A controllable recovery method, device, equipment and medium for low-sampling spectrum situation map

Abstract

The present invention discloses a controllable restoration method, device, equipment and medium for a low-sampling spectrum situation map, which belongs to the low-sampling image restoration technology. The method includes: obtaining a low-sampling image; restoring the low-sampling image using different interpolation methods to obtain images with different degrees of interpolation adaptation; filtering to find a more satisfactory image from the adapted image, and converting the found image into a conditional image of an image-generated image; inputting the conditional image into a diffusion model based on particle filtering for processing, and outputting the final restored image. The present invention enhances the capabilities of existing text and image-to-image diffusion models through interpolation adaptation, enhances the weak prior of two-dimensional images, not only improves the overall performance, but also enables the diffusion model to effectively handle complex scenes involving weak prior appearances. The present invention proposes integrating a diffusion model adaptation method combined with particle filtering into a pre-trained image-to-image diffusion model to further improve the restoration quality.

Description

Technical Field

[0001] The present invention relates to low-sampling image restoration technology, and in particular to a controllable restoration method, device, equipment and medium for a low-sampling spectrum situation map. Background Art

[0002] With the rapid advancement of generative techniques, researchers have increasingly focused on diffusion models for visual generation tasks, particularly with significant progress in text-to-image (T2I). However, generating high-quality visual content from text or images remains a challenging task. This requires not only generating visually realistic and smooth appearances from a given text or image, but also simultaneously considering diversity and controllability. Early developments in image diffusion models focused on a single guide, either from an image or text. Stable diffusion models leverage augmented training data to enhance the capabilities of diffusion models, thereby improving generation results. ControlNet, GLIGEN, and IP-Adapter extend these approaches by integrating multiple types of conditional guides, such as text, images, edge maps, and depth maps, providing greater flexibility in image generation. However, despite their versatility, these methods tend to rely primarily on text or other conditions to generate diverse image appearances, and struggle to produce high-quality results when the image appearances are undersampled. For example, if the sampling rate is 30% to 50%, the corresponding conditional images may produce satisfactory results. However, when the sampling rate is changed to 5%, the models often fail to achieve high-quality restoration quality. This weak prior on the conditional image introduces additional challenges to the diffusion process.

[0003] Additionally, some researchers have discovered the great potential of deep learning-based techniques. For example, the UNet-based generative adversarial network (GAN) image generator. However, this approach often encounters challenges related to mode collapse, significantly limiting its effectiveness. Secondly, generative techniques based on autoregressive models that learn data distributions have been proposed. While this can improve data quality and stabilize model convergence, it requires considerable computational resources. Summary of the Invention

[0004] In order to at least solve one of the technical problems existing in the prior art to a certain extent, the object of the present invention is to provide a method, device, equipment and medium for controllable recovery of a low-sampling spectrum situation map based on a diffusion model.

[0005] The first technical solution adopted by the present invention is:

[0006] A controllable recovery method for a low-sampling spectrum situation map comprises the following steps:

[0007] Get a downsampled image;

[0008] Different interpolation methods are used to restore low-sampled images to enhance the weak prior of appearance in the image generation process and obtain images with different degrees of interpolation adaptation;

[0009] Filter the adapted images to find a satisfactory image, and convert the found image into an image-to-image (I2I) conditional image;

[0010] The conditional image is input into the diffusion model based on particle filtering for processing, and the final restored image is output.

[0011] Furthermore, the step of inputting the conditional image into a diffusion model based on particle filtering for processing and outputting a final restored image includes:

[0012] Set the time step B and particle size S. In general diffusion models, S defaults to 1, which means that the average sample is regarded as the final result. If S is large enough, the final error becomes smaller.

[0013] When the diffusion model is reversed, as the time step t changes from N to 1, an average sample is generated from the distribution Where, i∈[1,S];

[0014] Through the particle filtering method, the samples Randomly select S middle samples as replacement samples;

[0015] These S replacement samples are expressed as

[0016] From S replacement samples The final satisfactory image χ0 is obtained.

[0017] Furthermore, the particle filtering method works as follows:

[0018] Particle initialization and state prediction: The initial particles represent the initial hypothetical state of the data. In image restoration tasks, particles can represent the initial low-quality or noisy image. Each particle has its own weight, which is usually equal initially. Based on the state transition model, the particle state is predicted (noise is added), allowing the particle to evolve during the next diffusion process.

[0019] The forward process of the diffusion model: Similar to the standard diffusion model, noise is gradually added to the hypothetical state of the particle, causing the state to gradually diffuse to a completely random state;

[0020] Particle weight update: In each diffusion step, the particle weight is updated based on the observation (such as low-resolution or partial data) and the current particle state. The weight update reflects the consistency between the particle state and the true observation value. The higher the weight, the more likely the particle state is to be close to the true target data.

[0021] Resampling and replacement: When the weights of some particles are low, particles are regenerated through the resampling mechanism;

[0022] Reverse diffusion process: In the reverse process of the diffusion model, particles gradually remove noise based on the observed data; at this time, the particle filter controls the denoising process through the particle state estimation, making the final state closer to the actual target data.

[0023] Furthermore, different interpolation methods include the nearest neighbor interpolation method (NN), the ordinary kriging interpolation method (OK), or the radial basis function interpolation method (RBF).

[0024] Furthermore, a suitable interpolation method is selected: one or more interpolation methods are selected for restoration according to the specific characteristics and data distribution of the low-sampled image;

[0025] Fusion of multiple interpolation methods: Combining multiple interpolation methods to generate multiple restored versions of the image, thereby improving the diversity of restoration; each interpolation method can adapt to different regions to improve the overall effect.

[0026] Furthermore, the selecting one or more interpolation methods for restoration includes:

[0027] For simple scenes, select the nearest neighbor interpolation method for restoration; for complex or random scenes, select the Kriging interpolation method or the radial basis function interpolation method for restoration.

[0028] Furthermore, filtering and finding a more satisfactory image from the adapted images includes:

[0029] Index screening criteria: Screening the images generated by interpolation according to the predetermined index evaluation criteria;

[0030] Automatic scoring mechanism: Automatically score multiple interpolated images and select the image that is closest to the real scene;

[0031] Final screening: Based on the index screening criteria or automatic scoring results, one or more images are selected as the initial conditions for subsequent processing. These images will serve as the conditional input for the image generation process, ensuring that the final restored image has high quality.

[0032] The second technical solution adopted by the present invention is:

[0033] A controllable recovery device for a low-sampling spectrum situation map, comprising:

[0034] Image acquisition module, acquiring low-sampled images;

[0035] The interpolation adaptation module uses different interpolation methods to restore the low-sampled image to enhance the weak prior of appearance in the image generation process and obtain images with different degrees of interpolation adaptation;

[0036] The image filtering module filters the adapted image to find a more satisfactory image and converts the found image into an image-to-image (I2I) conditional image;

[0037] The image restoration module inputs the conditional image into the diffusion model based on particle filtering for processing and outputs the final restored image.

[0038] The third technical solution adopted by the present invention is:

[0039] An electronic device comprises a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, at least one program, the code set or the instruction set is loaded and executed by the processor to implement a controllable recovery method of a low-sampling spectrum situation map as described above.

[0040] The fourth technical solution adopted by the present invention is:

[0041] A computer-readable storage medium stores at least one instruction, at least one program, code set or instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by a processor to implement a controllable recovery method of a low-sampling spectrum situation map as described above.

[0042] The fifth technical solution adopted by the present invention is:

[0043] A computer program product or computer program includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the above-mentioned method for controllable recovery of a low-sampling spectrum situation map.

[0044] The present invention enhances the capabilities of existing text-to-image and image-to-image diffusion models through interpolation adaptation, strengthening the weak priors of two-dimensional images. This not only improves overall performance but also enables the diffusion model to effectively handle complex scenes involving weak appearance priors. The present invention proposes a diffusion model adaptation method that combines particle filtering and integrates it into a pre-trained image-to-image diffusion model, further improving restoration quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following introduction is made to the drawings of the embodiments of the present invention or the related technical solutions in the prior art. It should be understood that the drawings introduced below are only for the convenience of clearly describing some embodiments of the technical solutions of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative work.

[0046] Figure 1 This is a schematic diagram of the results of the image restoration method of the ControlNet diffusion model;

[0047] Figure 2 Schematic diagram of a controllable recovery algorithm for a low-sampling spectrum situation map based on a diffusion model in an embodiment of the present invention;

[0048] Figure 3 This is the interpolation adaptation phase process in an embodiment of the present invention;

[0049] Figure 4 This is the process of the diffusion model adaptation phase in an embodiment of the present invention;

[0050] Figure 5 This is a flowchart of the steps of a controllable recovery method of a low-sampling spectrum situation map in an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention and are not to be construed as limiting the present invention. The step numbers in the following embodiments are provided for ease of explanation only and do not limit the order of the steps. The order of execution of the steps in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0052] In the description of the present invention, it should be understood that descriptions involving orientations, such as up, down, front, back, left, right, etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, they cannot be understood as limitations on the present invention.

[0053] In the description of the present invention, "several" means one or more, "many" means more than two, "greater than," "less than," and "exceed" are understood to exclude the number itself, while "above," "below," and "within" are understood to include the number itself. The use of "first" and "second" in the description is solely for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance, implicitly specifying the number of the indicated technical features, or implicitly specifying the order of the indicated technical features.

[0054] In the description of the present invention, unless otherwise clearly defined, terms such as setting, installing, and connecting should be understood in a broad sense, and technicians in the relevant technical field can reasonably determine the specific meanings of the above terms in the present invention based on the specific content of the technical solution.

[0055] To address existing technical issues, this paper proposes a technical solution that generates high-quality and highly consistent images through interpolation adaptation and particle filtering. In short, the goal of this paper is to enhance the weak priors provided by text or images by incorporating an additional interpolation adaptation. This interpolation adaptation ensures better appearance details of the image before inputting it into the diffusion model. Furthermore, to effectively improve image restoration quality, a particle filtering method is combined with the diffusion model to adapt it, enabling it to obtain accurate posterior sampling.

[0056] The great success of diffusion models in high-dimensional data such as image generation provides ideas for this. These models provide strong data priors. Specifically, the stable diffusion model has achieved good results by optimizing the training dataset, greatly improving the efficiency of image restoration. The rapid progress of generation technology has promoted significant progress and continuous innovation in various fields of image generation, such as text-to-image generation (T2I), image-to-image generation (I2I), and ControlNet, which has increased the diversity of image generation by using additional conditional information, such as depth, posture, trajectory and frequency. Figure 1 As shown in , the edge map is used as the conditional image, but these techniques are all in the direction of diverse image-to-image generation (see Figure 1 The first row), while the present invention mainly tends to controllable generation of low-sampled images to images without text (see Figure 1 The generated text for the first row of images is "a blue bird", while the generated text for the second row of images is "Restore original image".

[0057] Diffusion Posterior Sampling (DPS) offers a novel approach to this problem, extending the diffusion solver to effectively handle general noisy (non-)linear inverse problems through an approximation of posterior sampling. Interestingly, the resulting posterior sampling scheme is a hybrid of diffusion sampling and manifold-constrained gradients, without the strict measurement consistency projection step, resulting in more optimal generation paths in noisy environments compared to previous studies.

[0058] The goal of this invention is to restore high-quality original images using the latest T2I diffusion model. It also uses interpolation adaptation to mitigate the limitations of the T2I model in complex scenes due to weak prior image appearance, thereby improving the visual accuracy of image restoration. Secondly, this invention utilizes diffusion model adaptation, a clever technique that can skillfully refine the diffusion model's solution space for image restoration. This not only improves overall performance but also enables the image diffusion model to expertly locate complex scenes involving weak prior appearance. The interpolated and adapted images are used to guide the T2I model, eliminating the need for additional training or fine-tuning, thereby optimizing overall processing speed.

[0059] Example 1

[0060] like Figure 5 As shown, this embodiment provides a controllable recovery method for a low-sampling spectrum situation map, comprising the following steps:

[0061] S1. Obtain a low-sampled image.

[0062] First, collect low-sampled images as source images.

[0063] S2. Use different interpolation methods to restore the low-sampled image to enhance the weak prior of appearance in the image generation process and obtain images with different degrees of interpolation adaptation.

[0064] An auxiliary interpolation adaptation method is used to enhance the weak prior of appearance in the entire complex image generation process, and obtain images with different degrees of interpolation adaptation. Different interpolation methods include nearest neighbor interpolation method (NN), ordinary kriging interpolation method (OK) or radial basis function interpolation method (RBF).

[0065] Exemplarily, step S2 specifically includes the following steps:

[0066] S21. Select an appropriate interpolation method: Based on the specific characteristics and data distribution of the low-sampled image, select one or more interpolation methods for restoration. Specifically, for simple scenes, select the nearest neighbor interpolation method for restoration; for complex or random scenes, select the Kriging interpolation method or the radial basis function interpolation method for restoration.

[0067] S22. Fusion of multiple interpolation methods: Combine multiple interpolation methods to generate multiple restored versions of the image, thereby improving the diversity of restoration; among them, each interpolation method can adapt to different areas to improve the overall effect.

[0068] S3. Filter and find a more satisfactory image from the adapted image, and convert the found image into a conditional image of the image-generated image.

[0069] A more satisfactory image is found by filtering the interpolation adaptation method and converting it into an image-to-image (I2I) conditional image. As an optional implementation, step S3 specifically includes the following steps:

[0070] S31, index screening criteria: screening the images generated by interpolation according to predetermined index evaluation criteria;

[0071] S32, automatic scoring mechanism: automatically scoring multiple interpolated images and selecting the image that is closest to the real scene;

[0072] S33, Final Screening: Based on the index screening criteria or automatic scoring results, one or more images are selected as the initial conditions for subsequent processing. These images will serve as the conditional input for the image generation process to ensure that the final restored image has high quality.

[0073] S4. Input the conditional image into the diffusion model based on particle filtering for processing, and output the final restored image.

[0074] The diffusion model with particle filtering is used to better maintain the strong consistency of I2I image generation, and finally restore a more satisfactory image.

[0075] In some embodiments, step S4 specifically includes:

[0076] S41. Set the time step N and particle size S. In general diffusion models, S defaults to 1, which means that the average sample is regarded as the final result. If S is large enough, the final error becomes smaller.

[0077] S42. When the diffusion model is reversed, an average sample is generated from the distribution at each step as the time step t changes from N to 1.

[0078] S43, through the particle filtering method, from the sample Randomly select S middle samples as replacement samples;

[0079] S44, these S replacement samples are expressed as

[0080] S45, replace samples from S The final satisfactory image χ0 is obtained.

[0081] As an optional implementation, step S43 specifically includes the following steps:

[0082] S431. Particle initialization and state prediction: Initial particles represent the initial hypothetical state of the data. In image restoration tasks, particles can represent the initial low-quality or noisy image. Each particle has its own weight, which is usually equal initially. Based on the state transition model, the particle state is predicted (with noise added), allowing the particle to evolve during the next diffusion process.

[0083] S432, Forward process of diffusion model: Similar to the standard diffusion model, noise is gradually added to the hypothetical state of the particle, causing the state to gradually diffuse to a completely random state;

[0084] S433, Particle Weight Update: In each diffusion step, the particle weight is updated based on the observation (such as low-resolution or partial data) and the current particle state. The weight update reflects the consistency between the particle state and the true observation value. The higher the weight, the more likely the particle state is to be close to the true target data.

[0085] S434, Resampling and Replacement: When the weights of some particles are low, particles are regenerated through the resampling mechanism;

[0086] S435, Reverse Diffusion Process: In the reverse process of the diffusion model, particles gradually remove noise based on the observed data; at this time, the particle filter controls the denoising process through the state estimation of the particles, making the final state closer to the actual target data.

[0087] The method of this embodiment not only improves the overall performance, but also enables the diffusion model to effectively handle complex scenes involving weak prior appearance without the need for additional training or fine-tuning.

[0088] The above method is explained in detail below with reference to the accompanying drawings and specific embodiments.

[0089] See also Figure 2 , Figure 2 A controllable restoration method for 2D low-sampling spectrum situation map based on diffusion model is presented. The method specifically includes interpolation adaptation stage and diffusion model adaptation stage.

[0090] (1) Interpolation adaptation stage

[0091] The goal of this process is to interpolate the input image. In this embodiment, we mainly consider the interpolation quality and interpolation time to obtain the desired image. Ultimately, we select the most satisfactory method as our interpolation adaptation method. This stage specifically includes the following steps:

[0092] Step 1: Collect downsampled images.

[0093] Step 2: Get the adapted image. First, restore the collected image using different interpolation methods to get the adapted image, such as Figure 3 shown.

[0094] Step 3: Select the adapted images interpolated by different methods for filtering. Consider the random uncertainty and interpolation effect brought by different interpolation methods, and finally obtain the most satisfactory image.

[0095] (2) Diffusion model adaptation stage

[0096] Particle filter-based methods use a customized proposal distribution to obtain more accurate samples. Figure 4 As shown, this stage specifically includes the following steps:

[0097] Step 1: Input image y0, set time step N, particle size S;

[0098] Step 2: Generate samples q(y 1:N |y0) in the sample {y k} sequence, k∈[1,N];

[0099] Step 3: According to the sequence {y k Generate sample

[0100] Step 4: Standard diffusion model, with each step where k changes from N to 1, a sample is generated from the distribution The particle filtering method is used to extract samples from Randomly select S replacement samples from ; denote these S replacement samples as

[0101] Step 5: From S new samples The final image χ0 is obtained from .

[0102] In summary, the method of the present invention has the following advantages and beneficial effects compared to the existing technical solutions:

[0103] (1) The present invention enhances the capabilities of existing text and image-to-image diffusion models through interpolation adaptation. It first applies an auxiliary interpolation method, the specific task of which is to enhance the weak prior of the two-dimensional image.

[0104] (2) The present invention proposes a diffusion model adaptation method combined with particle filtering and integrated into a pre-trained image-to-image diffusion model to further improve the restoration quality.

[0105] (3) The system framework of the present invention has good low-sampling image recovery capabilities for diffusion models. Furthermore, the framework can be used in any scenario. It only needs to be fully deployed and fine-tuned before customization can be achieved to meet the needs of different scenarios. Furthermore, the method of the present invention only requires image collection and fine-tuning, which is highly reproducible and can lay the foundation for subsequent real-time testing.

[0106] Example 2

[0107] This embodiment provides a controllable recovery device for a low-sampling spectrum situation map, including:

[0108] Image acquisition module, acquiring low-sampled images;

[0109] The interpolation adaptation module uses different interpolation methods to restore the low-sampled image to enhance the weak prior of appearance in the image generation process and obtain images with different degrees of interpolation adaptation;

[0110] The image filtering module filters the adapted image to find a more satisfactory image and converts the found image into a conditional image of the image-generated image;

[0111] The image restoration module inputs the conditional image into the diffusion model based on particle filtering for processing and outputs the final restored image.

[0112] Since the device is a controllable recovery device for a low-sampling spectrum situation diagram according to an embodiment of the present invention, and the principle of solving the problem by the device is similar to that of the method, the implementation of the device can refer to the implementation process of the above-mentioned method embodiment, and the repeated parts will not be repeated.

[0113] Example 3

[0114] An embodiment of the present invention further provides an electronic device, comprising a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the following Figure 5 A controllable recovery method for low-sampling spectrum situation map is shown.

[0115] It is understood that the memory may include random access memory (RAM) or read-only memory (ROM). Optionally, the memory includes a non-transitory computer-readable storage medium. The memory may be used to store instructions, programs, codes, code sets, or instruction sets. The memory may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function, instructions for implementing the various method embodiments described above, etc.; the data storage area may store data created based on the use of the server, etc.

[0116] The processor may include one or more processing cores. The processor utilizes various interfaces and circuits to connect various components within the server. It executes various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory, as well as accessing data stored in memory. Optionally, the processor may be implemented using at least one of the following hardware forms: digital signal processing (DSP), field-programmable gate array (FPGA), and programmable logic array (PLA). The processor may integrate one or a combination of a central processing unit (CPU) and a modem. The CPU primarily processes the operating system and application programs, while the modem handles wireless communications. It is understood that the modem may not be integrated into the processor and may be implemented separately via a single chip.

[0117] Since the electronic device is an electronic device corresponding to a controllable recovery method of a low-sampling spectrum situation map in an embodiment of the present invention, and the principle of solving the problem by the electronic device is similar to that of the method, the implementation of the electronic device can refer to the implementation process of the above-mentioned method embodiment, and the repeated parts will not be repeated.

[0118] Example 4

[0119] An embodiment of the present invention further provides a computer-readable storage medium, wherein the storage medium stores at least one instruction, at least one program, code set or instruction set, and the at least one instruction, the at least one program, the code set or instruction set is loaded and executed by a processor to implement the following Figure 5 A controllable recovery method for low-sampling spectrum situation map is shown.

[0120] Those skilled in the art will appreciate that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program. The program can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0121] Since the storage medium is the storage medium corresponding to the controllable recovery method of a low-sampling spectrum situation map in an embodiment of the present invention, and the principle of solving the problem by the storage medium is similar to that of the method, the implementation of the storage medium can refer to the implementation process of the above-mentioned method embodiment, and the repeated parts will not be repeated.

[0122] Example 5

[0123] In some possible implementations, various aspects of the method of the embodiments of the present invention may also be implemented in the form of a program product, which includes program code. When the program product is run on a computer device, the program code is used to cause the computer device to execute the steps of the method for controllable recovery of a low-sampling spectrum situation map according to various exemplary embodiments of the present application described above in this specification. The executable computer program code or "code" for executing each embodiment may be written in a high-level programming language such as C, C++, C#, Smalltalk, Java, JavaScript, Visual Basic, Structured Query Language (e.g., Transact-SQL), Perl, or in various other programming languages.

[0124] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0125] In the description of this specification, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0126] The above embodiments are intended only to illustrate the technical concepts and features of the present invention. Their purpose is to enable those skilled in the art to understand the contents of the present invention and implement them accordingly. They are not intended to limit the scope of protection of the present invention. Any equivalent changes or modifications made based on the essence of the present invention are intended to be covered by the scope of protection of the present invention.

Claims

1. A controllable recovery method for a low-sampling spectrum situation map, characterized in that: The following steps are involved: Get a downsampled image; Different interpolation methods are used to restore low-sampled images to enhance the weak prior of appearance in the image generation process and obtain images with different degrees of interpolation adaptation; Filter the adapted image to find a more satisfactory image, and convert the found image into a conditional image of the image-generated image; The conditional image is input into the diffusion model based on particle filtering for processing, and the final restored image is output; Among them, the particle filter method works as follows: Particle initialization and state prediction: The initial particles represent the initial hypothetical state of the data. In image restoration tasks, particles represent the initial low-quality or noisy image. Based on the state transition model, the particle state is predicted, allowing the particles to evolve during the next diffusion process. The forward process of the diffusion model: gradually add noise to the hypothetical state of the particle, so that the state gradually diffuses to a completely random state; Particle weight update: In each diffusion step, the particle weight is updated based on the observation and the current state of the particle; Resampling and replacement: When the weights of some particles are low, particles are regenerated through the resampling mechanism; Reverse diffusion process: In the reverse process of the diffusion model, particles gradually remove noise based on the observed data; particle filtering controls the denoising process through particle state estimation, making the final state closer to the actual target data.

2. The method for controllable restoration of a low-sampling spectrum situation map according to claim 1, characterized in that: The step of inputting the conditional image into a diffusion model based on particle filtering for processing and outputting a final restored image comprises: Set the time step , particle size ; When the diffusion model is in reverse process, as the time step from Each time it becomes 1, an average sample is generated from the distribution ; Through the particle filtering method, the samples Random selection The middle samples are used as replacement samples; Put this The replacement samples are represented as ; from Replacement samples Get the final satisfactory image .

3. The method for controllable restoration of a low-sampling spectrum situation map according to claim 1, characterized in that: Different interpolation methods include nearest neighbor, ordinary kriging, or radial basis functions.

4. The method for controllable restoration of a low-sampling spectrum situation map according to claim 1, characterized in that: The method of restoring the low-sampled image by using different interpolation methods includes: Select appropriate interpolation method: According to the specific characteristics and data distribution of the low-sampled image, select one or more interpolation methods for restoration; Fusion of multiple interpolation methods: Combining multiple interpolation methods to generate multiple restored versions of the image, thereby improving the diversity of restoration.

5. The method for controllable restoration of a low-sampling spectrum situation map according to claim 4, characterized in that: The selecting one or more interpolation methods for restoration includes: For simple scenes, select the nearest neighbor interpolation method for restoration; for complex or random scenes, select the Kriging interpolation method or the radial basis function interpolation method for restoration.

6. The method for controllable restoration of a low-sampling spectrum situation map according to claim 1, characterized in that: The filtering to find a more satisfactory image from the adapted image includes: Index screening criteria: Screening the images generated by interpolation according to the predetermined index evaluation criteria; Automatic scoring mechanism: Automatically score multiple interpolated images and select the image that is closest to the real scene; Final screening: Based on the index screening criteria or automatic scoring results, one or more images are selected as the initial conditions for subsequent processing.

7. A controllable recovery device for low-sampling spectrum situation diagram, characterized in that: include: Image acquisition module, acquiring low-sampled images; The interpolation adaptation module uses different interpolation methods to restore the low-sampled image to enhance the weak prior of appearance in the image generation process and obtain images with different degrees of interpolation adaptation; The image filtering module filters the adapted image to find a more satisfactory image and converts the found image into a conditional image of the image-generated image; The image restoration module inputs the conditional image into the diffusion model based on particle filtering for processing and outputs the final restored image; Among them, the particle filter method works as follows: Particle initialization and state prediction: The initial particles represent the initial hypothetical state of the data. In image restoration tasks, particles represent the initial low-quality or noisy image. Based on the state transition model, the particle state is predicted, allowing the particles to evolve during the next diffusion process. The forward process of the diffusion model: gradually add noise to the hypothetical state of the particle, so that the state gradually diffuses to a completely random state; Particle weight update: In each diffusion step, the particle weight is updated based on the observation and the current state of the particle; Resampling and replacement: When the weights of some particles are low, particles are regenerated through the resampling mechanism; Reverse diffusion process: In the reverse process of the diffusion model, particles gradually remove noise based on the observed data; particle filtering controls the denoising process through particle state estimation, making the final state closer to the actual target data.

8. An electronic device, characterized in that: The electronic device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that The storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the method according to any one of claims 1 to 6.

Images