Frequency domain adaptive super-resolution reconstruction method

Abstract

The present invention discloses a frequency domain adaptive super-resolution reconstruction method, comprising the following steps: obtaining an original image; and processing the original image through a trained frequency domain adaptive super-resolution reconstruction model to obtain a reconstructed image. In the frequency domain adaptive super-resolution reconstruction method, the multi-scale frequency domain feature extraction module of the present invention can extract multi-scale features from feature maps at different levels by combining a feature pyramid structure with Fourier transform. This multi-scale feature extraction helps the model capture information at different levels of the image and enhance its ability to recognize details. The frequency pyramid attention network decomposes the feature map into different frequency components, calculates the importance of each frequency component, and weights these components through attention weights, so that the model can pay more attention to high-frequency information that has a greater impact on the quality of image reconstruction.

Description

Technical Field

[0001] The invention belongs to the technical field of image enhancement and relates to a frequency domain adaptive super-resolution reconstruction method. Background Art

[0002] Currently, super-resolution reconstruction methods can be roughly divided into traditional methods and deep learning methods based on their development history. Traditional super-resolution reconstruction methods can be further categorized into interpolation-based methods, reconstruction-based methods, and learning-based methods. Interpolation-based super-resolution methods primarily use basis functions or interpolation kernels to approximate lost feature information, thereby achieving image reconstruction. Typical interpolation methods include nearest neighbor interpolation, bilinear interpolation, and bicubic interpolation. However, these methods are relatively simple and often result in jagged artifacts, loss of detail, and blurring in the reconstructed image, failing to effectively restore high-frequency information. Reconstruction-based super-resolution methods first simulate the image degradation process to generate a low-quality degraded image. They then introduce prior constraints on the high-resolution image through regularization methods and perform optimization to generate the corresponding high-resolution image. Typical reconstruction methods include iterative backprojection, maximum a posteriori probability, and convex set projection. However, these methods are often computationally expensive and struggle to select appropriate constraints or fully utilize prior information about the image, limiting their effectiveness and adaptability in practical applications. With the advancement of deep learning, models such as convolutional neural networks, graph neural networks, and generative adversarial networks extract feature information from input images, ultimately using reconstruction modules to generate corresponding high-resolution images. While the performance and effectiveness of these models are gradually improving during image reconstruction, they face challenges such as excessive memory consumption and computational complexity.

[0003] The emergence of generative adversarial networks (GANs) enables the generation of high-quality, realistic samples. Therefore, GANs can be used for high-resolution image generation and effectively capture details and structural information in the data. The Super-Resolution Generative Adversarial Network (SRGAN) introduced GANs to the field of super-resolution reconstruction. SRGAN uses adversarial training between a generator and a discriminator to learn how to recover a high-resolution image from a low-resolution image. Since its introduction, SRGAN has attracted widespread attention in the super-resolution field. However, its shortcomings are not negligible: the training process is unstable, difficult to converge, and prone to mode collapse; the recovery of details may be imperfect, and the generated images may contain artifacts or unnatural textures; and the high computational resource consumption leads to slow training and generation speeds. To further improve the performance of SRGAN, researchers have innovated and improved multiple aspects, including network architecture, loss function, training strategy, multi-scale feature fusion, auxiliary model introduction, application-specific optimization, and inference acceleration. Through these contributions, SRGAN has achieved significant progress in image super-resolution. The generated images are not only more visually natural and rich in details, but also have a more stable and efficient training process. Existing super-resolution reconstruction technologies usually process images in the spatial domain. The higher the image resolution, the greater the complexity of the algorithm, requiring more computing resources and running time to complete the reconstruction task. Currently, more emphasis is placed on restoration effects while ignoring model complexity, resulting in an unstable training process and difficulty in convergence.

[0004] In summary, the existing technology has the problem of over-emphasizing reconstruction quality, resulting in excessive model complexity and low reconstruction efficiency. Summary of the Invention

[0005] The purpose of the present invention is to provide a frequency domain adaptive super-resolution reconstruction method, which solves the problem in the prior art that too much emphasis is placed on reconstruction quality, resulting in excessive model complexity and low reconstruction efficiency.

[0006] The technical solution adopted by the present invention is a frequency domain adaptive super-resolution reconstruction method, comprising the following steps:

[0007] S1, obtain the original image;

[0008] S2. The original image is processed by the trained frequency domain adaptive super-resolution reconstruction model to obtain a reconstructed image.

[0009] The present invention is also characterized in that:

[0010] The frequency domain adaptive super-resolution reconstruction model includes a shallow feature extraction module, a multi-scale frequency domain feature extraction module, and a feature fusion module;

[0011] The multi-scale frequency domain feature extraction module includes several convolutional layers, feature enhancement modules and frequency pyramid attention networks;

[0012] The feature enhancement module includes several convolutional layers and activation units connected in sequence, and the activation unit is a Gaussian error linear unit activation function.

[0013] S2 includes the following steps:

[0014] S2.1, shallow feature extraction module extracts shallow features of the original image through nonlinear mapping;

[0015] S2.2, several stacked multi-scale frequency domain feature extraction modules process the shallow features of the original image to obtain the deep features of the original image;

[0016] S2.3. The feature fusion module fuses the deep features of the original image with the shallow features of the original image and rearranges them through sub-pixel convolution to obtain a reconstructed image.

[0017] The process of multi-scale frequency domain feature extraction module processing shallow features of original image is as follows:

[0018] A1. Process the shallow features of the original image through several convolutional layers to obtain the distilled features of the original image;

[0019] A2. The feature enhancement module processes the distilled features of the original image to obtain a multi-scale feature map of the original image;

[0020] A3. Process the multi-scale feature map of the original image through the frequency pyramid attention network to obtain the deep features of the original image.

[0021] A3 includes the following steps:

[0022] A3.1. Convert the original image multi-scale feature map from the spatial domain to the frequency domain through Fourier transform.

[0023] A3.2. The frequency domain features of the real and imaginary parts of the original image's multi-scale feature map are extracted through convolutional layers. These features are then enhanced and merged based on the attention weight coefficients to obtain enhanced detail features of the original image.

[0024] A3.3. Convert the original image enhancement details from the frequency domain to the spatial domain through inverse Fourier transform;

[0025] A3.4. The enhanced detail features of the original image are fused with the multi-scale feature map of the original image to obtain the deep features of the original image.

[0026] The attention weight coefficient is calculated by the fast Fourier transform loss function;

[0027] The calculation formula of the fast Fourier transform loss function is as follows:

[0028]

[0029] in, represents the fast Fourier transform loss value, represents the loss weight coefficient, represents the real part of the predicted image in the frequency domain, represents the real part of the target image in the frequency domain, represents the imaginary part of the predicted image in the frequency domain, Represents the imaginary part of the target image in the frequency domain.

[0030] The trained frequency domain adaptive super-resolution reconstruction model is obtained through the following steps:

[0031] B1. Obtain training and test sets and set model training indicators;

[0032] B2. Use the training set to train the frequency domain adaptive super-resolution reconstruction model;

[0033] B3. Input the test set into the trained frequency domain adaptive super-resolution reconstruction model to obtain a test result. If the test result meets the model training index, the training is completed and a trained frequency domain adaptive super-resolution reconstruction model is obtained. If the test result does not meet the model training index, repeat steps B2-B3.

[0034] Model training metrics include peak signal-to-noise ratio and structural similarity index.

[0035] The reconstructed image has a higher resolution than the original image.

[0036] The beneficial effects of the present invention are as follows: the multi-scale frequency domain feature extraction module in the present invention can extract multi-scale features from feature maps of different levels by combining the feature pyramid structure and Fourier transform. This multi-scale feature extraction helps the model capture different levels of information of the image and enhance the ability to recognize details; the frequency pyramid attention network decomposes the feature map into different frequency components, calculates the importance of each frequency component, and weights these components through attention weights, so that the model can pay more attention to high-frequency information that has a greater impact on the quality of image reconstruction; deep information capture and feature enhancement are achieved through multiple stacked multi-scale frequency domain feature extraction modules, and the model can deeply capture Obtaining deep information in the network is conducive to improving the details and quality of the reconstructed image, enabling the model to better understand and reconstruct the complex structure in the image, especially when processing high-resolution images, which can significantly improve the reconstruction effect; combining the fast Fourier transform loss function and sub-pixel upsampling operation, it can effectively reconstruct high-resolution images. The fast Fourier transform loss function reduces the gap between the generated image and the real image in the frequency domain, while sub-pixel upsampling improves the resolution of the reconstructed image while maintaining the details and quality of the image. This method of combining frequency domain analysis and spatial domain upsampling not only improves the efficiency of image reconstruction, but also optimizes the visual quality of the reconstructed image. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 2 is a schematic diagram of the network structure of a frequency domain adaptive super-resolution reconstruction model in the frequency domain adaptive super-resolution reconstruction method in an embodiment;

[0038] Figure 2 2 is a schematic diagram of the network structure of a multi-scale frequency domain feature extraction module in a frequency domain adaptive super-resolution reconstruction model in an embodiment;

[0039] Figure 3 2 is a schematic diagram of the network structure of the feature enhancement module in the frequency domain adaptive super-resolution reconstruction model in an embodiment;

[0040] Figure 4 2 is a schematic diagram of the network structure of the frequency pyramid attention network in the frequency domain adaptive super-resolution reconstruction model in the embodiment. DETAILED DESCRIPTION

[0041] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0042] The frequency domain adaptive super-resolution reconstruction method comprises the following steps:

[0043] S1, obtain the original image;

[0044] S2, processing the original image through the trained frequency domain adaptive super-resolution reconstruction model to obtain a reconstructed image;

[0045] S2.1, shallow feature extraction module extracts shallow features of the original image through nonlinear mapping;

[0046] S2.2, several stacked multi-scale frequency domain feature extraction modules process the shallow features of the original image to obtain the deep features of the original image;

[0047] The process of multi-scale frequency domain feature extraction module processing shallow features of original image is as follows:

[0048] A1. Process the shallow features of the original image through several convolutional layers to obtain the distilled features of the original image;

[0049] A2. The feature enhancement module processes the distilled features of the original image to obtain a multi-scale feature map of the original image;

[0050] A3. Process the multi-scale feature map of the original image through the frequency pyramid attention network to obtain the deep features of the original image;

[0051] A3.1. Convert the original image multi-scale feature map from the spatial domain to the frequency domain through Fourier transform.

[0052] A3.2. The frequency domain features of the real and imaginary parts of the original image's multi-scale feature map are extracted through convolutional layers. These features are then enhanced and merged based on the attention weight coefficients to obtain enhanced detail features of the original image.

[0053] The attention weight coefficient is calculated by the fast Fourier transform loss function; the calculation formula of the fast Fourier transform loss function is as follows:

[0054]

[0055] in, represents the fast Fourier transform loss value, represents the loss weight coefficient, represents the real part of the predicted image in the frequency domain, represents the real part of the target image in the frequency domain, represents the imaginary part of the predicted image in the frequency domain, Represents the imaginary part of the target image in the frequency domain;

[0056] A3.3. Convert the original image enhancement details from the frequency domain to the spatial domain through inverse Fourier transform;

[0057] A3.4. Fusing the original image’s enhanced detail features with the original image’s multi-scale feature map to obtain the original image’s deep features.

[0058] S2.3, the feature fusion module fuses the deep features of the original image with the shallow features of the original image and rearranges them through sub-pixel convolution to obtain a reconstructed image;

[0059] The resolution of the reconstructed image is higher than the original image;

[0060] The trained frequency domain adaptive super-resolution reconstruction model is obtained through the following steps:

[0061] B1. Obtain training and test sets and set model training indicators;

[0062] Model training metrics include peak signal-to-noise ratio and structural similarity index;

[0063] B2. Use the training set to train the frequency domain adaptive super-resolution reconstruction model;

[0064] B3. Input the test set into the trained frequency domain adaptive super-resolution reconstruction model to obtain a test result. If the test result meets the model training index, the training is completed and a trained frequency domain adaptive super-resolution reconstruction model is obtained. If the test result does not meet the model training index, repeat steps B2-B3.

[0065] The frequency domain adaptive super-resolution reconstruction model includes a shallow feature extraction module, a multi-scale frequency domain feature extraction module, and a feature fusion module;

[0066] The multi-scale frequency domain feature extraction module includes several convolutional layers, feature enhancement modules and frequency pyramid attention networks;

[0067] The feature enhancement module includes several convolutional layers and activation units connected in sequence, and the activation unit is a Gaussian error linear unit activation function.

[0068] The present invention uses several multi-scale frequency domain feature extraction modules to extract deeper and more abstract feature information from the original image; the multi-scale frequency domain feature extraction modules can enhance the representation ability of the feature map, thereby helping the model learn to recover clear images from various transformations; the multi-scale frequency domain feature extraction modules use adaptive group convolution and multi-level attention mechanisms to enhance feature extraction and fusion; in the multi-scale frequency domain feature extraction modules, the frequency pyramid attention network is used to enhance the extraction and fusion of feature information of the original image; the residual connection is processed by the feature enhancement module to retain the details and texture information of the image; the frequency pyramid attention network weights the frequency domain features so that the network can pay more attention to important feature parts; the frequency pyramid attention network enhances the detail recovery capability in the image super-resolution task by combining Fourier transform and convolution operations.

[0069] In the frequency pyramid attention network of the present invention, the input image is converted from the spatial domain to the frequency domain to capture the high-frequency detail information of the image. Then, the real and imaginary parts in the frequency domain are processed separately through three layers of convolution to extract frequency domain features and enhance details. The processed real and imaginary parts are then merged, and the frequency domain signal is restored to the spatial domain through inverse Fourier transform. Finally, the input image is fused with the processed image to improve the image quality, especially in terms of details and textures, and enhance the clarity and detail recovery effect of the image. The frequency pyramid attention network of the present invention can improve the quality of image reconstruction through the advantages of frequency domain enhancement, multi-scale feature processing, and adaptive feature selection. At the same time, compared with the traditional spatial domain convolution method, it has obvious advantages in computational efficiency, detail recovery and robustness, especially in processing high-frequency details and reducing noise.

[0070] The frequency pyramid attention network processes images in the frequency domain and uses convolution to enhance high-frequency information such as texture and edges in the frequency domain. The fast Fourier transform loss function can ensure that image details are enhanced during training by optimizing the difference between the real and imaginary parts in the frequency domain. The combination of the two can significantly improve image detail recovery, image quality and convergence speed.

[0071] The present invention calculates the difference between the predicted image and the target image in the frequency domain through the fast Fourier transform loss function to obtain the attention weight coefficient: the image is converted from the spatial domain to the frequency domain through Fourier transform, and then the absolute error between the real part and the imaginary part is calculated respectively to measure the similarity between the predicted image and the real image in the frequency domain. Finally, the error of the real and imaginary parts is summed and multiplied by the loss coefficient to obtain the fast Fourier transform loss function value as the attention weight coefficient. Since the difference in the frequency domain can often better capture the details, texture and high-frequency components of the image, the fast Fourier transform loss function helps to improve the detail recovery effect of the image.

[0072] The training set used in this invention is the high-quality DIV2k dataset, whose images each contain extremely rich texture information, making it ideal for super-resolution reconstruction training data. The test set uses the widely used Set5 and Set14 test sets in super-resolution reconstruction. The Set5 test set consists of images of people and animals, while the Set14 test set includes images of people, animals, plants, and posters. The original images used for model training, i.e., low-resolution images, are obtained by downsampling the original clear images using interpolation.

[0073] Peak Signal-to-Noise Ratio (PSNR) and Structure Similarity Index (SSIM) are two commonly used methods for evaluating the quality of image super-resolution reconstruction technology. The higher the PSNR and SSIM values, the closer the reconstructed image is to the original image, and the better the reconstruction effect.

[0074] This paper combines a feature pyramid structure with the Fourier transform to design a multi-scale frequency-domain feature extraction module (Multi-Scale Frequency-Aware Block, MFAB). This module uses a frequency pyramid attention network (FPAN) to decompose the feature map into different frequency components, calculate the importance of each frequency component, and weight the different frequency components using attention weight coefficients. The weighted frequency components are recombined to obtain an adjusted feature map. Multiple stacked MFAB modules are then used to capture deeper information in the network. The frequency pyramid attention network is combined with a fast Fourier transform loss function and sub-pixel upsampling operations to reconstruct a high-resolution image.

[0075] Example 1

[0076] This embodiment proposes a frequency domain adaptive super-resolution reconstruction method, including the following steps:

[0077] S1, obtain the original image;

[0078] S2. The original image is processed by the trained frequency domain adaptive super-resolution reconstruction model to obtain a reconstructed image.

[0079] Example 2

[0080] This embodiment proposes a frequency domain adaptive super-resolution reconstruction method, including the following steps:

[0081] S1, obtain the original image;

[0082] S2, processing the original image through the trained frequency domain adaptive super-resolution reconstruction model to obtain a reconstructed image;

[0083] like Figure 1 As shown in , the frequency domain adaptive super-resolution reconstruction model includes a shallow feature extraction module, a multi-scale frequency domain feature extraction module, and a feature fusion module; Figure 2 As shown in , the multi-scale frequency domain feature extraction module includes convolutional layers such as c1_d, c2_d, c3_d, feature enhancement module and frequency pyramid attention network; Figure 3As shown in FIG, the feature enhancement module includes a Conv_3 convolution layer, a Conv_x convolution layer and an activation unit connected in sequence, and the activation unit is a Gaussian error linear unit activation function.

[0084] Example 3

[0085] This embodiment proposes a frequency domain adaptive super-resolution reconstruction method, including the following steps:

[0086] S1, obtain the original image;

[0087] S2, processing the original image through the trained frequency domain adaptive super-resolution reconstruction model to obtain a reconstructed image;

[0088] S2.1, shallow feature extraction module extracts shallow features of the original image through nonlinear mapping;

[0089] S2.2, seven stacked multi-scale frequency domain feature extraction modules process the shallow features of the original image to obtain the deep features of the original image;

[0090] S2.3. The feature fusion module fuses the deep features of the original image with the shallow features of the original image and rearranges them through sub-pixel convolution to obtain a reconstructed image.

[0091] Example 4

[0092] This embodiment proposes a frequency domain adaptive super-resolution reconstruction method, including the following steps:

[0093] S1, obtain the original image;

[0094] S2, processing the original image through the trained frequency domain adaptive super-resolution reconstruction model to obtain a reconstructed image;

[0095] S2.1, shallow feature extraction module extracts shallow features of the original image through nonlinear mapping;

[0096] S2.2, seven stacked multi-scale frequency domain feature extraction modules process the shallow features of the original image to obtain the deep features of the original image;

[0097] The process of multi-scale frequency domain feature extraction module processing shallow features of original image is as follows:

[0098] A1. Process the shallow features of the original image through the convolution layer to obtain the distilled features of the original image;

[0099] A2. The feature enhancement module processes the distilled features of the original image to obtain a multi-scale feature map of the original image;

[0100] A3, the original image multi-scale feature map is passed through Figure 4The frequency pyramid attention network shown in the figure is used to obtain the deep features of the original image;

[0101] S2.3. The feature fusion module fuses the deep features of the original image with the shallow features of the original image and rearranges them through sub-pixel convolution to obtain a reconstructed image.

[0102] Example 5

[0103] This embodiment proposes a frequency domain adaptive super-resolution reconstruction method, including the following steps:

[0104] S1, obtain the original image;

[0105] S2, processing the original image through the trained frequency domain adaptive super-resolution reconstruction model to obtain a reconstructed image;

[0106] S2.1, shallow feature extraction module extracts shallow features of the original image through nonlinear mapping;

[0107] S2.2, seven stacked multi-scale frequency domain feature extraction modules process the shallow features of the original image to obtain the deep features of the original image;

[0108] The process of multi-scale frequency domain feature extraction module processing shallow features of original image is as follows:

[0109] A1. Process the shallow features of the original image through the convolution layer to obtain the distilled features of the original image;

[0110] A2. The feature enhancement module processes the distilled features of the original image to obtain a multi-scale feature map of the original image;

[0111] A3, the original image multi-scale feature map is passed through Figure 4 The frequency pyramid attention network shown in the figure is used to obtain the deep features of the original image;

[0112] A3.1. Convert the original image multi-scale feature map from the spatial domain to the frequency domain through Fourier transform.

[0113] A3.2. The frequency domain features of the real and imaginary parts of the original image's multi-scale feature map are extracted through convolutional layers. These features are then enhanced and merged based on the attention weight coefficients to obtain enhanced detail features of the original image.

[0114] The attention weight coefficient is calculated by the fast Fourier transform loss function; the calculation formula of the fast Fourier transform loss function is as follows:

[0115]

[0116] in, represents the fast Fourier transform loss value, represents the loss weight coefficient, represents the real part of the predicted image in the frequency domain, represents the real part of the target image in the frequency domain, represents the imaginary part of the predicted image in the frequency domain, Represents the imaginary part of the target image in the frequency domain;

[0117] A3.3. Convert the original image enhancement details from the frequency domain to the spatial domain through inverse Fourier transform;

[0118] A3.4. Fusing the original image’s enhanced detail features with the original image’s multi-scale feature map to obtain the original image’s deep features.

[0119] S2.3. The feature fusion module fuses the deep features of the original image with the shallow features of the original image and rearranges them through sub-pixel convolution to obtain a reconstructed image.

[0120] Example 6

[0121] This embodiment proposes a frequency domain adaptive super-resolution reconstruction method, including the following steps:

[0122] S1, obtain the original image;

[0123] S2, processing the original image through the trained frequency domain adaptive super-resolution reconstruction model to obtain a reconstructed image;

[0124] The resolution of the reconstructed image is higher than the original image;

[0125] The trained frequency domain adaptive super-resolution reconstruction model is obtained through the following steps:

[0126] B1. Obtain training and test sets and set model training indicators;

[0127] Model training metrics include peak signal-to-noise ratio and structural similarity index;

[0128] B2. Use the training set to train the frequency domain adaptive super-resolution reconstruction model;

[0129] B3. Input the test set into the trained frequency domain adaptive super-resolution reconstruction model to obtain a test result. If the test result meets the model training index, the training is completed and a trained frequency domain adaptive super-resolution reconstruction model is obtained. If the test result does not meet the model training index, repeat steps B2-B3.

[0130] To better verify the effectiveness and accuracy of the proposed algorithm, the FDASRN (frequency-domain adaptive super-resolution reconstruction method) of the present invention is compared with other super-resolution reconstruction methods such as Bicubic (bicubic interpolation), SRCNN (Super-Resolution Convolutional Neural Network, super-resolution convolutional neural network), ESPCN (Efficient Sub-Pixel Convolutional Neural Network, efficient sub-pixel convolutional neural network), and LapSRN (Laplacian Pyramid Super-Resolution Network). All methods use the objective standards of PSNR and SSIM to quantitatively evaluate the reconstruction results of each network model, among which the optimal value is displayed in bold.

[0131] Table 1 PSNR (dB) values ​​of different algorithms on Set5 and Set14 datasets

[0132]

[0133] Table 2 SSIM values ​​of different algorithms on Set5 and Set14 datasets

[0134]

[0135] As shown in Tables 1 and 2, when the scaling factors are 2, 3, and 4, respectively, the evaluation values ​​of the proposed algorithm and the comparison method on the test dataset show that the PSNR and SSIM values ​​of the proposed algorithm are superior to those of the comparison algorithm. This proves that the proposed network structure significantly improves the performance of the comparison algorithm and can achieve better reconstruction results.

Claims

1. Frequency domain adaptive super-resolution reconstruction method, characterized in that: The following steps are involved: S1, obtain the original image; S2, processing the original image through the trained frequency domain adaptive super-resolution reconstruction model to obtain a reconstructed image; S2.1, shallow feature extraction module extracts shallow features of the original image through nonlinear mapping; S2.2, several stacked multi-scale frequency domain feature extraction modules process the shallow features of the original image to obtain the deep features of the original image; The process of the multi-scale frequency domain feature extraction module processing the shallow features of the original image is as follows: A1. Process the shallow features of the original image through several convolutional layers to obtain the distilled features of the original image; A2. The feature enhancement module processes the distilled features of the original image to obtain a multi-scale feature map of the original image; A3. Process the multi-scale feature map of the original image through the frequency pyramid attention network to obtain the deep features of the original image; A3.

1. Convert the original image multi-scale feature map from the spatial domain to the frequency domain through Fourier transform. A3.

2. The frequency domain features of the real and imaginary parts of the original image's multi-scale feature map are extracted through convolutional layers. These features are then enhanced and merged based on the attention weight coefficients to obtain enhanced detail features of the original image. The attention weight coefficient is calculated by the fast Fourier transform loss function; The calculation formula of the fast Fourier transform loss function is as follows: in, represents the fast Fourier transform loss value, represents the loss weight coefficient, represents the real part of the predicted image in the frequency domain, represents the real part of the target image in the frequency domain, represents the imaginary part of the predicted image in the frequency domain, Represents the imaginary part of the target image in the frequency domain; A3.

3. Convert the original image enhancement details from the frequency domain to the spatial domain through inverse Fourier transform; A3.

4. Fusing the original image’s enhanced detail features with the original image’s multi-scale feature map to obtain the original image’s deep features. S2.

3. The feature fusion module fuses the deep features of the original image with the shallow features of the original image and rearranges them through sub-pixel convolution to obtain a reconstructed image.

2. The frequency domain adaptive super-resolution reconstruction method according to claim 1, characterized in that: The frequency domain adaptive super-resolution reconstruction model includes a shallow feature extraction module, a multi-scale frequency domain feature extraction module, and a feature fusion module; The multi-scale frequency domain feature extraction module includes several convolutional layers, feature enhancement modules and frequency pyramid attention networks; The feature enhancement module includes several convolutional layers and activation units connected in sequence, and the activation unit is a Gaussian error linear unit activation function.

3. The frequency domain adaptive super-resolution reconstruction method according to claim 1, characterized in that: The trained frequency domain adaptive super-resolution reconstruction model is obtained by the following steps: B1. Obtain training and test sets and set model training indicators; B2. Use the training set to train the frequency domain adaptive super-resolution reconstruction model; B3. Input the test set into the trained frequency domain adaptive super-resolution reconstruction model to obtain a test result. If the test result meets the model training index, the training is completed and a trained frequency domain adaptive super-resolution reconstruction model is obtained. If the test result does not meet the model training index, repeat steps B2-B3.

4. The frequency domain adaptive super-resolution reconstruction method according to claim 3, characterized in that: The model training indicators include peak signal-to-noise ratio and structural similarity index.

5. The frequency domain adaptive super-resolution reconstruction method according to any one of claims 1 to 4, characterized in that: The reconstructed image has a higher resolution than the original image.

Images