A core image denoising method, device and equipment

Abstract

The embodiment of the present specification relates to the technical field of digital core, and especially relates to a core image denoising method, device and equipment, comprising: constructing a diffusion denoising model according to sample core image data; the diffusion denoising model comprises a converter and a denoiser; configuring the denoiser according to a plurality of sampling times to obtain a plurality of corresponding denoisers; inputting target core image data into the converter to convert the noise of the target core image data; inputting the target core image data after noise conversion into the plurality of denoisers to obtain a plurality of candidate denoising core images; fusing the plurality of candidate denoising core images according to the observability of the plurality of candidate denoising core images to obtain a denoising core image of the target core image data; the observability is used to represent the denoising deviation degree of the candidate denoising core image.

Description

Technical Field

[0001] The embodiments of this specification relate to the field of digital core technology, and more particularly to a method, device, and apparatus for denoising a core image. Background Art

[0002] Digital core technology is a key tool for analyzing reservoir rocks. High-precision scanning instruments are used to acquire core images, and specialized software is used to construct digital core models, providing a powerful tool for subsequent analysis of rock seepage parameters. However, due to external factors, the scanning process inevitably includes a variety of complex real-world noise, making it difficult to effectively restore the true pore structure of the physical core during subsequent modeling. Core image denoising algorithms aim to remove image noise while preserving as much of the core's true structural detail as possible.

[0003] Existing digital core denoising algorithms primarily include linear filtering, nonlinear filtering, wavelet transform filtering, and total variation regularization. Core images are characterized by low contrast, nonuniform illumination, and the superposition of multiple real-world noise types (such as long-tail noise, Poisson noise, and impulse noise). However, these algorithms assume that the noise is uniform or of a specific type, and may not be able to handle this complex real-world noise present in core images. Furthermore, the structural complexity of core images, such as the irregular pore distribution, makes it difficult for these algorithms to distinguish between noise and the actual core structure, resulting in loss of detail after denoising. Specifically, linear filtering algorithms such as mean filtering use local weighted averaging for denoising, but this can destroy the microscopic pore structure, blurring the texture and affecting subsequent geological analysis (such as porosity calculation). Nonlinear filtering algorithms such as median filtering and non-local mean filtering are effective against impulse noise, but are less effective against Gaussian noise and are prone to failure in areas of low core repeatability. Wavelet denoising relies on a fixed basis function, making it difficult to match the irregular texture of core images, and improper threshold selection can easily lead to artifacts or loss of detail. Algorithms such as total variation regularization rely on manually designed priors, which makes it difficult to model the complex characteristics of cores and has low robustness to unknown noise types.

[0004] Therefore, how to overcome the problems of poor ability to distinguish between noise and real core structure, poor ability to handle composite real-world noise in core images, and unsatisfactory denoising quality in existing methods, and propose a core image denoising method with strong ability to distinguish between noise and real core structure, strong ability to handle composite real-world noise in core images, and high denoising quality is a key issue that needs to be solved urgently. Summary of the Invention

[0005] The purpose of the embodiments of this specification is to provide a core image denoising method, device and equipment to overcome the problems of poor ability to distinguish between noise and real core structure, poor ability to handle composite real-world noise, and unsatisfactory denoising quality in existing core image denoising methods.

[0006] On the one hand, an embodiment of the present specification provides a core image denoising method, comprising: constructing a diffusion denoising model based on sample core image data; the diffusion denoising model includes a converter and a denoiser; configuring the denoiser according to multiple sampling times to obtain corresponding multiple denoisers; inputting target core image data into the converter to perform noise conversion on the target core image data; inputting the noise-converted target core image data into the multiple denoisers to obtain multiple candidate denoised core images; based on the observability of the multiple candidate denoised core images, fusing the multiple candidate denoised core images to obtain a denoised core image of the target core image data; the observability is used to represent the degree of denoising deviation of the candidate denoised core images.

[0007] On the other hand, an embodiment of the present specification provides a core image denoising device, comprising: a construction module for constructing a diffusion denoising model based on sample core image data; the diffusion denoising model includes a converter and a denoiser; a configuration module for configuring the denoiser according to multiple sampling times to obtain corresponding multiple denoisers; a conversion module for inputting target core image data into the converter to perform noise conversion on the target core image data; an input module for inputting the target core image data after noise conversion into the multiple denoisers to obtain multiple candidate denoised core images; a fusion module for fusing multiple candidate denoised core images based on the observability of the multiple candidate denoised core images to obtain a denoised core image of the target core image data; the observability is used to represent the degree of denoising deviation of the candidate denoised core images.

[0008] On the other hand, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the core image denoising method.

[0009] According to the technical solution provided by the core image data of the embodiment of the present specification, the diffusion denoising model can be constructed according to the sample core image data; the diffusion denoising model includes a converter and a denoiser; the denoiser is configured according to a plurality of sampling times to obtain a plurality of denoising devices; the target core image data is input into the converter to convert the target core image data; the target core image data after noise conversion is input into the plurality of denoising devices to obtain a plurality of candidate denoising core images; according to the observability of the plurality of candidate denoising core images, the plurality of candidate denoising core images are fused to obtain the denoising core image of the target core image data; the observability is used to represent the denoising deviation degree of the candidate denoising core image. Compared with the prior art, the embodiment of the present specification can accurately distinguish the noise and core structure in the core image data through the diffusion denoising model, and then convert the complex real world noise therein. In addition, the embodiment of the present specification can also configure different denoising devices according to the sampling times to obtain a plurality of candidate denoising core images, and fuse the plurality of candidate denoising core images according to the denoising deviation degree of the plurality of candidate denoising core images, thereby reducing the influence of random deviation in the plurality of candidate denoising core images on the denoising result, and greatly improving the denoising quality of the core image data. BRIEF DESCRIPTION OF DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present specification or the prior art, the drawings needed in the embodiment or prior art description will be briefly introduced as follows.

[0011] Figure 1 is a flowchart of a core image denoising method provided by the embodiment of the present specification;

[0012] Figure 2 is a denoising and noise adding schematic diagram of a core image based on a diffusion model provided by the embodiment of the present specification;

[0013] Figure 3 is a fusion method schematic diagram of a plurality of candidate denoising core images provided by the embodiment of the present specification;

[0014] Figure 4 is a method flowchart for constructing a diffusion denoising model provided by the embodiment of the present specification;

[0015] Figure 5 is a construction method schematic diagram of a noise conversion model provided by the embodiment of the present specification;

[0016] Figure 6 is a construction method schematic diagram of a tiled diffusion model provided by the embodiment of the present specification;

[0017] Figure 7FIG. 1 is a schematic diagram of a construction method of a converter in a diffusion denoising model provided by an embodiment of the present specification;

[0018] Figure 8 FIG. 1 is a schematic diagram of a diffusion denoising model constructed based on a tiling diffusion model and an encoder and a sampler of the diffusion denoising model provided by an embodiment of the present specification;

[0019] Figure 9 FIG. 1 is a flowchart of a Gaussian noise core image generation method provided by an embodiment of the present specification;

[0020] Figure 10 FIG. 1 is a schematic diagram of target core image data of core A obtained by using a visible light surface scanner provided by an embodiment of the present specification;

[0021] Figure 11 FIG. 1 is a schematic diagram of a denoised core image obtained by using a core image denoising method to denoise target core image data of core A provided by an embodiment of the present specification;

[0022] Figure 12 FIG. 1 is a schematic diagram of target core image data of core B obtained by using a CT scanner provided by an embodiment of the present specification;

[0023] Figure 13 FIG. 1 is a schematic diagram of a denoised core image obtained by using a core image denoising method to denoise target core image data of core B provided by an embodiment of the present specification;

[0024] Figure 14 FIG. 1 is a structural composition schematic diagram of a core image denoising device provided by an embodiment of the present specification;

[0025] Figure 15 FIG. 1 is a structural composition schematic diagram of a computer device provided by an embodiment of the present specification. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the present specification will be described clearly and completely below with reference to the drawings in the embodiments of the present specification. Obviously, the described embodiments are only part of the embodiments of the present specification, rather than all the embodiments. Based on the embodiments in the present specification, all other embodiments obtained by a person of ordinary skill in the art without creative labor should fall within the scope of protection of the present specification.

[0027] Digital core images can contain details such as pores, cracks, and mineral grains with irregular shapes and variable scales. Existing methods are prone to over-smoothing, resulting in the loss of key geological information. Noise in digital core images can come from scanning equipment (such as X-ray and CT imaging) or environmental interference, showing spatial heterogeneity (such as local high-noise areas). Existing methods assume that the noise is uniformly distributed (such as global threshold segmentation), making it difficult to deal with it in a targeted manner. The internal structure of the core has low contrast and blurred boundaries. Existing edge detection algorithms are prone to misjudging noise as edges or are unable to distinguish between real core structure and noise. In addition, digital core images often contain Gaussian-like noise, Poisson noise (photon counting noise), and impulse noise (sensor anomalies). Existing methods are only designed for a single noise type and cannot effectively handle mixed noise.

[0028] To address these issues, the diffusion model can gradually remove noise through multiple iterations, avoiding the crude estimates of single-step denoising. It is particularly well-suited for preserving weak edges and tiny pores (such as nanoscale pores). The diffusion model can also be trained on a large number of core images to implicitly learn geological characteristics such as pore distribution and fracture orientation, resulting in an extremely accurate distinction between noise and true core structure. Furthermore, based on a generative framework, the diffusion model can synthesize plausible microstructures (for example, repairing pore connections obscured by noise).

[0029] Figure 1 This is a flowchart of a core image denoising method provided in an embodiment of this specification. The specific implementation includes the following steps:

[0030] S101: Constructing a diffusion denoising model based on sample core image data; the diffusion denoising model includes a converter and a denoiser.

[0031] In some embodiments, the sample core image data includes a sample core image to be denoised and a corresponding sample denoised core image; a noise conversion model can be constructed based on the sample core image to be denoised; the noise conversion model includes an encoder and a decoder, the encoder is used to perform noise conversion on the sample core image to be denoised, and the decoder is used to recover the sample core image to be denoised from the noise conversion result; a tiled diffusion model can be constructed based on the sample denoised core image; the tiled diffusion model includes a diffuser and a sampler; the diffuser is used to add noise to the sample denoised core image, and the sampler is used to denoise the added noise result; the converter can be constructed based on the encoder of the noise conversion model; the denoiser can be constructed based on the sampler of the tiled diffusion model.

[0032] The diffusion denoising model achieves precise noise removal through a phased approach. Similar to the hierarchical processing mechanism of the human visual system, the diffusion denoising model extracts features from noisy core images, transforms their representation, and then performs progressive cleanup in the feature space. This two-stage design not only conforms to the theoretical framework of signal processing but also effectively addresses the unsatisfactory performance of existing end-to-end denoising models in core image denoising scenarios.

[0033] Sample core image data can include a sample undenoised core image and a corresponding sample denoised core image. The sample undenoised core image can be a denoised core image obtained by scanning the sample core, while the sample denoised core image can be a clean core image obtained by denoising the sample undenoised core image. Specifically, the sample undenoised core image can be obtained from the following channels: 1) Raw output from core scanning equipment: This includes unprocessed data directly generated by equipment such as X-rays, CT scanners (typically with a resolution of 0.5-5μm / voxel), and visible light surface scanners. These core images generally contain inherent device noise, such as X-ray quantum noise (Poisson distribution) and electronic circuit readout noise (long-tail distribution). 2) Field-collected images: Images of the core surface taken with a digital camera or mobile phone. Noise primarily comes from uneven ambient lighting, lens contamination, and JPEG compression artifacts. This type of data is characterized by uneven spatial distribution of noise, often accompanied by localized occlusions. 3) Degraded archived sample images: Low-quality core images resulting from multiple copies or compressed transmissions may contain mixed noise (impulse noise + Gaussian noise) and blurring. Denoised core images can be obtained through the following methods: 1) Professional equipment re-inspection: Re-imaging the same core sample using higher-precision scanning equipment; 2) Expert manual restoration: Geologists use tools such as Photoshop to perform pixel-by-pixel corrections; 3) Rescanning after physical cleaning: Re-acquiring a core image after surface cleaning of the core sample.

[0034] A noise conversion model based on an autoencoder structure can be trained using sample denoised core images. The model consists of two symmetrical parts: an encoder and a decoder. During the training phase, the encoder can perform noise conversion on the input sample denoised core images, converting the complex real-world noise in the sample denoised core images into Gaussian noise. Correspondingly, the decoder can perform inverse noise conversion on the noise conversion result output by the encoder, that is, recovering the sample denoised core images from the noise conversion result through hierarchical sampling. The noise conversion model can master the ability to convert real-world noise to theoretical Gaussian noise through self-supervised learning. The noise conversion model can specifically adopt network architectures such as ResNet, Transformer, and GNN, in which the encoder part can include multiple (for example, 5-7) downsampling layers, and the decoder part can adopt a mirror-symmetrical upsampling structure, gradually performing inverse noise conversion through transposed convolution or interpolation operations. This will not be described in detail here.

[0035] The tiled diffusion model can be a generative model built based on sample denoised core images. It can learn the essential distribution of core image data by defining the forward diffusion process and the reverse denoising process. The tiled diffusion model can implicitly learn geological characteristics such as pore distribution and fracture orientation through a large number of sample denoised core images, thereby obtaining the ability to accurately distinguish between noise and real core structure. The model can include two parts: diffuser and sampler, which correspond to the two opposite processes of noise addition and removal respectively. Figure 2 The figure shows a schematic diagram of core image denoising and noise addition based on a diffusion model. The diffuser can be considered a Markov chain process, which gradually transforms a clear image into a Gaussian noise image over T time steps. At each time step t, the system adds noise of a specific intensity according to a predefined noise schedule (such as a cosine schedule). Through this progressive noise addition strategy, the diffuser can establish a complete transition path from a clear image to a pure noise image. The sampler implements the inverse denoising process, essentially learning a conditional probability distribution. Similarly, through a progressive denoising strategy, the denoiser can establish a complete transition path from a pure noise image to a clear image. This involves gradually removing noise through multiple iterations, avoiding the crude estimation of single-step denoising and helping to preserve weak edges and tiny pores (such as nanoscale pores). The tiled diffusion model can be specifically a diffusion network based on stable diffusion. By minimizing the difference between predicted and actual noise, the network gradually masters the dynamic denoising process. The tiled diffusion model can gradually remove noise through multiple iterations, avoiding the crude estimation of single-step denoising and is particularly suitable for preserving weak edges and tiny pores (such as nanoscale pores).

[0036] The diffusion denoising model consists of two modules: a converter and a denoiser. Specifically, the encoder in the noise conversion model, constructed using sample denoised core images, can be used as the converter. The sampler in the tiled diffusion model, constructed using sample denoised core images, can be used as the denoiser. The converter maps the noisy image in the original pixel space to a Gaussian noise feature space more suitable for denoising. The denoiser, based on the principles of the diffusion model, performs controllable noise removal in the feature space. Its advantage lies in its ability to fine-tune the denoising intensity through multiple iterations, avoiding the loss of detail caused by a single denoising step. The two modules work together to form a complete "encode-clean-reconstruct" processing chain, efficiently removing complex real-world noise from the denoised core images while maintaining image structural integrity.

[0037] S102: configuring the denoiser according to various sampling times to obtain corresponding various denoisers.

[0038] In some embodiments, a sampling sequence of a preset length can be generated based on a Monte Carlo algorithm; the sampling sequence includes multiple sampling times; and the denoiser is configured according to the multiple sampling times in the sampling sequence to obtain multiple denoisers with different numbers of sampling blocks.

[0039] The primary factor affecting core image distortion and perception is the randomness introduced by the Gaussian noise in the additional noise term. Excessive randomness can lead to significant distortion in the final denoised core image. In theory, moderate randomness can promote convergence to better results. Therefore, based on the Monte Carlo algorithm, the number of samplings during the denoising process can be adjusted to control randomness. The candidate denoised core images obtained through multiple inferences are averaged to converge to higher-quality results. A higher sampling number allows for more refinement and introduces more randomness, resulting in more detailed and perceptually better-quality core images.

[0040] The Monte Carlo algorithm can generate a series of numerical sequences that conform to preset rules by simulating a large number of random events and randomly sampling within a given parameter range based on the principle of probability statistics. The total number of sampling times N of the denoiser of the diffusion denoising model (that is, the total number of sampling blocks contained therein, and each sampling block performs one sampling) can be obtained. For a given confidence level a (0.7 <a<1),可以认为[aN,N]区间中每个为整数的采样次数对应的岩心图像都可以作为一个目标岩心图像数据的一个候选的去噪岩心图像,因为经过至少A(A可以表示aN向下取整的结果)次采样后,可以认为目标岩心图像数据中的噪声水平已经收敛。

[0041] Specifically, a Monte Carlo algorithm can be used to generate a sampling sequence of a preset length (for example, 10,000). The generated sampling sequence includes multiple sampling times, and the denoiser is configured for each sampling time. For example, the denoiser contains 1,000 sampling blocks, and the given confidence level is set to 0.8. After at least 1,000*0.8 sampling times, it can be considered that the noise level in the target core image data has converged. A Monte Carlo algorithm can be used to generate a sampling sequence of length 2,000. For a sampling time with a value of 810 in the sampling sequence, the first 810 sampling blocks in the denoiser can be retained. Based on the first 810 sampling blocks in the denoiser, a sampler with a sampling time of 810 is obtained. By configuring a plurality of different sampling times in the sampling sequence according to this process, a plurality of denoisers with different numbers of sampling blocks can be obtained.

[0042] S103: Input the target core image data into the converter to perform noise conversion on the target core image data.

[0043] In some embodiments, target core image data may be input into the converter to perform noise conversion on the target core image data; the target core image data may include a target core image to be denoised.

[0044] The target core image data can include the target core image to be denoised. This target core image can be input into a converter, which converts the various complex real-world noises present in the target core image to theoretical Gaussian noise. This conversion process not only preserves key information such as fracture texture and pore structure in the target core image, but also introduces a Gaussian noise pattern that matches the desired pattern, providing a sound data foundation for subsequent denoising.

[0045] S104: Inputting the noise-converted target core image data into the multiple denoisers to obtain multiple candidate denoised core images.

[0046] In some embodiments, the target core image data after noise conversion can be input into the multiple denoisers with different numbers of sampling blocks to obtain multiple candidate denoised core images.

[0047] The target core image data after noise conversion can be input into a variety of denoisers with different numbers of sampling blocks. Each denoiser samples the target core image data after noise conversion according to the number of sampling blocks it contains, generating multiple candidate denoised core images. Specifically, a denoiser containing 810 sampling blocks can sample the target core image data after noise conversion 810 times and obtain candidate denoised core images after 810 samples. Each candidate denoised core image strictly corresponds to a specific sampling number in the sampling sequence, meaning that it is the image result obtained after being processed by the denoiser at that sampling number. This correspondence ensures the traceability and interpretability of the denoising process, providing rich data dimensions for subsequent analysis and evaluation. In addition, by adjusting the sampling number in the denoising process to control randomness and obtaining multiple candidate denoised core images through inference of multiple different denoisers, a data foundation is laid for fusing multiple candidate denoised core images to obtain higher quality results.

[0048] S105: According to the observability of the plurality of candidate denoised core images, the plurality of candidate denoised core images are fused to obtain a denoised core image of the target core image data; the observability is used to indicate the denoising deviation degree of the candidate denoised core images.

[0049] In some embodiments, based on the observability of each candidate denoised core image, the following formula may be used to fuse multiple candidate denoised core images to obtain a denoised core image of the target core image data:

[0050]

[0051] Where x represents the target core image data, y fin represents the denoised core image of the target core image data, represents the i-th candidate denoised core image, represents the observability of the i-th candidate denoised core image, and M represents the total number of candidate denoised core images.

[0052] Each sampling process of the denoiser is a random process, and therefore each candidate denoised core image inevitably carries a certain amount of random bias. According to the law of large numbers, the impact of this random bias on the quality of the denoised core image can be completely eliminated if the number of inferences is sufficiently large. Therefore, by weighting multiple candidate denoised core images using corresponding observable properties, the random bias in these candidate denoised core images can be minimized, thereby improving the quality of the denoised core image.

[0053] The observability of each candidate denoised core image represents the degree of denoising bias of the candidate denoised core image, that is, the degree of bias of the residual small noise in the target core image data after denoising. Observability can be specifically expressed as the probability of obtaining the candidate denoised core image from the target core image to be denoised. Based on the known standard Gaussian noise model N(0,I) and the given target core image data x, the corresponding noise level σ can be determined and a modified Gaussian noise model can be constructed. Based on the modified Gaussian noise model The observability of each candidate denoised core image can be directly calculated

[0054] Please refer to Figure 3 , the observability of each candidate denoised core image can be As each candidate denoised core image The weighting coefficient is . According to the law of large numbers, when the total number of candidate denoised core images, M, approaches infinity, the weighted average of all candidate denoised core images and their observables can completely eliminate random deviations in the candidate denoised core images. M can be set to a sufficiently large integer (e.g., 10,000). In this case, the weighted average of all candidate denoised core images and their observables can essentially eliminate random deviations in the candidate denoised core images, thereby significantly improving the quality of the final denoised core image.

[0055] Figure 4 This is a flow chart of a method for constructing a diffusion denoising model provided in an embodiment of this specification. The specific implementation includes the following steps:

[0056] S401: Constructing a noise conversion model based on a sample core image to be denoised; the noise conversion model includes an encoder and a decoder, the encoder is used to perform noise conversion on the sample core image to be denoised, and the decoder is used to recover the sample core image to be denoised from the noise conversion result.

[0057] In some embodiments, the sample denoised core image can be input into a preset Gaussian noise adder to obtain a Gaussian noise core image; based on the sample core image to be denoised and the sample denoised core image, the noise conversion model can be constructed using the following formula:

[0058]

[0059] Where x represents the sample core image to be denoised, y represents the Gaussian noise core image, ε~Ν(0,I), represents the encoder, θ1 represents the parameters of the encoder, represents the decoder, θ2 represents the parameters of the decoder, σ represents the noise level of the sample core image to be denoised, Ν(0,I) represents the standard Gaussian distribution, λ represents the hyperparameter, and R(y) represents the regularization function.

[0060] During the scanning process of core images, real-world noise will be introduced into the core images. This real-world noise does not follow a Gaussian distribution. Because real-world noise is generated by a mixture of various random sources and various non-random sources in the real world, its distribution mechanism is difficult to learn. Although the distribution mechanism of real-world noise is difficult to learn, noise conversion can be used to convert various noise types into Gaussian noise. By setting constraints in the noise conversion model Constraining the encoder ensures that it learns the transformation relationship between complex real-world noise in the sample core image to be denoised and theoretical Gaussian noise. Based on this transformation relationship, complex real-world noise in the core image to be denoised that does not conform to the Gaussian distribution can be converted into Gaussian noise, laying a solid data foundation for subsequent Gaussian noise denoising.

[0061] Please refer to Figure 5 , the sample denoising model can be input into the preset Gaussian denoiser to obtain the Gaussian noise core image. Based on the sample core image to be denoised and the Gaussian noise core image, a noise conversion model can be constructed. Specifically, the noise conversion model can include two symmetrical parts: an encoder and a decoder. The sample core image to be denoised can be used as the input of the encoder, and the decoder takes the output of the encoder as input and outputs the prediction result of the sample core image to be denoised, which can be obtained through the loss function Constraining the encoder and decoder ensures that the noise conversion model can learn the potential feature distribution of the sample core image to be denoised. Furthermore, the output of the encoder can be further analyzed through the loss function Constrain the encoder to ensure that the encoder can learn the conversion relationship between complex real-world noise and theoretical Gaussian noise in the sample core image to be denoised. In addition, the regularization constraint can be used to balance and loss function Ensure the overall generalization and robustness of the noise conversion model.

[0062] S402: constructing a tiled diffusion model based on the sample denoised core image; the tiled diffusion model includes a diffuser and a sampler; the diffuser is used to add noise to the sample denoised core image, and the sampler is used to denoise the denoised result.

[0063] In some embodiments, based on the sample denoised core image, the following formula can be used to construct a tiled diffusion model:

[0064]

[0065] Where y0 represents the initial image block of the sample denoised core image, y t represents the image block of the sample denoised core image after the t-th time step of the diffuser denoising / the t-th time step of the sampler, ε~Ν(0,I), Ν(0,I) represents the standard Gaussian distribution, ε θ represents the tile diffusion model, θ represents the parameters of the tile diffusion model, and [O] represents adjacent image blocks and The overlapping area pixels, represents the edge area pixels of the image block, and λ1 and λ2 represent hyper parameters.

[0066] The diffusion model divides each image into multiple blocks and processes them separately, and then integrates them through a coordination mechanism. This will inevitably cause fragmentation when the blocks are spliced ​​together and sudden changes in the block edges. It can force the prediction values ​​of adjacent blocks to be similar in the overlapping parts, thereby ensuring that the overlapping areas of adjacent blocks are consistent and improving the quality of core image generation. The sudden changes at the edge of the blocks can be penalized to make the transition between blocks more natural and improve the quality of core image generation.

[0067] Please refer to Figure 6 The tiled diffusion model can be a generative model built based on the sample denoised core image, which can include two modules: diffuser and sampler. The diffuser can be considered as a Markov chain process, which gradually converts the clear image into a Gaussian noise image through T time steps. At each time step t, noise of a specific intensity can be added according to a predefined noise schedule. Through this progressive noise addition strategy, the diffuser can establish a complete transition path from a clear image to a pure noise image. The sampler can implement the reverse denoising process, the essence of which is to learn a conditional probability distribution. Similarly, through the progressive denoising strategy, the denoiser can establish a complete transition path from a pure noise image to a clear image. The tiled diffusion model can be based on diffusion networks such as DDPM and Stable Diffusion, and the loss function Minimize the difference between the predicted noise and the actual noise, so that the network can gradually master the dynamic process of denoising. In addition, the tiled diffusion model essentially divides the large image into multiple small areas (tiles) for separate processing, and then integrates them through a coordination mechanism. Therefore, the loss function can be used Ensure local consistency between different blocks, ensure the generation quality within each block, and use loss functions Ensure the global coordination of the entire image and ensure seamless connection between blocks.

[0068] S403: Construct the converter based on the encoder of the noise conversion model.

[0069] In some embodiments, the converter can be constructed based on the encoder of the noise conversion model. Considering the loss function used in the noise conversion model Constrained encoder. Therefore, it can be assumed that the encoder can learn the conversion relationship between complex real-world noise in the sample denoised core image and theoretical Gaussian noise. Therefore, the encoder in the noise conversion model can be directly used as the converter in the diffusion denoising model. The converter can be used to perform noise conversion on the input denoised core image.

[0070] S404: Construct the denoiser based on the sampler of the tiled diffusion model.

[0071] In some embodiments, the noise level corresponding to each sampling block in the sampler may be calculated using the following formula:

[0072]

[0073] Where t∈[0,T], T+1 represents the total number of sampling blocks in the sampler, and d t represents the noise level corresponding to the t-th sampling block, represents the noise attenuation coefficient corresponding to the Tt-th time step to which the t-th sampling block belongs; the degree of matching between the noise level corresponding to each sampling block and the noise level of the target core image data can be calculated; the time step of the sampling block corresponding to the maximum degree of matching can be used as a time step threshold; in the sampler, a plurality of sampling blocks whose time steps are less than or equal to the time step threshold can be determined; and the denoiser can be constructed based on the plurality of sampling blocks whose time steps are less than or equal to the time step threshold.

[0074] Different sampling blocks in the sampler correspond to different noise levels. By calculating the matching degree between the noise level corresponding to each sampling block and the noise level of the target core image data, and taking the time step of the sampling block corresponding to the maximum matching degree as the time step threshold, a denoiser can be constructed based on multiple sampling blocks whose time steps are less than or equal to the time step threshold. This achieves the alignment of the denoising capability of the denoiser with the noise level of the target core image data, which helps to improve the quality of the denoised core image generated by the denoiser.

[0075] The sampler of the tiled diffusion model can include multiple reverse sampling blocks. Multiple reverse sampling blocks correspond to multiple time steps in a Markov chain process. Assuming there are T+1 reverse sampling blocks, the t-th reverse sampling block uniquely corresponds to the Tt-th time step in the Markov chain process. The noise attenuation coefficient corresponding to the Tt-th time step can be obtained The noise level of the core image after sampling at the Ttth time step can be expressed as

[0076] The denoising capability of the sampler at time step Tt can be aligned with the noise level of the target core image data to obtain the best denoised core image. Figure 7 , the noise level of the target core image data can be obtained, and the formula The noise level corresponding to each sampling block is obtained, and the absolute deviation between the noise level of the target core image data and the noise level corresponding to each reverse sampling block is calculated. The reverse sampling block and the corresponding time step corresponding to the minimum absolute deviation can be determined, and this time step can be used as a time step threshold. In the sampler, multiple sampling blocks with time steps less than or equal to the time step threshold can be determined. These sampling blocks with time steps less than or equal to the time step threshold can be directly used as denoisers.

[0077] In some embodiments, a tiled diffusion model may be constructed based on the sample denoised core image. The tiled diffusion model includes a diffuser and a sampler. The diffuser is used to add noise to the sample denoised core image, and the sampler is used to denoise the added noise. Based on the sampler, a Gaussian noise model is constructed using the following formula:

[0078] p(x|y)~Ν(y,σ 2 I);

[0079] Wherein, x represents the target core image data, y represents the denoised core image, σ represents the noise level of the target core image data, and N(0, I) represents the standard Gaussian distribution. Each candidate denoised core image can be input into the Gaussian noise model to obtain the observability of the candidate denoised core image.

[0080] By constructing a Gaussian noise model based on a pre-trained sampler, the accurate quantification of the difference between the target core image data and each candidate denoised core image is achieved, laying the foundation for the subsequent weighted average fusion of multiple candidate denoised core images.

[0081] The tiled diffusion model can include two parts: diffuser and sampler, which correspond to the two opposite processes of noise addition and removal. Figure 2The figure shows a schematic diagram of denoising and denoising of core images based on the diffusion model. The diffuser can be considered as a Markov chain process, which gradually converts the clear image into a Gaussian noise image through T time steps. At each time step t, the system adds noise of a specific intensity according to a predefined noise schedule (such as cosine scheduling). Through this progressive denoising strategy, the diffuser can establish a complete transition path from a clear image to a pure noise image. The sampler can implement the inverse denoising process, which is essentially to learn a conditional probability distribution. Similarly, through the progressive denoising strategy, the denoiser can establish a complete transition path from a pure noise image to a clear image. The sampler is essentially equivalent to a known standard Gaussian noise model N(0,I). Therefore, based on the given target core image data x, the corresponding noise level σ can be determined, and a modified Gaussian noise model p(x|y)~N(y,σ 2 I), wherein Each candidate denoised core image can be input into the modified Gaussian noise model The observability of the candidate denoised core image is obtained

[0082] The encoder in the noise conversion model can be directly used as the converter of the diffusion denoising model, and the denoiser of the diffusion denoising model can be constructed using the above process based on the sampler in the tiled diffusion model. Please refer to Figure 8 Schematic diagram of the diffusion denoising model constructed in .

[0083] In some embodiments, a sample denoised core image included in the sample core image data can be input into a preset Gaussian noise adder to generate a Gaussian noise core image. The quality of the generated Gaussian noise core image determines whether the converter in the subsequent noise conversion model can learn the accurate conversion method from real-world noise to theoretical Gaussian noise. Figure 9 This is a flowchart of a method for generating a Gaussian noise core image provided in an embodiment of this specification. The specific implementation includes the following steps:

[0084] S901: Construct a Gaussian noise filter.

[0085] In some embodiments, a Gaussian noise adder may be constructed using the following formula:

[0086]

[0087] In the formula, σ(x,y) represents the preset basic noise intensity, α represents the adaptive adjustment coefficient, Indicates the gradient magnitude of the pixel (x, y).

[0088] For high-texture areas in core images, such as pores, cracks and other complex structures, these areas often contain rich geological information, but are also more susceptible to noise interference, causing effective signals to be obscured by the complex background. To this end, these high-texture areas are identified by calculating the gradient amplitude of the pixel points. The gradient amplitude reflects the severity of the grayscale change in the image. High gradient amplitude areas usually correspond to details such as edges and textures in the image. Subsequently, an adaptive adjustment coefficient is introduced. This coefficient is dynamically adjusted according to the size of the gradient amplitude, so that the noise variance in high-texture areas can be moderately increased. This strategy is not a simple addition of noise, but an intelligent regulation based on the local characteristics of the image. It aims to highlight the characteristics of high-texture areas and prevent the noise from being submerged by the background, thereby enhancing the effect of subsequent data enhancement operations and making geological features such as pores and cracks more clearly discernible in the image.

[0089] On the other hand, for low-texture areas in core images, such as smooth areas with homogeneous minerals, excessive noise interference not only fails to aid feature extraction but can also introduce artifacts, affecting the accuracy of image analysis. Therefore, during noise control, gradient amplitude information can be used to identify these low-texture areas and adaptively adjust the coefficient to reduce the noise intensity in these areas. By reducing the noise variance, artifacts caused by excessive noise interference in smooth areas can be reduced while preserving the overall image structure, ensuring image authenticity and accuracy.

[0090] S902: Input the sample denoised core image into a Gaussian noise adder to obtain an initial noisy core image.

[0091] In some embodiments, a sample denoised core image can be input into a Gaussian noise maker to generate an initial noisy core image. By coupling the pixel gradient amplitude with an adaptive adjustment coefficient, the Gaussian noise maker can achieve refined management of the core image noise distribution. This not only highlights the geological features of high-texture areas but also preserves the image quality of low-texture areas, providing a solid and accurate foundation for subsequent Gaussian noise core image generation.

[0092] S903: Smoothing the initial noisy core image based on the Markov random field.

[0093] In some embodiments, based on the Markov random field, the following formula can be used to smooth the initial noisy core image:

[0094]

[0095] Where, represents a single-point Gaussian constraint on the initial noisy core image based on a Markov random field, represents the neighborhood consistency constraint of the initial noisy core image based on the Markov random field, and β represents the smoothness constraint factor.

[0096] By combining the Markov random field theory and imposing single-point Gaussian constraints and neighborhood consistency constraints, the noise variance of the pore structure area in the initial noisy core image can be effectively reduced, and a higher quality Gaussian noise core image can be obtained.

[0097] The core image can be modeled as a two-dimensional Markov random field, where the grayscale value of each pixel is regarded as a random variable. These random variables are interconnected through a specific potential energy function to reflect the spatial smoothness and structural consistency of the image. A single-point Gaussian constraint can be imposed on the initial noisy core image. The core idea of ​​the single-point Gaussian constraint is that the grayscale value of the noise at each pixel in the image should follow a known statistical distribution (such as a Gaussian distribution). Therefore, based on this prior knowledge, a prior probability based on the Gaussian distribution is set for each pixel, which serves as a soft constraint in the process of adjusting the noise distribution. Through the single-point Gaussian constraint, the grayscale value of the pixel can be gradually adjusted so that it is as close as possible to the ideal state described by the Gaussian noise distribution while satisfying the observed data, thereby initially suppressing the noise.

[0098] Complex regions, such as pore structures, exhibit high noise variance and large variations in grayscale values ​​between pixels. A neighborhood consistency constraint can be further introduced. This constraint assumes strong spatial correlation between adjacent pixels in an image, meaning that the grayscale values ​​of adjacent pixels should be somewhat similar. By constructing a potential energy function that incorporates neighborhood information, the denoised image is encouraged to maintain the rationality of individual pixel grayscale values ​​while also maintaining the consistency of grayscale values ​​within the neighborhood as much as possible. This neighborhood consistency constraint effectively utilizes the spatial structural information of the image, promotes smoothing and restoration of complex regions such as pore structures, and further reduces the noise variance in these regions.

[0099] S904: Based on DCT transformation, frequency suppression is performed on the initial noisy core image using a preset frequency threshold.

[0100] In some embodiments, the initial noisy core image may be subjected to a DCT transform to obtain spectrum data corresponding to the initial noisy core image. Based on a preset frequency threshold, the following formula may be used to perform frequency suppression on the spectrum data corresponding to the initial noisy core image:

[0101]

[0102] Where N(u,v) represents the transformation coefficient at the horizontal frequency u and the vertical frequency v in the spectrum data corresponding to the initial noise core image, γ represents the preset attenuation factor, and T represents the preset frequency threshold.

[0103] The initial noisy core image can be viewed as a two-dimensional digital signal matrix, where each element represents the grayscale value of the pixel at the corresponding location in the image. Subsequently, by applying the DCT algorithm, this spatial domain image data can be converted to the frequency domain, yielding the spectral data corresponding to the initial noisy core image. The DCT transform concentrates most of the image's energy in a few low-frequency components, while concentrating image detail and noise in the majority of high-frequency components.

[0104] After obtaining the spectral data, it can be finely processed based on a preset frequency threshold. Specifically, based on the horizontal and vertical frequency amplitudes of each frequency component in the spectral data, a comparison with the preset frequency threshold can be used to determine whether to suppress that frequency component. The preset frequency threshold can be the maximum frequency component of theoretical Gaussian distributed noise obtained based on prior knowledge of Gaussian noise. If the sum of the horizontal and vertical frequency amplitudes of a frequency component in the spectral data is greater than the preset frequency threshold, the corresponding transform coefficient can be reduced to attenuate that frequency component, ensuring that the noise in the core image conforms more closely to a Gaussian distribution.

[0105] S905: Fusing the smoothing result of the initial noise core image and the frequency suppression result of the initial noise core image to obtain a Gaussian noise core image.

[0106] In some embodiments, the following formula may be used to fuse the smoothing result of the initial noisy core image and the frequency suppression result of the initial noisy core image to obtain a Gaussian noise core image:

[0107] I fin =(1-τ edge )·I MRF +τ edge I DCT ;

[0108] Where, represents the binary edge mask result of the initial noisy core image calculated using the Canny edge detection operator, represents the smoothing result of the initial noisy core image, and represents the frequency suppression result of the initial noisy core image.

[0109] For the initial noisy core image, the Canny edge detection operator can be used to extract edge features. The Canny edge detection operator systematically analyzes the image through multi-stage filtering (including Gaussian smoothing noise reduction, gradient magnitude calculation, non-maximum suppression, and dual-threshold edge connection), ultimately generating a binary edge mask image. This mask image accurately demarcates significant edge regions in the original image using a 0-1 pixel value, where pixels with a value of 1 correspond to edge locations and pixels with a value of 0 represent non-edge regions.

[0110] The generated binary edge mask can be used as a dynamic weighting coefficient to construct a composite image fusion model. Smoothing effectively eliminates high-frequency noise but may result in blurred edges, while frequency suppression preserves edge detail but may leave residual low-frequency noise. Based on this, a pixel-by-pixel weighted fusion strategy is implemented, using the binary edge mask as a weight control parameter. In edge regions (mask value 1), the edge details of the frequency suppression result are retained, while in non-edge regions (mask value 0), the overall structure of the smoothing result is prioritized. This fusion mechanism achieves the dual effects of edge preservation and Gaussian noise addition.

[0111] In some embodiments, the denoised core image of the target core image data may be input into a preset core image super-resolution model to obtain an enhanced denoised core image of the target core image data.

[0112] Based on the preset core image super-resolution model, the local core texture and global pore structure in the denoised core image can be further optimized, making the core edge and texture clearer and more natural, and achieving the generation of high-quality, detail-rich and naturally transitioned ultra-high-resolution enhanced denoised core images.

[0113] The preset core image super-resolution model can be based on the Swin-Conv-Unet (SCUNet) network architecture. This architecture cleverly combines the advantages of the Swin Transformer and convolutional neural networks (CNNs). Through the structural design of the codec, it achieves layer-by-layer extraction and reconstruction of image features. In particular, each sub-block of the codec utilizes a Swin-Conv module, which not only inherits the powerful local feature modeling capabilities of the residual convolution layer but also incorporates the non-local feature capture mechanism of the Swin Transformer module. This enables the model to capture the complex structure and texture variations of the core image globally, thereby achieving more accurate image reconstruction. The denoised core image of the target core image data can be input into the preset core image super-resolution model to obtain an enhanced denoised core image of the target core image data. In addition, the preset core image super-resolution model can perform multi-scale feature fusion based on skip connections, further improving the detail and sharpness of the denoised core image. Specifically, by transferring and fusing feature information between different scales, the model can better capture subtle structural and texture changes in core images, so that the output enhanced denoised core images at ultra-high resolution not only have fine texture features but also maintain the consistency and coherence of the overall structure.

[0114] The preset core image super-resolution model can accurately identify and enhance local core texture features in core images. These textures are important indicators of the core's internal structure and are of great significance for geological analysis and oil and gas exploration. At the same time, the preset core image super-resolution model can also effectively restore the global pore structure of the core image, making the core's edge contours clearer and more natural, avoiding the edge blurring or detail loss that can occur with traditional denoising methods. Through this series of optimizations, the resulting high-quality, detailed, and naturally transitioned ultra-high-resolution enhanced denoised core images not only enhance the image's visual quality but also provide a more accurate and reliable data foundation for subsequent geological analysis.

[0115] The following provides two specific examples of using the core image denoising method provided in this specification to denoise a core image:

[0116] Example 1: Use visible light surface scanner to obtain target core image data of core A, please refer to Figure 10 Based on the core image denoising method provided in this manual, the target core image data of core A is denoised to obtain a denoised core image. Please refer to Figure 11 .

[0117] Example 2: Use CT scanner to obtain target core image data of core B, please refer to Figure 12Based on the core image denoising method provided in this manual, the target core image data of core B is denoised to obtain a denoised core image. Please refer to Figure 13 .

[0118] The core image denoising method provided in the embodiments of this specification can construct a diffusion denoising model based on sample core image data; the diffusion denoising model includes a converter and a denoiser; the denoiser is configured according to multiple sampling times to obtain corresponding multiple denoisers; the target core image data is input into the converter to perform noise conversion on the target core image data; the noise-converted target core image data is input into the multiple denoisers to obtain multiple candidate denoised core images; based on the observability of the multiple candidate denoised core images, the multiple candidate denoised core images are fused to obtain a denoised core image of the target core image data; the observability is used to represent the degree of denoising deviation of the candidate denoised core images. Compared with existing methods, the embodiments of this specification can accurately distinguish between noise and core structure in core image data through the diffusion denoising model, and then perform noise conversion on the composite real-world noise therein. In addition, the embodiments of this specification can also configure different denoisers according to the number of sampling times to obtain multiple candidate denoised core images, and fuse multiple candidate denoised core images according to the degree of denoising deviation of the multiple candidate denoised core images, thereby reducing the impact of random deviations in the multiple candidate denoised core images on the denoising results, greatly improving the denoising quality of the core image data.

[0119] Based on the above core image denoising method, this specification also proposes an embodiment of a core image denoising device. Figure 14 As shown, the core image denoising device 1400 may specifically include the following modules:

[0120] The construction module 1401 is used to construct a diffusion denoising model based on the sample core image data; the diffusion denoising model includes a converter and a denoiser.

[0121] A configuration module 1402 is configured to configure the denoiser according to a plurality of sampling times to obtain a plurality of corresponding denoisers;

[0122] The conversion module 1403 is used to input the target core image data into the converter to perform noise conversion on the target core image data.

[0123] The input module 1404 is configured to input the noise-converted target core image data into the multiple denoisers to obtain multiple candidate denoised core images.

[0124] The fusion module 1405 is used to fuse the multiple candidate denoised core images according to their observability to obtain a denoised core image of the target core image data; the observability is used to indicate the denoising deviation degree of the candidate denoised core images.

[0125] In some embodiments, the above-mentioned construction module 1401 can be specifically used for sample core images to be denoised to construct a noise conversion model; the noise conversion model includes an encoder and a decoder, the encoder is used to perform noise conversion on the sample core images to be denoised, and the decoder is used to recover the sample core images to be denoised from the noise conversion results; based on the sample denoised core images, a tiled diffusion model is constructed; the tiled diffusion model includes a diffuser and a sampler; the diffuser is used to add noise to the sample denoised core images, and the sampler is used to denoise the added noise results; based on the encoder of the noise conversion model, the converter is constructed; based on the sampler of the tiled diffusion model, the denoiser is constructed.

[0126] In some embodiments, the construction module 1401 may also be used to input the sample denoised core image into a preset Gaussian noise adder to obtain a Gaussian noise core image; and construct a noise conversion model based on the sample denoised core image and the Gaussian noise core image using the following formula:

[0127]

[0128] Where x represents the sample core image to be denoised, y represents the Gaussian noise core image, ε~Ν(0,I), represents the encoder, θ1 represents the parameters of the encoder, represents the decoder, θ2 represents the parameters of the decoder, σ represents the noise level of the sample core image to be denoised, Ν(0,I) represents the standard Gaussian distribution, λ represents the hyperparameter, and R(y) represents the regularization function.

[0129] In some embodiments, the construction module 1401 may be further configured to construct a tiled diffusion model based on the sample denoised core image using the following formula:

[0130]

[0131] Where y0 represents the initial image block of the sample denoised core image, y t represents the image block of the sample denoised core image after the t-th time step of the diffuser denoising / the t-th time step of the sampler, ε~Ν(0,I), Ν(0,I) represents the standard Gaussian distribution, ε θ represents the tile diffusion model, θ represents the parameters of the tile diffusion model, and [O] represents adjacent image blocks and The overlapping area pixels, represents the edge area pixels of the image block, and λ1 and λ2 represent hyper parameters.

[0132] In some embodiments, the construction module 1401 may be further configured to calculate the noise level corresponding to each sampling block in the sampler using the following formula:

[0133]

[0134] Where t∈[0,T], T+1 represents the total number of sampling blocks in the sampler, and d t represents the noise level corresponding to the t-th sampling block, Representing the noise attenuation coefficient corresponding to the Tt-th time step to which the t-th sampling block belongs; calculating the matching degree between the noise level corresponding to each sampling block and the noise level of the target core image data; taking the time step of the sampling block corresponding to the maximum matching degree as the time step threshold; in the sampler, determining a plurality of sampling blocks whose time steps are less than or equal to the time step threshold; constructing the denoiser based on the plurality of sampling blocks whose time steps are less than or equal to the time step threshold.

[0135] In some embodiments, the above-mentioned configuration module 1402 can be specifically used to generate a sampling sequence of preset length based on the Monte Carlo algorithm; the sampling sequence includes multiple sampling times; the denoiser is configured according to the multiple sampling times in the sampling sequence to obtain multiple denoisers with different numbers of sampling blocks.

[0136] In some embodiments, the conversion module 1403 may be specifically configured to input the noise-converted target core image data into the multiple denoisers with different numbers of sampling blocks to obtain multiple candidate denoised core images.

[0137] In some embodiments, the fusion module 1405 may be specifically configured to fuse multiple candidate denoised core images using the following formula based on the observability of each candidate denoised core image to obtain a denoised core image of the target core image data:

[0138]

[0139] Where x represents the target core image data, y fin represents the denoised core image of the target core image data, represents the i-th candidate denoised core image, represents the observability of the i-th candidate denoised core image, and M represents the total number of candidate denoised core images.

[0140] In some embodiments, the fusion module 1405 may be further configured to construct a tiled diffusion model based on the sample denoised core image; the tiled diffusion model includes a diffuser and a sampler; the diffuser is configured to add noise to the sample denoised core image, and the sampler is configured to denoise the added noise result; based on the sampler, a Gaussian noise model is constructed using the following formula:

[0141] p(x|y)~Ν(y,σ 2 I);

[0142] Where x represents the target core image data, y represents the denoised core image, σ represents the noise level of the target core image data, and N(0, I) represents the standard Gaussian distribution. Each candidate denoised core image is input into the Gaussian noise model to obtain the observability of the candidate denoised core image.

[0143] As can be seen from the above, based on the core image denoising device provided in the embodiments of this specification, a diffusion denoising model can be constructed based on sample core image data; the diffusion denoising model includes a converter and a denoiser; the denoiser is configured according to multiple sampling times to obtain corresponding multiple denoisers; the target core image data is input into the converter to perform noise conversion on the target core image data; the noise-converted target core image data is input into the multiple denoisers to obtain multiple candidate denoised core images; based on the observability of the multiple candidate denoised core images, the multiple candidate denoised core images are fused to obtain a denoised core image of the target core image data; the observability is used to represent the degree of denoising deviation of the candidate denoised core images. Compared with existing methods, the embodiments of this specification can accurately distinguish between noise and core structure in core image data through the diffusion denoising model, and then perform noise conversion on the composite real-world noise therein. In addition, the embodiments of this specification can also configure different denoisers according to the number of sampling times to obtain multiple candidate denoised core images, and fuse multiple candidate denoised core images according to the degree of denoising deviation of the multiple candidate denoised core images, thereby reducing the impact of random deviations in the multiple candidate denoised core images on the denoising results, greatly improving the denoising quality of the core image data.

[0144] It should be noted that the units, devices or modules described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. For the convenience of description, the above devices are described in terms of functions and are divided into various modules and described separately. Of course, when implementing this specification, the functions of each module can be implemented in the same or multiple software and / or hardware, or the module that implements the same function can be implemented by a combination of multiple sub-modules or sub-units. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0145] The embodiments of this specification also provide a computer device for a core image denoising method, comprising a processor and a memory for storing processor executable instructions. When the processor is specifically implemented, it can perform the following steps according to the instructions: constructing a diffusion denoising model based on sample core image data; the diffusion denoising model includes a converter and a denoiser; configuring the denoiser according to multiple sampling times to obtain corresponding multiple denoisers; inputting target core image data into the converter to perform noise conversion on the target core image data; inputting the target core image data after noise conversion into the multiple denoisers to obtain multiple candidate denoised core images; based on the observability of the multiple candidate denoised core images, fusing the multiple candidate denoised core images to obtain a denoised core image of the target core image data; the observability is used to represent the degree of denoising deviation of the candidate denoised core images.

[0146] In order to complete the above instructions more accurately, refer to Figure 15 As shown, the embodiment of this specification also provides another specific computer device 1500, wherein the computer device 1500 includes a network communication port 1501, a processor 1502 and a memory 1503, and the above structures are connected through internal cables so that each structure can perform specific data interaction.

[0147] The processor 1502 can be specifically configured to construct a diffusion denoising model according to sample core image data; the diffusion denoising model comprises a converter and a denoiser; the denoiser is configured according to a plurality of sampling times to obtain a plurality of corresponding denoisers; target core image data is input into the converter to perform noise conversion on the target core image data; the target core image data after noise conversion is input into the plurality of denoisers to obtain a plurality of candidate denoised core images; the plurality of candidate denoised core images are fused according to observability of the plurality of candidate denoised core images to obtain a denoised core image of the target core image data; and the observability is used to represent a denoising deviation degree of the candidate denoised core image.

[0148] The memory 1503 can be specifically configured to store corresponding instruction programs.

[0149] In this embodiment, the network communication port 1501 can be a virtual port that is bound with different communication protocols, so as to send or receive different data. For example, the network communication port can be a port responsible for web data communication, can also be a port responsible for FTP data communication, and can also be a port responsible for mail data communication. In addition, the network communication port can also be an entity communication interface or a communication chip. For example, it can be a wireless mobile network communication chip such as GSM, CDMA, etc.; it can also be a Wifi chip; and it can also be a Bluetooth chip.

[0150] In this embodiment, the processor 1502 can be implemented in any appropriate manner. For example, the processor can take the form of, for example, a microprocessor or processor and a computer readable medium storing computer readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, an Application Specific Integrated Circuit (ASIC), a programmable logic controller, and an embedded microcontroller, and the like. The present specification is not limited thereto.

[0151] In this embodiment, the memory 1503 comprises a volatile memory and a non-volatile memory. The memory 1503 can comprise a plurality of levels, and in a digital system, as long as it can save binary data, it can be a memory; in an integrated circuit, a circuit without a physical form and having a storage function is also called a memory, such as RAM, FIFO, and the like; and in a system, a storage device with a physical form is also called a memory, such as a memory stick, a TF card, and the like.

[0152] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0153] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0154] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0155] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0156] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A core image denoising method, characterized in that: The method comprises: A diffusion denoising model is constructed based on sample core image data; the diffusion denoising model includes a converter and a denoiser; the sample core image data includes a sample denoised core image; constructing the diffusion denoising model includes: constructing a tiled diffusion model based on the sample denoised core image; the tiled diffusion model includes a diffuser and a sampler; the diffuser is used to add noise to the sample denoised core image, and the sampler is used to denoise the added noise result; constructing the denoiser based on the sampler of the tiled diffusion model, including: calculating the noise level corresponding to each sampling block in the sampler using the following formula: Where, , Indicates the total number of sampling blocks in the sampler, Indicates the The noise level corresponding to the sampling block, Indicates the The sampling block belongs to a noise attenuation coefficient corresponding to a time step; calculating the matching degree between the noise level corresponding to each sampling block and the noise level of the target core image data; using the time step of the sampling block corresponding to the maximum matching degree as a time step threshold; determining, in the sampler, a plurality of sampling blocks whose time steps are less than or equal to the time step threshold; and constructing the denoiser based on the plurality of sampling blocks whose time steps are less than or equal to the time step threshold; Configuring the denoiser according to a plurality of sampling times to obtain a corresponding plurality of denoisers; Inputting the target core image data into the converter to perform noise conversion on the target core image data; Inputting the noise-converted target core image data into the multiple denoisers to obtain multiple candidate denoised core images; According to the observability of the multiple candidate denoised core images, the multiple candidate denoised core images are fused to obtain a denoised core image of the target core image data; the observability is used to represent the denoising deviation degree of the candidate denoised core images.

2. The method according to claim 1, characterized in that The sample core image data also includes a sample core image to be denoised corresponding to the sample denoised core image; The constructing of the diffusion denoising model further includes: Constructing a noise conversion model based on the sample core image to be denoised; the noise conversion model includes an encoder and a decoder, the encoder is used to perform noise conversion on the sample core image to be denoised, and the decoder is used to recover the sample core image to be denoised from the noise conversion result; The converter is constructed based on an encoder of the noise conversion model.

3. The method according to claim 2, characterized in that The noise conversion model is constructed based on the sample core image to be denoised, including: The sample denoised core image is input into a preset Gaussian noise adder to obtain a Gaussian noise core image; Based on the sample core image to be denoised and the Gaussian noise core image, the noise conversion model is constructed using the following formula: ； Where, represents the sample core image to be denoised, represents the Gaussian noise core image, , represents the encoder, , represents the parameters of the encoder, Describes the decoder, Denotes the decoder parameters, represents the noise level of the sample core image to be denoised, represents the standard Gaussian distribution, represents the hyperparameter, represents the regularization function.

4. The method according to claim 2, characterized in that The step of constructing a tiled diffusion model based on the sample denoised core image includes: Based on the sample denoised core image, the tiled diffusion model is constructed using the following formula: ； Where, represents the initial image partition of the sample denoised core image, Indicates the diffuser Step noise / sampler Image blocks of sample denoised core images after denoising for time steps, , represents the standard Gaussian distribution, represents the tiled diffusion model, represents the parameters of the tiled diffusion model, Represents adjacent image blocks and The overlapping area pixels, Represents the edge area pixels of the image block, and represents a hyperparameter.

5. The method according to claim 1, characterized in that: The denoiser is configured according to a plurality of sampling times to obtain a plurality of corresponding denoisers, including: Generate a sampling sequence of a preset length based on a Monte Carlo algorithm; the sampling sequence includes multiple sampling times; Configuring the denoiser according to various sampling times in the sampling sequence to obtain various denoisers with different numbers of sampling blocks; The target core image data after noise conversion is input into the multiple denoisers to obtain multiple candidate denoised core images, including: The target core image data after noise conversion is input into the multiple denoisers with different numbers of sampling blocks to obtain multiple candidate denoised core images.

6. The method according to claim 1, characterized in that The step of fusing the plurality of candidate denoised core images according to the observability of the plurality of candidate denoised core images to obtain a denoised core image of the target core image data comprises: According to the observability of each candidate denoised core image, multiple candidate denoised core images are fused using the following formula to obtain the denoised core image of the target core image data: ； Where, represents the target core image data, represents the denoised core image of the target core image data, Indicates the candidate denoised core images, Indicates the Observability of candidate denoised core images, Represents the total number of candidate denoised core images.

7. The method according to claim 1, characterized in that The method further comprises: Constructing a tiled diffusion model based on the sample denoised core image; the tiled diffusion model includes a diffuser and a sampler; the diffuser is used to add noise to the sample denoised core image, and the sampler is used to denoise the added noise result; Based on the sampler, the Gaussian noise model is constructed using the following formula: ； Where, represents the target core image data, represents the denoised core image, represents the noise level of the target core image data, represents the standard Gaussian distribution; Each candidate denoised rock core image is input into the Gaussian noise model to obtain the observability of the candidate denoised rock core image.

8. A core image denoising device, characterized in that: The device comprises: A construction module is provided for constructing a diffusion denoising model based on sample core image data; the diffusion denoising model includes a converter and a denoiser; the sample core image data includes a sample denoised core image; constructing the diffusion denoising model includes: constructing a tiled diffusion model based on the sample denoised core image; the tiled diffusion model includes a diffuser and a sampler; the diffuser is used to add noise to the sample denoised core image, and the sampler is used to denoise the added noise result; constructing the denoiser based on the sampler of the tiled diffusion model, including: calculating the noise level corresponding to each sampling block in the sampler using the following formula: Where, , Indicates the total number of sampling blocks in the sampler, Indicates the The noise level corresponding to the sampling block, Indicates the The sampling block belongs to a noise attenuation coefficient corresponding to a time step; calculating the matching degree between the noise level corresponding to each sampling block and the noise level of the target core image data; using the time step of the sampling block corresponding to the maximum matching degree as a time step threshold; determining, in the sampler, a plurality of sampling blocks whose time steps are less than or equal to the time step threshold; and constructing the denoiser based on the plurality of sampling blocks whose time steps are less than or equal to the time step threshold; A configuration module, configured to configure the denoiser according to a plurality of sampling times to obtain a plurality of corresponding denoisers; a conversion module, configured to input the target core image data into the converter to perform noise conversion on the target core image data; an input module, configured to input the noise-converted target core image data into the plurality of denoisers to obtain a plurality of candidate denoised core images; A fusion module is used to fuse multiple candidate denoised core images according to their observability to obtain a denoised core image of the target core image data; the observability is used to represent the denoising deviation degree of the candidate denoised core images.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

Images