Video super-resolution reconstruction method, device, computer equipment and storage medium

Abstract

The present invention discloses a video super-resolution reconstruction method, apparatus, computer equipment, and storage medium. The method comprises: obtaining a video sequence to be super-resolution reconstructed; performing denoising on the video sequence using a lightweight residual denoising network; inputting the denoised video sequence into a video super-resolution network, and having the video super-resolution network output a super-resolution reconstruction feature map corresponding to the video sequence. The video super-resolution reconstruction method based on a lightweight denoising network of the present invention introduces a residual denoising network to denoise the video sequence, then constructs a video super-resolution network with bidirectional propagation and optical flow alignment, and reconstructs the denoised video sequence, thereby improving the video super-resolution reconstruction effect. The use of the lightweight residual denoising network can also accelerate the reconstruction operation speed.

Description

Technical Field

[0001] The present invention relates to the field of video processing technology, and in particular to a video super-resolution reconstruction method, device, computer equipment and storage medium. Background Art

[0002] As multimedia products become increasingly popular, varying degrees of distortion occur during video acquisition, compression, transmission, and storage, resulting in poor final video quality. Therefore, super-resolution reconstruction algorithms are needed to obtain clearer video content. Compared with traditional video super-resolution reconstruction algorithms, deep learning-based algorithms have become a more mainstream research method, with widespread applications in video refurbishment, online video playback, intelligent security, medical imaging, and other fields. However, distorted videos contain a large amount of noise, and existing research has directly performed super-resolution reconstruction on distorted videos, amplifying the impact of noise and achieving unsatisfactory results.

[0003] Video super-resolution involves converting low-resolution videos into high-resolution videos through enhancement. Its applications are extensive, including video refurbishment, smart security, medical imaging, and the popular online video player. This involves transmitting low-resolution videos over low-bandwidth conditions and then restoring them to high-resolution on the playback end using a video super-resolution algorithm.

[0004] Compared to image super-resolution reconstruction, video super-resolution reconstruction presents additional challenges, requiring consideration of frame continuity and processing speed. Currently, most video super-resolution reconstruction methods directly employ image super-resolution algorithms, using only the information from the current frame as reconstruction content. This results in less than ideal reconstruction results. Summary of the Invention

[0005] The embodiments of the present invention provide a video super-resolution reconstruction method, apparatus, computer equipment and storage medium, aiming to improve the video super-resolution reconstruction effect and operation speed.

[0006] In a first aspect, an embodiment of the present invention provides a video super-resolution reconstruction method, comprising:

[0007] Obtaining a video sequence to be super-resolution reconstructed;

[0008] Using a lightweight residual denoising network to perform denoising on the video sequence;

[0009] The denoised video sequence is input into a video super-resolution network, and the video super-resolution network outputs a super-resolution reconstructed feature map corresponding to the video sequence.

[0010] In a second aspect, an embodiment of the present invention provides a video super-resolution reconstruction device, comprising:

[0011] A video sequence acquisition unit, used to acquire a video sequence to be super-resolution reconstructed;

[0012] a video sequence denoising unit, configured to perform denoising on the video sequence using a lightweight residual denoising network;

[0013] The super-resolution reconstruction unit is used to input the denoised video sequence into a video super-resolution network, and the video super-resolution network outputs a super-resolution reconstruction feature map corresponding to the video sequence.

[0014] In a third aspect, an embodiment of the present invention provides 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 video super-resolution reconstruction method as described in the first aspect is implemented.

[0015] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the video super-resolution reconstruction method as described in the first aspect is implemented.

[0016] Embodiments of the present invention provide a video super-resolution reconstruction method, apparatus, computer device, and storage medium. The method comprises: obtaining a video sequence to be super-resolution reconstructed; performing denoising on the video sequence using a lightweight residual denoising network; inputting the denoised video sequence into a video super-resolution network, and having the video super-resolution network output a super-resolution reconstruction feature map corresponding to the video sequence. The video super-resolution reconstruction method based on a lightweight denoising network in the embodiment of the present invention introduces a residual denoising network to denoise the video sequence, then constructs a video super-resolution network with bidirectional propagation and optical flow alignment, and reconstructs the denoised video sequence, thereby improving the video super-resolution reconstruction effect. The lightweight residual denoising network is also used to accelerate the reconstruction operation speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 A schematic diagram of a flow chart of a video super-resolution reconstruction method provided by an embodiment of the present invention;

[0019] Figure 2A schematic diagram of a sub-process of a video super-resolution reconstruction method provided by an embodiment of the present invention;

[0020] Figure 3 A schematic block diagram of a video super-resolution reconstruction device provided by an embodiment of the present invention;

[0021] Figure 4 A schematic block diagram of a video super-resolution reconstruction apparatus provided by an embodiment of the present invention;

[0022] Figure 5 A schematic diagram of the network structure of a lightweight residual denoising network in a video super-resolution reconstruction method provided by an embodiment of the present invention;

[0023] Figure 6 A schematic diagram comparing test results of a video super-resolution reconstruction method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0025] It will be understood that when used in this specification and the appended claims, the terms “comprises” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0026] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the present invention. As used in the specification and appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0027] It should be further understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0028] See below Figure 1 , Figure 1 A schematic flow chart of a video super-resolution reconstruction method provided by an embodiment of the present invention specifically includes steps S101 to S103.

[0029] S101, obtaining a video sequence to be super-resolution reconstructed;

[0030] S102, performing noise reduction processing on the video sequence using a lightweight residual noise reduction network;

[0031] S103: Input the denoised video sequence into a video super-resolution network, and the video super-resolution network outputs a super-resolution reconstructed feature map corresponding to the video sequence.

[0032] In this embodiment, the acquired video sequence is first denoised using a lightweight residual denoising network. The frames in the denoised video sequence are then super-reconstructed using a video super-resolution network to obtain corresponding feature maps. The video super-resolution reconstruction method based on a lightweight denoising network in this embodiment introduces a residual denoising network to denoise the video sequence. A video super-resolution network with bidirectional propagation and optical flow alignment is then constructed to reconstruct the denoised video sequence, thereby improving the video super-resolution reconstruction effect. The lightweight residual denoising network is also used to accelerate the reconstruction process.

[0033] In one embodiment, step S102 includes:

[0034] The video sequence is denoised using the first denoising module, the second denoising module and the third denoising module in the lightweight residual denoising network in sequence; wherein the first denoising module includes a first convolution and a LeakyReLu activation function layer, the second denoising module includes multiple consecutive denoising layers, and each denoising layer includes a second convolution, a ReLu activation function, and a second convolution in sequence, and the third denoising module includes a third convolution layer; the first and third convolutions are both 3*3*c*64 convolutions, the second convolution is 3*3*64*64 convolution, and c is the number of image channels;

[0035] Get the frames in the video sequence after noise reduction according to the following formula

[0036]

[0037] Where, D(I t ) represents the residual output of the lightweight residual denoising network D, I t represents the tth frame in the video sequence.

[0038] Since the residual between the noise image and the original image is very small, and according to the theory in ResNet, when the residual is 0, the mapping between the networks is equivalent to the identity mapping, it is a network structure that is very easy to train. Real noise is a very small part, that is, the residual between the noisy image and the original image is very small, so residual learning is very suitable for image restoration. Therefore, in order to design a smooth noise reduction network that is more in line with real scenes, this embodiment applies residual learning to video noise reduction. That is to say, the noise reduction process is performed by the lightweight residual noise reduction network, and the overall network design is an end-to-end network output. The difference is that the input of this embodiment is the current frame, and the output is the residual map. The noisy image undergoes a series of convolution processes, and finally generates a residual map containing only noise.

[0039] Since it is for video denoising, in order to achieve real-time effect, the network depth can be set to 20 to meet performance requirements. In the denoising process, the convolution kernel size is set to 3*3 and the pooling layer is removed. The specific network structure of the lightweight residual denoising network is as follows: Figure 5 shown.

[0040] Figure 5 The network structure shown includes three parts, namely the first noise reduction module, the second noise reduction module and the third noise reduction module, wherein:

[0041] The first denoising module: Conv(3*3*c*64)+LeakyReLu (c represents the number of image channels), that is, the first convolution and LeakyReLu activation function layer;

[0042] The second denoising module: Conv(3*3*64*64)+ReLu+Conv(3*3*64*64), i.e. the second convolution, ReLu activation function, and the second convolution;

[0043] The third denoising module: Conv(3*3*c*64), which is the third convolutional layer.

[0044] The tth frame in the video sequence is counted as D(I t ), and use D to represent the process of lightweight residual denoising network, then the output of lightweight residual denoising network D is residual D(I t ), denoised frame:

[0045]

[0046] The optimization objective is the mean squared error (MSE) between the residual image and the network output. This network design eliminates the real image from the original noisy image in the hidden layer. To make the strength of the noise reduction controllable, the noise reduction can be repeated.

[0047] In one embodiment, if Figure 2 As shown, the step S103 includes steps S201 to S204.

[0048] S201, extracting spatial propagation features from frames in a video sequence using a bidirectional propagation method;

[0049] In this step, considering that one-way propagation obtains less information, a two-way propagation method is used to obtain more information of the frames in the video sequence, thereby improving the final video output quality.

[0050] In a specific embodiment, step S201 includes:

[0051] The forward propagation features of the frames in the video sequence are extracted according to the following formula and backpropagation features

[0052]

[0053] Where, F b and F f represent the forward propagation and backward propagation functions respectively, represents the current frame in the video sequence, express The previous frame, express The next frame, express The forward propagation characteristics of Indicates the next frame Here, the input of the bidirectional propagation is the lightweight residual denoising output.

[0054] S202, performing alignment processing on the spatial propagation features based on optical flow alignment;

[0055] Unlike image super-resolution reconstruction, this step requires image alignment for video. Misalignment will hinder convergence, resulting in poor results. Therefore, similar to flow-based methods, this embodiment employs optical flow alignment to align the extracted spatial propagation features in feature space.

[0056] In a specific embodiment, step S202 includes:

[0057] The spatial propagation features are aligned according to the following formula to obtain the alignment variables after alignment:

[0058]

[0059]

[0060] In the formula, G represents the optical flow estimation module, Y represents the spatial mapping module, represents the optical flow estimation variable, Indicates the current frame The previous and / or next frame of Represents intermediate variables The previous frame variable and / or next frame variable.

[0061] S203, inputting the aligned spatial propagation features into a residual module, and outputting intermediate variables of frames in the video sequence from the residual;

[0062] In this step, the aligned spatial propagation features are further adjusted through the residual module.

[0063] In a specific embodiment, step S203 includes:

[0064] The intermediate variables of the frames in the video sequence are calculated according to the following formula:

[0065]

[0066] Where R {b,f} Represents the residual module.

[0067] S204 , performing up-sampling processing on the intermediate variable based on pixel reorganization to obtain the super-resolution reconstruction feature map.

[0068] In this step, the intermediate variables It is sent to the upsampling module U, which generates a high-resolution current frame H by pixel reorganization. t .

[0069] In a specific embodiment, step S204 includes:

[0070] The intermediate variables are upsampled according to the following formula to obtain the current frame H of the super-resolution reconstruction feature map: t :

[0071]

[0072] Where U represents the upsampling module, Represents the current frame H of the super-resolution reconstruction feature map t The forward propagation characteristics of Represents the current frame H of the super-resolution reconstruction feature map t Backward propagation features.

[0073] In this embodiment, considering the importance of distant frame information and the limited information obtained through one-way propagation, bidirectional propagation is used to fully utilize the information in the video sequence. The spatial features of optical flow are then used for alignment. The aligned features are then passed through a residual module to output intermediate variables. Finally, based on pixel reconstruction, the low-resolution feature map is upsampled to obtain a high-resolution feature map.

[0074] In a specific application scenario, to better simulate the distortion caused by real-world scenes, this embodiment collects video data from commonly used video websites and then compresses it to construct a training dataset. Furthermore, operations such as blurring, noise, resizing, and JPEG compression are randomly added to the training dataset to produce low-resolution, noisy videos. This training dataset is then input into the video super-resolution reconstruction method provided by this embodiment of the present invention for training and learning.

[0075] The final experimental results are compared with Figure 6 As shown in Figure 2, we selected the first, 11th, and 21st frames from a set of videos as references. Figure 6 The first row shows the original video, the second row shows the result without noise reduction, and the third row shows the result of super-resolution with noise reduction. These are the experimental results of the video super-resolution reconstruction method provided by the embodiments of the present invention. As can be seen from the comparison figures, the video super-resolution reconstruction method provided by the embodiments of the present invention produces a very clear and natural result, superior to the result without noise reduction.

[0076] Furthermore, the video super-resolution reconstruction method provided by the present invention is trained on a large-scale dataset without distinguishing between different scenes. However, for different scenes, different datasets can be collected to train the video super-resolution reconstruction method provided by the present invention, thereby improving the reconstruction effect and adapting to the needs of various processing scenarios.

[0077] Figure 3 A schematic block diagram of a video super-resolution reconstruction device 300 provided in an embodiment of the present invention, the device 300 includes:

[0078] The video sequence acquisition unit 301 is used to acquire a video sequence to be super-resolution reconstructed;

[0079] a video sequence denoising unit 302, configured to perform denoising on the video sequence using a lightweight residual denoising network;

[0080] The super-resolution reconstruction unit 303 is configured to input the denoised video sequence into a video super-resolution network, and the video super-resolution network outputs a super-resolution reconstruction feature map corresponding to the video sequence.

[0081] In one embodiment, the video sequence noise reduction unit 302 includes:

[0082] A residual denoising unit, configured to sequentially perform denoising on the video sequence using the first denoising module, the second denoising module, and the third denoising module in the lightweight residual denoising network; wherein the first denoising module includes a first convolution and a LeakyReLu activation function layer, the second denoising module includes multiple consecutive denoising layers, and each denoising layer sequentially includes a second convolution, a ReLu activation function, and a second convolution; the third denoising module includes a third convolution layer; the first and third convolutions are both 3*3*c*64 convolutions, the second convolution is a 3*3*64*64 convolution, and c is the number of image channels;

[0083] The frame acquisition unit is used to obtain the frames in the video sequence after noise reduction according to the following formula

[0084]

[0085] Where, D(I t ) represents the residual output of the lightweight residual denoising network D, I t represents the tth frame in the video sequence.

[0086] In one embodiment, if Figure 4 As shown, the super-resolution reconstruction unit 303 includes:

[0087] A bidirectional propagation unit 401 is configured to extract spatial propagation features from frames in a video sequence using a bidirectional propagation method;

[0088] an optical flow alignment unit 402, configured to perform alignment processing on the spatial propagation features based on optical flow alignment;

[0089] a variable output unit 403, configured to input the aligned spatial propagation features into a residual module, and output intermediate variables of frames in the video sequence from the residual;

[0090] The pixel reassembly unit 404 is configured to perform upsampling processing on the intermediate variable based on pixel reassembly to obtain the super-resolution reconstructed feature map.

[0091] In one embodiment, the bidirectional communication unit 401 includes:

[0092] The forward and backward extraction unit is used to extract the forward propagation features of the frames in the video sequence according to the following formula: and backpropagation features

[0093]

[0094]

[0095] Where, F b and F f represent the forward propagation and backward propagation functions respectively, represents the current frame in the video sequence, express The previous frame, express The next frame, express The forward propagation characteristics of Indicates the next frame Backward propagation features.

[0096] In one embodiment, the optical flow alignment unit 402 includes:

[0097] An alignment processing unit is used to align the spatial propagation features according to the following formula to obtain an alignment variable after alignment processing:

[0098]

[0099]

[0100] In the formula, G represents the optical flow estimation module, Y represents the spatial mapping module, represents the optical flow estimation variable, Indicates the current frame The previous and / or next frame of Represents intermediate variables The previous frame variable and / or next frame variable.

[0101] In one embodiment, the variable output unit 403 includes:

[0102] A variable calculation unit is used to calculate the intermediate variables of the frames in the video sequence according to the following formula:

[0103]

[0104] Where R {b,f} Represents the residual module.

[0105] In one embodiment, the pixel recombining unit 404 includes:

[0106] The sampling processing unit is used to perform up-sampling processing on the intermediate variable according to the following formula to obtain the current frame H of the super-resolution reconstruction feature map t :

[0107]

[0108] Where U represents the upsampling module, Represents the current frame H of the super-resolution reconstruction feature map t The forward propagation characteristics of Represents the current frame H of the super-resolution reconstruction feature map t Backward propagation features.

[0109] Since the embodiments of the apparatus part correspond to the embodiments of the method part, please refer to the description of the embodiments of the method part for the embodiments of the apparatus part, and they will not be repeated here.

[0110] The present invention also provides a computer-readable storage medium having a computer program stored thereon. When executed, the computer program can implement the steps provided in the above embodiments. The storage medium can include a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, among other media capable of storing program code.

[0111] The present invention also provides a computer device that may include a memory and a processor. The memory stores a computer program, and when the processor calls the computer program in the memory, the steps provided in the above embodiment can be implemented. Of course, the computer device may also include various network interfaces, a power supply, and other components.

[0112] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the various embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of this application, several improvements and modifications can be made to this application, and these improvements and modifications also fall within the scope of protection of the claims of this application.

[0113] It should also be noted that, in this specification, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus comprising the element.

Claims

1. A video super-resolution reconstruction method, characterized in that: include: Obtaining a video sequence to be super-resolution reconstructed; Using a lightweight residual denoising network to perform denoising on the video sequence; Inputting the denoised video sequence into a video super-resolution network, and having the video super-resolution network output a super-resolution reconstructed feature map corresponding to the video sequence; The adopting a lightweight residual denoising network to perform denoising on the video sequence includes: The video sequence is denoised using the first denoising module, the second denoising module and the third denoising module in the lightweight residual denoising network in sequence; wherein the first denoising module includes a first convolution and a LeakyReLu activation function layer, the second denoising module includes multiple consecutive denoising layers, and each denoising layer includes a second convolution, a ReLu activation function, and a second convolution in sequence, and the third denoising module includes a third convolution layer; the first and third convolutions are both 3*3*c*64 convolutions, the second convolution is 3*3*64*64 convolution, and c is the number of image channels; Get the frames in the video sequence after noise reduction according to the following formula Where, D(I t ) represents the residual output of the lightweight residual denoising network D, I t represents the tth frame in the video sequence; The step of inputting the denoised video sequence into a video super-resolution network, and having the video super-resolution network output a super-resolution reconstruction feature map corresponding to the video sequence, includes: A two-way propagation method is used to extract spatial propagation features from frames in video sequences; Performing alignment processing on the spatial propagation features based on optical flow alignment; Inputting the aligned spatial propagation features into a residual module, and outputting intermediate variables of frames in the video sequence from the residual; The intermediate variables are up-sampled based on pixel reorganization to obtain the super-resolution reconstruction feature map.

2. The video super-resolution reconstruction method according to claim 1, characterized in that The method of extracting spatial propagation features from frames in a video sequence using a bidirectional propagation method includes: The forward propagation features of the frames in the video sequence are extracted according to the following formula and backpropagation features Where, F f and F b represent the forward propagation and backward propagation functions respectively, represents the current frame in the video sequence, express The previous frame, express The next frame, express The forward propagation characteristics of Indicates the next frame Backward propagation features.

3. The video super-resolution reconstruction method according to claim 2, characterized in that The aligning processing of the spatial propagation features based on optical flow alignment includes: The spatial propagation features are aligned according to the following formula to obtain the alignment variables after alignment: In the formula, G represents the optical flow estimation module, Y represents the spatial mapping module, represents the optical flow estimation variable, Indicates the current frame The previous and / or next frame of Represents intermediate variables The previous frame variable and / or next frame variable.

4. The video super-resolution reconstruction method according to claim 3, characterized in that Inputting the aligned spatial propagation features into a residual module, and outputting intermediate variables of frames in the video sequence from the residual, includes: The intermediate variables of the frames in the video sequence are calculated according to the following formula: Where R {b,f} Represents the residual module.

5. The video super-resolution reconstruction method according to claim 4, characterized in that: The upsampling process is performed on the intermediate variable based on pixel reorganization to obtain the super-resolution reconstruction feature map, including: The intermediate variables are upsampled according to the following formula to obtain the current frame H of the super-resolution reconstruction feature map: t : Where U represents the upsampling module, Represents the current frame H of the super-resolution reconstruction feature map t The forward propagation characteristics of Represents the current frame H of the super-resolution reconstruction feature map t Backward propagation features.

6. A video super-resolution reconstruction device, characterized in that: include: A video sequence acquisition unit, used to acquire a video sequence to be super-resolution reconstructed; a video sequence denoising unit, configured to perform denoising on the video sequence using a lightweight residual denoising network; A super-resolution reconstruction unit, configured to input the denoised video sequence into a video super-resolution network, and have the video super-resolution network output a super-resolution reconstruction feature map corresponding to the video sequence; The video sequence noise reduction unit comprises: A residual denoising unit, configured to sequentially perform denoising on the video sequence using the first denoising module, the second denoising module, and the third denoising module in the lightweight residual denoising network; wherein the first denoising module includes a first convolution and a LeakyReLu activation function layer, the second denoising module includes multiple consecutive denoising layers, and each denoising layer sequentially includes a second convolution, a ReLu activation function, and a second convolution; the third denoising module includes a third convolution layer; the first and third convolutions are both 3*3*c*64 convolutions, the second convolution is a 3*3*64*64 convolution, and c is the number of image channels; The frame acquisition unit is used to obtain the frames in the video sequence after noise reduction according to the following formula Where, D(I t ) represents the residual output of the lightweight residual denoising network D, I t represents the tth frame in the video sequence; The super-resolution reconstruction unit comprises: A bidirectional propagation unit, configured to extract spatial propagation features from frames in a video sequence using a bidirectional propagation method; an optical flow alignment unit, configured to perform alignment processing on the spatial propagation features based on optical flow alignment; a variable output unit, configured to input the aligned spatial propagation features into a residual module, and output intermediate variables of frames in the video sequence from the residual; A pixel reorganization unit is used to perform upsampling processing on the intermediate variable based on pixel reorganization to obtain the super-resolution reconstruction feature map.

7. A computer device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the video super-resolution reconstruction method according to any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the video super-resolution reconstruction method according to any one of claims 1 to 5 is implemented.

Images