Step S30: Based on the noise influence coefficient, the corresponding type of focusing noise in the initial interferometric image is removed using a corresponding denoising algorithm to obtain the target interferometric image.

[0035] It should be noted that denoising algorithms refer to signal processing methods used to suppress or eliminate specific types of focusing noise. Their selection and parameter configuration are adapted based on the noise impact coefficient to achieve targeted denoising.

[0036] Target interferometric image refers to an optimized interferometric image obtained after targeted removal of various types of focusing noise, which retains clear interference fringes and significantly reduces focusing error.

[0037] It is understood that, in this embodiment, based on the obtained noise influence coefficient, a corresponding denoising algorithm is matched for each type of focusing noise, and these algorithms are applied sequentially to the initial interference image to remove the corresponding type of focusing noise. After multi-stage, type-matched denoising processing, the final target interference image is generated. This effectively suppresses noise while preserving the high-frequency details and structural integrity of the interference fringes to the maximum extent. The overall process of focusing noise denoising in this embodiment can be referred to... Figure 5 .

[0038] Since this embodiment selects a denoising algorithm based on the noise influence coefficient to remove the corresponding type of focusing noise, that is, the denoising strategy corresponds one-to-one with the noise type and the degree of influence, rather than using a uniform single filtering method, it can avoid the problem of interference fringe blurring caused by excessive smoothing in traditional methods.

[0039] Furthermore, since different types of focusing noise have different statistical characteristics and physical causes, targeted algorithms can more accurately separate noise from effective signals, thereby reducing focusing errors while maintaining the fringe clarity required for wavefront curvature extraction and improving the overall focusing accuracy of wavefront curvature interferometry.

[0040] In one feasible implementation, the focusing noise includes speckle noise, additive noise, and phase jitter noise; the noise influence coefficient includes speckle coefficient and additive coefficient; and the specific implementation of removing the corresponding type of focusing noise from the initial interferometric image based on the noise influence coefficient using a corresponding denoising algorithm to obtain the target interferometric image can also be: Based on the speckle coefficients and a preset homomorphic filtering algorithm, speckle noise in the initial interferometric image is removed to obtain a first interferometric image with the speckle noise removed. Based on a preset phase compensation algorithm and a preset standard interferogram, phase jitter noise in the first interferometric image is removed to obtain a second interferometric image with the phase jitter noise removed. Based on the additive coefficients and a preset sparse basis tracking algorithm, additive noise in the second interferometric image is removed to obtain the target interferometric image with the additive noise removed.

[0041] It should be noted that speckle noise is a multiplicative noise generated by coherent light interference, manifesting as granular random intensity fluctuations in the interference image, which can mask the true fringe structure. Additive noise refers to random noise existing in the interference image in a superimposed manner; it is independent of the signal and is uniformly distributed across all pixels of the image. Phase jitter noise refers to random fluctuations in the phase of the interference fringes caused by factors such as system vibration, environmental disturbances, or instability of optical components, resulting in fringe position shifts or distortions. The speckle coefficient represents the noise influence coefficient of the degree to which speckle noise affects focusing error, and is used to guide the parameter settings of the homomorphic filtering algorithm. The additive coefficient represents the noise influence coefficient of the degree to which additive noise affects focusing error, and is used to determine optimization parameters such as the Lagrange coefficient in the sparse basis tracking algorithm.

[0042] Homomorphic filtering is a denoising method that converts multiplicative noise into an additive form by taking its logarithm, and then uses frequency domain filtering to separate low-frequency illumination and high-frequency reflection components. Phase compensation algorithms correct fringe distortion caused by phase jitter by aligning with a standard interferogram to achieve phase recovery. A standard interferogram refers to a reference interferogram with a stable and regular fringe structure acquired under ideal, undisturbed conditions, used for template matching in the phase compensation process. Sparse basis pursuit algorithms are denoising methods based on sparse signal representation theory, separating noise from effective signals by solving an optimization problem in a specific transform domain. The first interferogram refers to the intermediate result image obtained after removing speckle noise from the initial interferogram. The second interferogram refers to the intermediate result image obtained after removing phase jitter noise from the first interferogram.

[0043] Understandably, this embodiment first processes the initial interferometric image based on the speckle coefficient and a preset homomorphic filtering algorithm to effectively suppress speckle noise, obtaining a first interferometric image. Subsequently, a preset phase compensation algorithm and a standard interferogram are used to correct the phase jitter noise in the first interferometric image, and a second interferometric image is generated through fringe alignment and interpolation. Finally, additive noise is further removed from the second interferometric image based on additive coefficients and a preset sparse basis tracking algorithm, ultimately obtaining the target interferometric image. This achieves precise, phased, and type-adaptive removal of the three main types of focusing noise, significantly improving image quality while preserving interference fringe details.

[0044] By clearly distinguishing between three types of focusing noise—speckle noise, phase jitter noise, and additive noise—and configuring denoising algorithms that match the physical mechanisms of each, homomorphic filtering is used to address multiplicative speckle noise, phase compensation to address fringe distortion, and sparse basis tracking to address superimposed additive noise. Furthermore, the parameters of each algorithm are dynamically determined by the corresponding noise impact coefficient, thus ensuring that the denoising intensity is adapted to the actual degree of noise impact.

[0045] Furthermore, this embodiment adopts a sequential processing strategy, thereby avoiding mutual interference between algorithms, effectively protecting the high-frequency information and phase continuity of the interference fringes, and ultimately significantly reducing the overall focusing error and improving the wavefront curvature extraction accuracy without blurring the fringes.

[0046] In one feasible implementation, the specific implementation of removing speckle noise from the initial interferometric image based on the speckle coefficient and a preset homomorphic filtering algorithm to obtain a first interferometric image with the speckle noise removed can also be: Logarithmic operations are performed on the pixel values ​​of the initial interference image to convert the speckle noise in the initial interference image into additive speckle noise, resulting in a converted interference image. Based on the speckle coefficient, the filter cutoff frequency for filtering the speckle noise is determined. Based on the filter cutoff frequency, a preset Gaussian filtering algorithm is used to separate the low-frequency additive speckle noise and the high-frequency interference fringes in the converted interference image, resulting in the first interference image.

[0047] It should be noted that the logarithmic operation refers to performing a natural logarithmic or common logarithmic transformation on the grayscale value of each pixel in the initial interference image. This is used to convert multiplicative speckle noise into an additive form, facilitating subsequent linear filtering. Additive speckle noise, after the logarithmic operation, is transformed from multiplicative speckle noise into an additive component that can be superimposed on the signal, and its spectrum is mainly concentrated in the low-frequency region.

[0048] The filter cutoff frequency is a threshold parameter used in frequency domain filtering to distinguish between retained and suppressed frequency components. In this embodiment, it is determined by the speckle coefficient and is used to control the suppression strength of the Gaussian filter for low-frequency additive speckle noise. The Gaussian filtering algorithm is a low-pass filtering method based on the Gaussian function, used to smooth images and suppress low-frequency noise. In this embodiment, it is used to separate low-frequency additive speckle noise from high-frequency interference fringes.

[0049] It should be noted that this embodiment first performs a logarithmic operation on the pixel values ​​of the initial interference image, converting the speckle noise, which originally existed in a multiplicative form, into additive speckle noise, resulting in the converted interference image. Subsequently, a filter cutoff frequency for removing this additive speckle noise is determined based on the speckle coefficient, and a preset Gaussian filtering algorithm is configured based on this cutoff frequency. This Gaussian filter is used to perform frequency domain low-pass filtering on the converted interference image, effectively separating the low-frequency additive speckle noise components while retaining the high-frequency interference fringe information. Finally, an exponential restoration or other inverse transformation is used to obtain the first interference image with removed speckle noise. This achieves a physical mechanism-adaptive removal of speckle noise while preserving the key high-frequency fringe structure reflecting the wavefront curvature to the greatest extent possible.

[0050] This embodiment transforms multiplicative speckle noise into an additive form by taking the logarithm, making the originally nonlinearly coupled noise and signal a linearly separable superposition relationship, thus providing a basis for subsequent linear filtering methods.

[0051] Furthermore, since the filter cutoff frequency is determined based on the speckle coefficient, which reflects the actual impact of speckle noise on focusing error, this cutoff frequency can dynamically adapt to the noise intensity. When the noise impact is large, a lower cutoff frequency is used to enhance noise reduction; when the impact is small, more details are preserved to avoid over-smoothing. Thus, Gaussian filtering can accurately separate low-frequency additive speckle noise from high-frequency interference fringes in the frequency domain. This effectively suppresses speckle noise while fully preserving the fringe edges and high-frequency modulation information used for wavefront curvature calculation, significantly improving subsequent focusing accuracy.

[0052] In one feasible implementation, the specific implementation of removing phase jitter noise from the first interferogram based on a preset phase compensation algorithm and a preset standard interferogram to obtain a second interferogram with the phase jitter noise removed can also be: Standard fringes are extracted from the standard interferogram along its fringe direction and rotated to a horizontal position to obtain a standard fringe template. The first interferogram is then rotated and translated, and the interference fringes in the first interferogram are stitched together with the standard fringes in the standard fringe template to obtain a stitched interferogram. Based on a preset linear interpolation algorithm, the grayscale values ​​at the junction of the interference fringes and the standard fringes in the stitched interferogram are interpolated to obtain the second interferogram.

[0053] It should be noted that standard fringes refer to a segment of local interference fringes with a regular and stable phase distribution that is cut out from the fringe direction of a preset standard interferogram and used as a reference for phase alignment.

[0054] A standard fringe template refers to a normalized reference image formed by rotating the standard fringes to a horizontal position, facilitating geometric alignment and stitching with the image to be corrected. Rotation and translation refer to rigid body geometric transformation operations applied to the first interferogram, used to adjust the direction and position of its interference fringes to align it spatially with the standard fringe template. A stitched interferogram is a combined image formed by aligning the rotated and translated first interferogram with the standard fringe template in the fringe region; discontinuities in grayscale may exist at the stitching points. A linear interpolation algorithm refers to an algorithm that performs linear transition processing on the pixel grayscale values ​​at the stitching boundary between the interference fringes and the standard fringes in the stitched interferogram, used to eliminate stitching seams and achieve smooth connections.

[0055] Understandably, this embodiment first extracts a standard fringe from a preset standard interferogram along its fringe direction and rotates it to a horizontal position to form a standard fringe template. Then, a rotation and translation operation is performed on the first interferogram to align its internal interference fringes with the standard fringe template in both direction and position. The aligned first interferogram is then stitched together with the standard fringe template to generate a stitched interferogram. A preset linear interpolation algorithm is used to perform grayscale interpolation on the stitching boundary region between the interference fringes and the standard fringes in the stitched interferogram, smoothly transitioning grayscale differences. This results in a second interferogram with phase jitter noise removed, effectively correcting fringe phase shifts or distortions caused by system disturbances and restoring the continuity and regularity of the interference fringes.

[0056] Because this embodiment constructs the standard fringes in the standard interferogram as a horizontal standard fringe template, and rotates and translates the first interferogram to achieve fringe alignment, the fringe distortion or offset caused by phase jitter can be geometrically corrected by referring to the ideal template, thereby directly compensating for phase error in the spatial domain.

[0057] Furthermore, since a linear interpolation algorithm is used to smooth the grayscale at the stitching point after stitching, artificial boundaries or grayscale jumps introduced by hard stitching are avoided, ensuring the continuity and phase consistency of the interference fringes. Therefore, this embodiment does not rely on complex frequency domain estimation, but achieves effective suppression of phase jitter noise through geometric alignment and local interpolation. This significantly improves the phase stability of the interference image while preserving the original fringe structure, providing reliable input for subsequent high-precision wavefront curvature calculation.

[0058] In one feasible implementation, the specific implementation of removing additive noise from the second interferometric image based on the additive coefficients and a preset sparse basis tracking algorithm to obtain the target interferometric image with the additive noise removed can also be: Based on the additive coefficients, the Lagrange coefficients of the sparse basis pursuit algorithm are determined. Based on the preset observation matrix and the Lagrange coefficients, the target optimization function for removing the additive noise is determined. The target optimization function is the sum of the spatial norm and numerical fidelity of the denoised second interferometric image. The spatial norm is the sum of the pixel values ​​of the denoised second interferometric image, and the numerical fidelity is the pixel difference between the denoised second interferometric image and the second interferometric image. Based on the target optimization function, the additive noise in the second interferometric image is removed to obtain the target interferometric image with the additive noise removed.

[0059] It should be noted that the Lagrange multiplier is a regularization parameter used in the sparse basis pursuit algorithm to balance the weights between sparsity constraints and data fidelity in the objective optimization function. In this embodiment, its value is determined by additive coefficients. The observation matrix is ​​a preset linear transformation matrix used to map the interferometric image from the original pixel domain to the sparse representation domain. The objective optimization function is a mathematical expression used to guide the denoising process, defined as the sum of the spatial norm and numerical fidelity of the denoised image. The optimal denoising result is obtained by minimizing this function. The spatial norm is a regularization term used to constrain the overall strength or energy of the solution. Numerical fidelity refers to the pixel-level difference between the denoised second interferometric image and the original second interferometric image, used to ensure that the denoising result does not deviate too far from the original observation data.

[0060] It should also be noted that the calculation formula for the optimization target in this embodiment is as follows:

[0061] Where x is the denoised signal, A is the observation matrix, b is the noisy signal, and λ is the Lagrange coefficient.

[0062] Understandably, this embodiment first determines the Lagrange coefficients required for the sparse basis pursuit algorithm based on additive coefficients. Then, combining the preset observation matrix and these Lagrange coefficients, a target optimization function for removing additive noise is constructed. The target function consists of two parts: one is the spatial norm of the denoised second interferogram, used to constrain the overall characteristics of the solution; the other is the numerical fidelity, i.e., the pixel difference between the denoised image and the original second interferogram, used to maintain fidelity to the original data. Finally, by solving for the minimum value of this target optimization function, additive noise is effectively separated and suppressed from the second interferogram, outputting a target interferogram with additive noise removed. Thus, precise removal of additive noise is achieved while balancing image structure preservation and noise suppression.

[0063] Since the Lagrange coefficients are determined based on the additive coefficients, which reflect the actual impact of additive noise on focusing error, the constructed objective optimization function can dynamically adjust the denoising intensity. When the additive noise has a significant impact, higher weight is given to numerical fidelity to avoid excessive distortion, or sparsity constraints are appropriately strengthened to improve denoising capability.

[0064] Furthermore, since the target optimization function explicitly includes the joint optimization of spatial norm and numerical fidelity, it can suppress additive noise while preventing drastic shifts in the overall brightness or contrast of the image. Thus, by solving the optimization problem of the degree of influence of this adaptive noise, additive noise can be effectively removed without introducing artificial artifacts, ensuring the signal-to-noise ratio and wavefront information integrity of the target interferometric image, thereby supporting high-precision focusing.

[0065] In one embodiment, after the step of removing the corresponding type of focusing noise from the initial interferometric image based on the noise influence coefficient and using a corresponding denoising algorithm to obtain the target interferometric image, the method further includes: In this embodiment, 30 existing images each from low, medium, and high noise scenes were selected and divided into three groups (no noise reduction group, traditional single noise reduction group, and the adaptive group of this invention). The focus error rate, Strell ratio, and image contrast of the three groups were compared to verify the error reduction effect in medium and high noise scenes.

[0066] In one embodiment, after the step of removing the corresponding type of focusing noise from the initial interferometric image based on the noise influence coefficient and using a corresponding denoising algorithm to obtain the target interferometric image, the method further includes: Strong speckle patterns are simulated by rotating frosted glass, and speckle density is controlled by adjusting the rotation speed. Gray scale standard deviation is adjusted by adjusting the brightness of adjustable LED lights. Strong phase jitter is simulated by using a miniature vibration table, and stripe offset is controlled by adjusting the amplitude.

[0067] Specifically, in this embodiment, five noise-free template images are first collected for each type of extreme scenario, followed by 30 noisy images. These images are then input into an adaptive algorithm to calculate the focus error rate and verify the model's generalization ability.

[0068] In summary, this embodiment obtains an initial interferometric image containing multiple types of focusing noise, inputs the initial interferometric image into a preset noise error association model, and obtains the noise influence coefficients of each type of focusing noise on the focusing error of the initial interferometric image. The noise error association model is obtained by training a preset model based on preset interferometric image samples. The interferometric image samples include multiple types of focusing noise and the actual focusing error during focusing. Based on the noise influence coefficients, the corresponding type of focusing noise in the initial interferometric image is removed through a corresponding denoising algorithm.

[0069] To address the issue that single-filtering methods can blur the details of interference fringes, thus affecting the accuracy of wavefront curvature extraction and consequently reducing the focusing accuracy of wavefront curvature interferometry, this embodiment uses a noise error correlation model to determine the noise influence coefficient and performs targeted denoising based on this coefficient. Since the interference image samples used for training include multiple types of focusing noise and actual focusing errors, training the model with these samples allows it to learn the relationship between various types of noise and focusing errors, thereby determining the influence coefficients of different types of focusing noise on the focusing errors. Furthermore, this embodiment does not use a uniform single filter for denoising the interference images; instead, it uses a corresponding denoising algorithm based on the noise influence coefficient to specifically remove the relevant types of focusing noise, thus avoiding impacting the wavefront curvature extraction accuracy and improving the focusing accuracy of wavefront curvature interferometry.

[0070] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 6 Before step S20, the focusing noise optimization method further includes steps S01 to S03: Step S01: Obtain the interference image sample, wherein the interference image sample further includes quantization parameters of various types of focusing noise; It should be noted that quantification parameters refer to indicators used to numerically describe the intensity or characteristics of various focusing noises, including speckle contrast, additive noise standard deviation, phase jitter amplitude, etc., which serve as calculable representations of noise in interferometric image samples.

[0071] It should also be noted that this embodiment extracts at least 50 low / medium / high noise samples each from the database. Each sample contains three types of noise quantization parameters and corresponding focus error data, where low / medium / high represent the noise level, respectively. The focus error calculation formula is as follows:

[0072] in, It is the focusing error, and the root mean square error of wavefront curvature (RMS) is recorded simultaneously as an auxiliary indicator.

[0073] It is understood that this embodiment performs the operation of acquiring interference image samples. These interference image samples not only contain multiple types of focusing noise and their corresponding actual focusing errors, but also include quantified parameters of each type of focusing noise, that is, numerically representing the intensity or statistical characteristics of each type of noise. These samples will be used to subsequently train the model to be trained, establishing a mapping relationship between noise and focusing error, thereby providing a structured and quantifiable training data foundation for constructing a high-precision noise-error correlation model.

[0074] Since the interference image samples in this embodiment contain quantified parameters of various types of focusing noise, the model to be trained can directly utilize the correspondence between the numerical features of noise and the actual focusing error during the learning process. This avoids the ambiguity and uncertainty caused by relying solely on implicit learning of image pixels and improves the model's ability to identify different noise types and their degree of influence.

[0075] Furthermore, the introduction of quantization parameters makes the prediction of the noise impact coefficient interpretable and adjustable, providing a reliable basis for subsequent precise denoising based on this coefficient, and ultimately enhancing the robustness and generalization performance of the entire focusing noise optimization method.

[0076] Step S02: Based on the quantization parameters, the actual focusing error of the interference sample image, and the model to be trained, determine the predicted focusing error when focusing; It should be noted that the actual focus error is the true focus deviation value recorded by the system when acquiring interferometric image samples. The predicted focus error is an estimate of the focus error of the current sample, output by the model to be trained based on the input quantization parameters.

[0077] It is understood that this embodiment uses the quantization parameters and corresponding actual focusing errors in the acquired interference image samples as inputs and labels, and feeds them into the model to be trained. The model generates the predicted focusing error when focusing through the internal mapping mechanism, thereby providing comparable output results for subsequent model training and constituting the error feedback signal required for supervised learning.

[0078] Since this embodiment determines the predicted focus error based on the quantization parameters, the actual focus error of the interference image sample, and the model to be trained, that is, using the quantization parameters of noise as the independent variable and the actual focus error as the dependent variable, it drives the model to establish an explicit noise error mapping relationship, thereby making the model training process have clear physical meaning and supervision.

[0079] Furthermore, because the quantization parameters accurately characterize the intensity features of various types of focusing noise, the model can more accurately distinguish the independent or coupled contributions of different noise types to the focusing error, avoiding the misinterpretation of different noises. Therefore, the predicted focusing error generated in this embodiment is closer to the real physical process, laying the foundation for subsequent optimization of model parameters based on the difference between predicted and actual errors, ultimately improving the prediction accuracy and generalization ability of the noise error correlation model.

[0080] In one feasible implementation, the model to be trained is a random forest model, and the specific implementation of determining the predicted focus error during focusing based on the quantization parameters, the actual focus error of the interference sample image, and the model to be trained can also be: Based on a preset multiple linear regression algorithm, the quantization parameter is used as the independent variable and the actual focusing error is used as the dependent variable to determine the initial influence coefficient of each type of focusing noise on the actual focusing error. Based on the initial influence coefficient, the quantization parameter and the actual focusing error, the predicted focusing error is determined by the random forest model.

[0081] It should be noted that the random forest model is an ensemble learning algorithm composed of multiple decision trees, used in this embodiment to model the complex relationship between quantization parameters and focus error. The multiple linear regression algorithm is a statistical modeling method that outputs the weight coefficients of each variable by fitting a linear relationship between multiple independent variables and a dependent variable. The initial influence coefficients refer to the linear contribution weights of various types of focus noise to the actual focus error, initially estimated through multiple linear regression, and are used to assist or initialize the learning process of subsequent nonlinear models.

[0082] Understandably, this embodiment employs a pre-defined multiple linear regression algorithm, using the quantization parameters in the interferometric image samples as independent variables and the actual focusing error as the dependent variable, to fit the initial influence coefficients of various focusing noises on the actual focusing error. Subsequently, these initial influence coefficients, the original quantization parameters, and the actual focusing error are used together as input features or prior information and fed into a random forest model for training. The model then outputs the predicted focusing error during focusing, thereby retaining nonlinear modeling capabilities while introducing linear prior knowledge to improve model convergence speed and physical interpretability.

[0083] Since this implementation method obtains the initial influence coefficient through multiple linear regression, the coefficient reflects the initial linear correlation strength of various types of focusing noise with the actual focusing error, providing physically meaningful initialization guidance or feature enhancement for subsequent complex random forest models.

[0084] Furthermore, the random forest model, leveraging its nonlinear and high fault tolerance characteristics, can capture potential interaction effects between noises. The initial influence coefficients help the model focus more on noise dimensions that significantly contribute to the error, avoiding overfitting on low-correlation features. Therefore, this embodiment retains the flexibility of data-driven methods while incorporating the interpretability of statistical regression, resulting in more accurate and stable prediction focusing errors, thereby improving the overall performance and generalization ability of the noise error correlation model.

[0085] Step S03: Based on the predicted focus error and the actual focus error, train the model to be trained to obtain the noise error correlation model.

[0086] It should be noted that this embodiment uses goodness of fit R² and error prediction deviation as core indicators to ensure that the model can predict focusing error through noise parameters.

[0087] Understandably, this embodiment calculates the difference between the predicted focus error and the corresponding actual focus error in the interferometric image sample, and uses this difference as a loss signal to feed back to the model to be trained, driving iterative optimization of the model parameters. After multiple rounds of training, the model gradually learns the mapping relationship between the quantization parameters and the focus error, eventually converging into a noise error correlation model with generalization capabilities, thus completing the transformation from raw sample data to a deployable model.

[0088] Since this embodiment uses the difference between the predicted focus error and the actual focus error as a supervision signal to train the model, the model optimization objective is directly aligned with the core task of reducing focus error.

[0089] Furthermore, since the input features used for training include quantification parameters of various types of focusing noise, the model naturally learns the differentiated impact mechanism of different noise types on focusing accuracy in the process of minimizing the prediction error. The resulting noise error correlation model not only has high prediction accuracy, but can also implicitly or explicitly output the influence weights of various noises, providing a basis for subsequent targeted denoising.

[0090] In summary, this embodiment obtains interferometric image samples containing multiple types of focusing noise, their quantization parameters, and corresponding actual focusing errors. Based on these quantization parameters and actual focusing errors, it uses the model to be trained to generate predicted focusing errors, performs supervised training on the model, and finally obtains a noise error correlation model that can output the influence coefficients of various types of focusing noise.

[0091] Because this embodiment uses quantization parameters and actual focusing error as inputs, and trains the noise error correlation model by fusing linear priors and nonlinear modeling, the model can not only capture the independent contribution of various types of focusing noise to the focusing error, but also identify their potential coupling effects, thereby significantly improving the accuracy of noise influence coefficient estimation. Furthermore, since the entire training process directly aims to reduce focusing error, the obtained noise error correlation model can accurately guide the selection of subsequent denoising algorithm types and parameter configurations, avoiding the fringe blurring problem caused by traditional single filtering. This effectively suppresses multiple types of noise while preserving high-frequency details of the interference fringes, ultimately achieving a synergistic improvement in wavefront curvature extraction accuracy and autofocus performance.

[0092] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the focusing noise optimization method of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0093] This application provides a focus noise optimization device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the focus noise optimization method in the above embodiment 1.

[0094] The following is for reference. Figure 7 The diagram illustrates a structural schematic suitable for implementing the focus noise optimization device in the embodiments of this application. The focus noise optimization device in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, tablets, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 7 The focus noise optimization device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0095] like Figure 7 As shown, the focus noise optimization device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.) that can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the focus noise optimization device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. The communication device 1009 allows the focus noise optimization device to communicate wirelessly or wiredly with other devices to exchange data. Although the figure shows focus noise optimization devices with various systems, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0096] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0097] The focusing noise optimization device provided in this application, employing the focusing noise optimization method in the above embodiments, can solve the technical problem of low focusing accuracy in wavefront curvature interferometry. Compared with the prior art, the beneficial effects of the focusing noise optimization device provided in this application are the same as those of the focusing noise optimization method provided in the above embodiments, and other technical features in this focusing noise optimization device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0098] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0099] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0100] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the focus noise optimization method in the above embodiments.

[0101] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0102] The aforementioned computer-readable storage medium may be included in the focus noise optimization device; or it may exist independently and not assembled into the focus noise optimization device.

[0103] The aforementioned computer-readable storage medium carries one or more programs, which, when executed by the focusing noise optimization device, cause the focusing noise optimization device to perform the aforementioned focusing noise optimization method.

[0104] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0105] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0106] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0107] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described focusing noise optimization method, which can solve the technical problem of low focusing accuracy in wavefront curvature interferometry. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the focusing noise optimization method provided in the above embodiments, and will not be repeated here.

[0108] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the focus noise optimization method described above.

[0109] The computer program product provided in this application can solve the technical problem of low focusing accuracy in wavefront curvature interferometry. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the focusing noise optimization method provided in the above embodiments, and will not be repeated here.

[0110] All user-related data involved in this application was obtained with the user's permission or consent, as per [reference]. Figure 8 In other words, when this application is applied to a specific product or technology, user permission is required to acquire and process the relevant data, and the processing of the relevant data must comply with the relevant laws, regulations and regulatory standards of the relevant countries and regions.

[0111] The above description is only a part of the embodiments of this application and does not limit the scope of protection of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for optimizing focusing noise, characterized in that, The method includes: Acquire an initial interferometric image containing multiple types of focusing noise; The initial interferometric image is input into a preset noise error correlation model to obtain the noise influence coefficients of various types of focusing noise on the focusing error of the initial interferometric image. The noise error correlation model is obtained by training a preset model to be trained based on preset interferometric image samples. The interferometric image samples include multiple types of focusing noise and the actual focusing error during focusing. Based on the noise influence coefficient, the corresponding type of focusing noise in the initial interferometric image is removed using a corresponding denoising algorithm to obtain the target interferometric image.

2. The method as described in claim 1, characterized in that, The focusing noise includes speckle noise, additive noise, and phase jitter noise; the noise influence coefficient includes speckle coefficient and additive coefficient; the step of removing the corresponding type of focusing noise from the initial interferometric image based on the noise influence coefficient using a corresponding denoising algorithm to obtain the target interferometric image includes: Based on the speckle coefficient and the preset homomorphic filtering algorithm, speckle noise in the initial interference image is removed to obtain a first interference image with the speckle noise removed; Based on a preset phase compensation algorithm and a preset standard interferogram, phase jitter noise in the first interferogram is removed to obtain a second interferogram with the phase jitter noise removed. Based on the additive coefficients and the preset sparse basis tracking algorithm, additive noise in the second interference image is removed to obtain the target interference image with the additive noise removed.

3. The method as described in claim 2, characterized in that, The step of removing speckle noise from the initial interferometric image based on the speckle coefficient and a preset homomorphic filtering algorithm to obtain a first interferometric image with the speckle noise removed includes: The pixel values ​​of the initial interference image are logarithmically divided to convert the speckle noise in the initial interference image into additive speckle noise, thus obtaining the converted interference image. Based on the speckle coefficient, determine the filter cutoff frequency when filtering the speckle noise; Based on the filter cutoff frequency, the low-frequency additive speckle noise and the high-frequency interference fringes in the converted interference image are separated by a preset Gaussian filtering algorithm to obtain the first interference image.

4. The method as described in claim 2, characterized in that, The step of removing phase jitter noise from the first interferogram based on a preset phase compensation algorithm and a preset standard interferogram to obtain a second interferogram with the phase jitter noise removed includes: Standard fringes are extracted from the standard interferogram along the fringe direction, and the standard fringes are rotated to the horizontal direction to obtain a standard fringe template; The first interference image is rotated and translated, and the interference fringes in the first interference image are stitched together with the standard fringes in the standard fringes template to obtain the stitched interference image. Based on a preset linear interpolation algorithm, the grayscale values ​​at the junction of the interference fringes and the standard fringes in the stitched interference image are interpolated to obtain the second interference image.

5. The method as described in claim 2, characterized in that, The step of removing additive noise from the second interferometric image based on the additive coefficients and a preset sparse basis tracking algorithm to obtain the target interferometric image with the additive noise removed includes: Based on the additive coefficients, the Lagrange coefficients of the sparse basis pursuit algorithm are determined; Based on the preset observation matrix and the Lagrange coefficients, a target optimization function is determined when removing the additive noise. The target optimization function is the sum of the spatial norm and the numerical fidelity of the denoised second interferometric image. The spatial norm is the sum of the pixel values ​​of the denoised second interferometric image, and the numerical fidelity is the pixel difference between the denoised second interferometric image and the second interferometric image. Based on the target optimization function, additive noise in the second interference image is removed to obtain the target interference image with the additive noise removed.

6. The method as described in claim 1, characterized in that, The step prior to inputting the initial interferometric image into a preset noise error correlation model to obtain the noise influence coefficients of various types of focusing noise on the focusing error of the initial interferometric image, further includes: The interference image samples are obtained, wherein the interference image samples further include quantization parameters of various types of focusing noise; Based on the quantization parameters, the actual focusing error of the interference sample image, and the model to be trained, the predicted focusing error during focusing is determined. Based on the predicted focus error and the actual focus error, the model to be trained is trained to obtain the noise error correlation model.

7. The method as described in claim 6, characterized in that, The model to be trained is a random forest model. The step of determining the predicted focus error during focusing based on the quantization parameters, the actual focus error of the interference sample image, and the model to be trained includes: Based on a preset multiple linear regression algorithm, the quantization parameters are used as independent variables and the actual focusing error is used as the dependent variable to determine the initial influence coefficients of various types of focusing noise on the actual focusing error. Based on the initial influence coefficient, the quantization parameter, and the actual focusing error, the predicted focusing error during focusing is determined using the random forest model.

8. A focusing noise optimization device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the focus noise optimization method as described in any one of claims 1 to 7.

9. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the focusing noise optimization method as described in any one of claims 1 to 7.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the focus noise optimization method as described in any one of claims 1 to 7.

Images