Image processing method and device, equipment, medium and product

Abstract

The invention relates to the technical field of computers, and particularly provides an image processing method and device, equipment, a medium and a product. The method comprises the steps of obtaining a to-be-processed image; processing the to-be-processed image by using a pre-trained first model to obtain image features output by the first model; wherein the first model is used for extracting quantum features corresponding to the input features from a plurality of different scales, performing residual fusion on the input features and the quantum features, and outputting the input features and the quantum features. By extracting the quantum features from multiple scales and outputting the quantum features after residual fusion, more complex feature information can be extracted, the number of network layers is reduced, and the accuracy of the first model is ensured under the condition that large-scale parameters are not introduced. Therefore, the accuracy of image processing can be improved.

Description

Technical Field

[0001] The present disclosure belongs to the field of computer technology, and specifically relates to an image processing method, apparatus, device, medium, and product. Background Art

[0002] Image processing technology uses computers to process, analyze, and understand images to handle various patterns of objects and targets. It can be used in applications such as image classification and target tracking. Related technologies often use classic neural network models for image processing. However, these image processing methods have relatively simple feature extraction structures, resulting in insufficient feature extraction capabilities and the potential loss of important features during feature extraction. Consequently, this reduces image processing accuracy. Summary of the Invention

[0003] In view of the above problems, the present disclosure is proposed. The present disclosure provides an image processing method, apparatus, device, medium and product, which can improve the accuracy of image processing.

[0004] According to one aspect of the present disclosure, there is provided an image processing method, comprising: Get the image to be processed; Processing the image to be processed using a pre-trained first model to obtain image features output by the first model; The first model is used to extract quantum features corresponding to input features from multiple different scales, and perform residual fusion of the input features and the quantum features and output them.

[0005] Optionally, the first model includes: A plurality of quantum convolution kernel circuits of different scales, any one of which is used to extract features based on quantum computing and output classical features; Channel dimension, used to combine the classical features of multiple different scales; The residual connection layer is used to perform residual fusion on the input features and the concatenated classic features and output the result.

[0006] Optionally, the quantum convolution kernel circuit includes: an encoding unit, an entanglement unit and a measurement unit; Wherein, the encoding unit is used to perform quantum encoding on the input feature; The entanglement unit is used to extract features of the quantum code and output the quantum features; The measuring unit is used to extract the quantum feature and output the classical feature.

[0007] Optionally, the quantum convolution kernel circuit includes at least one of the following quantum gates: a Hadamard gate, a first rotation gate, a second rotation gate, and a third rotation gate, wherein the Hadamard gate is used to convert input features into superposition state data, and the rotation gate is used to convert angle coding.

[0008] Optionally, performing quantum encoding on the input feature includes: At the target scale, the quantum convolution kernel is used as a sliding window to determine the target feature map from the input features; Performing quantum coding on the pixel points of the target feature map based on the Hadamard gate to obtain a first quantum code for each pixel point, where the first quantum code is superposition state data; The first quantum code is angle-encoded based on at least one of the rotating gates to obtain a second quantum code for each of the pixel points.

[0009] Optionally, the method further includes: Using the second model, semi-supervised training is performed on the first model; The second model is a teacher model of the first model; the second model is used to generate a feature representation of the input feature.

[0010] Optionally, the first model is an independent feature extraction model, or the first model is a sub-model of the third model; The third model is used to process image tasks; wherein the image tasks include one or more of the following: image classification, image segmentation, image recognition, target detection, and feature extraction.

[0011] Optionally, obtaining the image to be processed includes: Get the original image; Segmenting the original image to obtain a plurality of image sub-blocks; Randomly masking the plurality of image sub-blocks to obtain uncovered image sub-blocks; The uncovered image sub-block is used as an image to be processed.

[0012] According to another aspect of the present disclosure, there is provided an image processing apparatus, comprising: An acquisition module, used for acquiring an image to be processed; A processing module is used to process the image to be processed using a pre-trained first model to obtain image features output by the first model; wherein the first model is used to extract quantum features corresponding to input features from multiple different scales, and to perform residual fusion of the input features and the quantum features and output them.

[0013] According to another aspect of the present disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned image processing method.

[0014] According to another aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. The program is executed by a processor to implement the above-mentioned image processing method.

[0015] According to another aspect of the present disclosure, a computer program product is provided, comprising a computer-readable code, or a non-volatile computer-readable storage medium carrying the computer-readable code. When the computer-readable code runs in a processor of an electronic device, the processor in the electronic device executes and implements the above-mentioned image processing method.

[0016] In the present disclosure, an image to be processed is obtained; the image to be processed is processed using a pre-trained first model to obtain image features output by the first model; the first model is used to extract quantum features corresponding to the input features at multiple scales, and the input features are fused with the quantum features for output. By extracting quantum features from multiple scales and outputting them after residual fusion, more complex feature information can be extracted, the number of network layers used can be reduced, and the accuracy of the first model can be guaranteed without introducing large-scale parameters. Therefore, the accuracy of image processing can be improved.

[0017] It is to be understood that both the foregoing general description and the following detailed description are exemplary, and are intended to provide further explanation of the technology as claimed. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The above and other purposes, features, and advantages of the present disclosure will become more apparent through a more detailed description of the embodiments of the present disclosure in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the present disclosure and are not intended to limit the present disclosure. In the drawings, the same reference numerals generally represent the same components or steps.

[0019] Figure 1 is a schematic diagram illustrating a communication system to which a communication network coverage assessment method according to an embodiment of the present disclosure is applied.

[0020] Figure 2 is a flowchart illustrating a communication network coverage assessment method according to an embodiment of the present disclosure.

[0021] Figure 3 4 is a flowchart further illustrating a process of determining sampling point information in a communication network coverage assessment method according to an embodiment of the present disclosure.

[0022] Figure 4 It is a flowchart further illustrating the comprehensive coverage status determination process in the communication network coverage assessment method according to an embodiment of the present disclosure.

[0023] Figure 5 A hardware block diagram of an electronic device provided in the present disclosure.

[0024] Figure 6 A schematic diagram of a computer program product provided by the present disclosure. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to more clearly understand the technical solution of the present application, the application scenario of the solution of the present application is first described below.

[0026] Image processing technology refers to the technology of using computers to process, analyze and understand images in order to handle targets and objects of various different patterns. It can be used in application scenarios such as image classification and target tracking and detection.

[0027] In related technologies, image processing often uses classic neural network models. However, these image processing methods have relatively simple feature extraction structures, resulting in poor robustness and generalization when processing complex samples. Furthermore, quantum gate circuits use a small number of qubits and a relatively simple circuit structure, resulting in insufficient image feature extraction capabilities. This can lead to the loss of important features during feature extraction, making it particularly difficult to capture deep semantic information in some images. Consequently, this reduces image processing accuracy.

[0028] To address the aforementioned technical issues, this disclosure provides an inventive concept for image processing: extracting quantum features from multiple scales and outputting them after residual fusion. This allows for the extraction of more complex feature information, reduces the number of network layers, and maintains the accuracy of the first model without introducing large-scale parameters. Consequently, image processing accuracy can be improved.

[0029] In order to make the purpose, technical solutions and advantages of the present disclosure more apparent, the following will describe in detail exemplary embodiments of the present disclosure with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all the embodiments of the present disclosure, and it should be understood that the present disclosure is not limited to the exemplary embodiments described herein.

[0030] Figure 1 A flow chart of an image processing method provided by the present disclosure. Figure 1 As shown, the method includes: S101: Acquire an image to be processed.

[0031] Specifically, in this embodiment, the original image can be divided into k sub-blocks, which are randomly masked and then screened, with the uncovered sub-blocks used as the processed images. This increases data diversity and helps the model learn more robust features, rather than relying on information from specific regions. Furthermore, masking some regions may force the model to use contextual information to infer the masked content, thereby improving the model's generalization ability.

[0032] S102: Process the image to be processed using the pre-trained first model to obtain image features output by the first model.

[0033] Specifically, the first model is used to extract quantum features corresponding to input features at multiple scales, perform residual fusion of the input features and the quantum features, and output them. The first model can be trained using samples of the image to be processed, so that the first model learns how to extract quantum features and outputs them after residual fusion. After the first model is trained, the input features of the image to be processed can be used to extract the corresponding quantum features.

[0034] For example, the first model includes: quantum convolution kernel circuits of multiple different scales, channel dimensions and residual connection layers.

[0035] Specifically, multiple quantum convolution kernel circuits of different scales are used, and any one of them is used to extract features based on quantum computing and output classical features. The quantum convolution kernel is usually a small matrix (such as 3x3 or 5x5) that is used to extract quantum features from the input features. This is achieved by sliding the quantum convolution kernel on the input data and calculating the sum of the element-wise products of the kernel and the corresponding part of the input features at each position to generate a new feature map. Each quantum convolution kernel only focuses on a small part of the input data. The quantum convolution kernel reuses the same weights on the entire input. The features extracted in this way are insensitive to small-scale translations of the input. The quantum convolution kernel circuit is the tool used by the quantum convolution kernel to implement quantum computing. By constructing the quantum convolution kernel circuit, quantum features are calculated and ultimately classical features in the form of bit values ​​are output, and subsequent operations are performed using the classical features.

[0036] You can The quantum convolution kernel of 1 is split into separable quantum convolution kernels in the x and y directions, using 2 quantum bits respectively, and the convolution kernel size is 1 2 and 2 1. This can significantly reduce the amount of computation. By using quantum convolution kernel circuits of multiple different scales, the corresponding classical features at the target scale are extracted.

[0037] Channel dimension, used to splice multiple classic features of different scales. Feature fusion of multiple classic features of different scales, splicing the classic features in the channel dimension. Residual connection layer, used to perform residual fusion of input features and spliced ​​classic features and output them. After 1 The convolution kernel of 1 keeps the number of channels after fusion and the number of input feature channels consistent, so that the residual network can be used to perform residual connection between the input features and the concatenated classic features.

[0038] After multiple rounds of training, a first model for image feature extraction based on the image samples to be processed can be obtained. This model can be used in visual tasks in multiple scenarios such as image classification, image segmentation, and object detection.

[0039] In the present disclosure, an image to be processed is obtained; the image to be processed is processed using a pre-trained first model to obtain image features output by the first model; the first model is used to extract quantum features corresponding to the input features at multiple scales, and the input features are fused with the quantum features for output. By extracting quantum features from multiple scales and outputting them after residual fusion, more complex feature information can be extracted, the number of network layers used can be reduced, and the accuracy of the first model can be guaranteed without introducing large-scale parameters. Therefore, the accuracy of image processing can be improved.

[0040] In one possible implementation, the quantum convolution kernel circuit includes at least one of the following quantum gates: a Hadamard gate, a first rotation gate, a second rotation gate, and a third rotation gate.

[0041] Figure 2 This is a schematic diagram of the structure of the quantum convolution kernel circuit provided by the present disclosure, such as Figure 2 As shown in Figure 2, the Hadamard gate (H) is used to convert input features into superposition data. This involves quantum encoding pixels and converting them into superposition states to achieve quantum parallelism in quantum computing. By placing one or more quantum bits in a superposition state, a quantum algorithm can perform calculations on multiple quantum states simultaneously, achieving higher computational efficiency than classical computing.

[0042] An example formula for the H-gate is shown below:

[0043] The first, second, and third revolving gates are used to convert angle encoding, that is, to angle encoding the encoding of the superposition state, thus achieving quantum entanglement. The first revolving gate can be RX angle encoding, rotating around the x-axis to change the amplitude of the quantum state; the second revolving gate can be RY angle encoding, rotating around the y-axis to change the amplitude; and the third revolving gate can be RZ angle encoding, rotating around the z-axis to change only the phase.

[0044] An example formula for RX angle encoding is shown below:

[0045] in, is the rotation angle, Encodes the RX angle.

[0046] An example formula for RY angle encoding is shown below:

[0047] in, is the rotation angle, Encodes the RY angle.

[0048] An example formula for RZ angle encoding is shown below:

[0049] in, is the rotation angle, It is the RZ angle code.

[0050] In addition, during the training process, the rotation angle of the previous step can be used as a training parameter of the quantum convolution kernel circuit. This embodiment uses Bloch sphere angle encoding, and other angle encoding methods or amplitude encoding and other encoding methods can also be used, which are not specifically limited in this disclosure.

[0051] In one possible implementation, a quantum convolution kernel circuit includes: an encoding unit, an entanglement unit, and a measurement unit; wherein the encoding unit is used to perform quantum encoding on input features; the entanglement unit is used to perform feature extraction on the quantum encoding and output quantum features; and the measurement unit is used to perform feature extraction on the quantum features and output classical features.

[0052] Example: At the target scale, the quantum convolution kernel is used as a sliding window to determine the target feature map from the input features. The pixel points of the target feature map are quantum-encoded based on the Hadamard gate to obtain the first quantum code of each pixel point, which is superposition state data. The first quantum code is angle-encoded based on at least one rotation gate to obtain the second quantum code of each pixel point.

[0053] Specifically, the quantum convolution kernel is used as a sliding window to select the target feature map that requires quantum computation from the input features. The pixels of the target feature map are quantum-encoded using an H-gate to obtain a first quantum code. This first quantum code is then regularized and projected onto the (0, π) interval. After quantum computation using at least one rotation gate, the final second quantum code is obtained.

[0054] The entanglement unit then extracts features from the quantum code, linking the states of the two qubits to form an entangled state and outputting quantum features. The measurement unit converts the quantum features into classical features, such as probability values, which are then used as the final output features of the sliding window.

[0055] In one possible implementation, the method further includes: The second model is used to perform semi-supervised training on the first model.

[0056] Specifically, the second model serves as the teacher model for the first model; the second model is used to generate feature representations of the input features. In this embodiment, the first model can be used as the student model in a teacher-student model. After the second model generates feature representations of the input features, feature transfer is performed on the second model using imitation, achieving semi-supervised training of the first model by the second model.

[0057] In addition, when training the second model, the feature extraction part can also use the image feature extraction method in the first model.

[0058] In a possible implementation, the first model is an independent feature extraction model, or the first model is a sub-model of the third model.

[0059] Specifically, the first model can be an independent model for image feature extraction, which can be used alone or in combination with other independent models. Alternatively, the first model can be a submodel of a third model. For example, the third model can be a teacher-student model, and the third model is used to process image tasks; wherein image tasks include one or more of the following: image classification, image segmentation, image recognition, object detection, and feature extraction. It is understood that feature extraction here refers to a model used for feature extraction within a larger model.

[0060] In a possible implementation, obtaining an image to be processed includes: An original image is obtained; the original image is segmented to obtain a plurality of image sub-blocks; the plurality of image sub-blocks are randomly masked to obtain uncovered image sub-blocks; and the uncovered image sub-blocks are used as images to be processed.

[0061] Specifically, the original image is divided into k image sub-blocks, for example, k=16, satisfying k=N N is enough, and then a certain proportion of image sub-blocks are randomly masked. Gray pixels can be filled instead of masking. After linear projection, their position codes are added, and the uncovered image sub-blocks are selected as the image to be processed.

[0062] Figure 3 Another flow chart of an image processing method provided by the present disclosure. Figure 3 As shown, the method includes: S301: Preprocess the image data set to obtain an image to be processed.

[0063] Specifically, the original images in the image dataset are segmented and masked, and the randomly obtained unmasked image data are subjected to data enhancement and normalization operations to obtain the image to be processed.

[0064] S302: Input the image to be processed into the teacher-student model based on the multi-scale residual quantum convolution kernel to perform shallow and deep multi-stage knowledge distillation.

[0065] Specifically, the image to be processed is fed into the teacher model. Through learning in the shallow network layers, a shallow feature representation is obtained. This feature is then transferred to the student model using imitation. The teacher model, trained on the shallow layers, is then fed into the deep layers for training. 80% of the sub-blocks of the original image are masked. The student model then generates a complete feature representation of the deep teacher model through a generation module. After distilling knowledge from the shallow and deep layers, the student model decodes the feature representation.

[0066] S303: Define a quantum convolution kernel circuit to perform quantum state encoding and gating operations on the image data in the sliding window.

[0067] Specifically, during the semi-supervised training of the teacher-student model, feature extraction can use the defined quantum convolution kernel circuit to perform quantum state encoding and gating operations on the image data in the sliding window to obtain the eigenvalues ​​after quantum calculation.

[0068] S304: Perform residual connection and multi-scale fusion on the features of quantum convolution, and calculate the knowledge distillation loss and supervised regularization loss.

[0069] Specifically, the eigenvalues ​​after multi-scale quantum convolution are fused, and the fused features are used to perform residual connection with the input features to calculate the knowledge distillation loss and supervised regularization loss.

[0070] S305: Back propagation updates parameters and outputs the classification results of the image dataset.

[0071] Specifically, the knowledge distillation loss and supervised regularization loss are used to update the model parameters of the teacher-student model and optimize the model to output accurate image dataset classification results.

[0072] This allows the use of a small amount of labeled data, reducing the cost of manual labeling and data cleaning, and extracting richer feature information while accelerating calculations through quantum convolution kernels. This is particularly effective for semi-supervised tasks that require deep mining of the data's own semantic information, and can be applied to a variety of image tasks.

[0073] Figure 4 This is a schematic diagram of the structure of an image processing device provided by the present disclosure. Figure 4 As shown, the device 400 includes: an acquisition module 410 and a processing module 420.

[0074] An acquisition module 410 is used to acquire an image to be processed; A processing module 420 is used to process the image to be processed using a pre-trained first model to obtain image features output by the first model; wherein the first model is used to extract quantum features corresponding to input features from multiple different scales, and to perform residual fusion of the input features with the quantum features and output them.

[0075] Optionally, the device comprises: At the target scale, the quantum convolution kernel is used as a sliding window to determine the target feature map from the input features; Performing quantum coding on the pixel points of the target feature map based on the Hadamard gate to obtain a first quantum code for each pixel point, where the first quantum code is superposition state data; The first quantum code is angle-encoded based on at least one of the rotating gates to obtain a second quantum code for each of the pixel points.

[0076] Optionally, the device further comprises: Using the second model, semi-supervised training is performed on the first model; The second model is a teacher model of the first model; the second model is used to generate a feature representation of the input feature.

[0077] Optionally, the acquisition module is used to: Get the original image; Segmenting the original image to obtain a plurality of image sub-blocks; Randomly masking the plurality of image sub-blocks to obtain uncovered image sub-blocks; The uncovered image sub-block is used as an image to be processed.

[0078] The present application also provides an electronic device to perform the above image processing method. Figure 5 It shows a schematic diagram of an electronic device provided by some embodiments of the present application. Figure 5As shown, the electronic device 5 includes: a processor 500, a memory 501, a bus 502 and a communication interface 503, and the processor 500, the communication interface 503 and the memory 501 are connected via the bus 502; the memory 501 stores a computer program that can be run on the processor 500, and when the processor 500 runs the computer program, it executes the image processing method provided in any of the aforementioned embodiments of the present application.

[0079] Memory 501 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between the device network element and at least one other network element is achieved through at least one communication interface 503 (which may be wired or wireless), and may utilize the Internet, a wide area network, a local area network, a metropolitan area network, or the like.

[0080] The bus 502 may be an ISA bus, a PCI bus, or an EISA bus. The bus may be divided into an address bus, a data bus, a control bus, etc. The memory 501 is used to store programs. The processor 500 executes the programs upon receiving execution instructions. The image processing method disclosed in any of the aforementioned embodiments of the present application may be applied to the processor 500 or implemented by the processor 500.

[0081] The processor 500 may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by hardware integrated logic circuits in the processor 500 or by software instructions. The above processor 500 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in conjunction with the embodiments of this application can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium mature in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory 501 , and the processor 500 reads the information in the memory 501 and completes the steps of the above method in combination with its hardware.

[0082] The electronic device provided in the embodiment of the present application and the image processing method provided in the embodiment of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, operated or implemented by them.

[0083] An embodiment of the present application also provides a computer-readable storage medium corresponding to the image processing method provided in the aforementioned embodiment. The computer-readable storage medium shown therein may be a CD having a computer program stored thereon. When the computer program is run by a processor, it will execute the image processing method provided in any of the aforementioned embodiments.

[0084] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical or magnetic storage media, which are not listed here one by one.

[0085] The computer-readable storage medium provided in the above-mentioned embodiments of the present application and the image processing method provided in the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0086] The present application also provides a computer program product 600. Figure 6 The computer program product carries a computer program 601, and the instructions included in the program code can be used to execute the steps of the image processing method described in the above method embodiment. For details, please refer to the above method embodiment, which will not be repeated here.

[0087] The computer program product may be implemented in hardware, software, or a combination thereof. In one embodiment, the computer program product is implemented as a computer storage medium. In another embodiment, the computer program product is implemented as a software product, such as a software development kit (SDK).

[0088] The basic principles of the present disclosure have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, and effects mentioned in this disclosure are merely illustrative and not restrictive, and should not be construed as necessarily possessed by each embodiment of the present disclosure. Furthermore, the specific details disclosed above are provided for illustrative purposes and to facilitate understanding, rather than as limitations. These details do not limit the present disclosure to necessarily being implemented using these specific details.

[0089] The block diagrams of the devices, devices, equipment, and systems involved in this disclosure are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As will be appreciated by those skilled in the art, these devices, devices, equipment, and systems can be connected, arranged, or configured in any manner. Words such as "include," "comprise," "have," and the like are open-ended words, meaning "including but not limited to," and can be used interchangeably therewith. The words "or" and "and" used herein refer to the words "and / or" and can be used interchangeably therewith, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to," and can be used interchangeably therewith.

[0090] Additionally, as used herein, "or" used in a list of items beginning with "at least one" indicates a separate list, so that, for example, a list of "at least one of A, B, or C" means A or B or C, or AB or AC or BC, or ABC (i.e., A and B and C). Furthermore, the word "exemplary" does not mean that the example described is preferred or better than other examples.

[0091] It should also be noted that in the system and method of the present disclosure, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present disclosure.

[0092] Various changes, substitutions, and modifications may be made to the technology described herein without departing from the teachings defined by the appended claims. Moreover, the scope of the claims of this disclosure is not limited to the specific aspects of the processes, machines, manufactures, compositions of things, means, methods, and actions described above. Currently existing or later developed processes, machines, manufactures, compositions of things, means, methods, or actions that perform substantially the same function or achieve substantially the same results as the corresponding aspects described herein may be utilized. Accordingly, the appended claims include within their scope such processes, machines, manufactures, compositions of things, means, methods, or actions.

[0093] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present disclosure. Therefore, the present disclosure is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0094] The above description has been provided for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present disclosure to the forms disclosed herein. Although a number of example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. An image processing method, characterized in that: include: Get the image to be processed; Processing the image to be processed using a pre-trained first model to obtain image features output by the first model; The first model is used to extract quantum features corresponding to input features from multiple different scales, and perform residual fusion of the input features and the quantum features and output them.

2. The method according to claim 1, characterized in that The first model includes: A plurality of quantum convolution kernel circuits of different scales, any one of which is used to extract features based on quantum computing and output classical features; Channel dimension, used to combine the classical features of multiple different scales; The residual connection layer is used to perform residual fusion on the input features and the concatenated classic features and output the result.

3. The method according to claim 2, characterized in that The quantum convolution kernel circuit includes: an encoding unit, an entanglement unit and a measurement unit; Wherein, the encoding unit is used to perform quantum encoding on the input feature; The entanglement unit is used to extract features of the quantum code and output the quantum features; The measuring unit is used to extract the quantum feature and output the classical feature.

4. The method according to claim 2, characterized in that The quantum convolution kernel circuit includes at least one of the following quantum gates: a Hadamard gate, a first rotation gate, a second rotation gate, and a third rotation gate. The Hadamard gate is used to convert input features into superposition state data, and the rotation gate is used to convert angle coding.

5. The method according to claim 4, characterized in that The performing quantum encoding on the input feature comprises: At the target scale, the quantum convolution kernel is used as a sliding window to determine the target feature map from the input features; Performing quantum coding on the pixel points of the target feature map based on the Hadamard gate to obtain a first quantum code for each pixel point, where the first quantum code is superposition state data; The first quantum code is angle-encoded based on at least one of the rotating gates to obtain a second quantum code for each of the pixel points.

6. The method according to any one of claims 1 to 5, characterized in that The method further comprises: Using the second model, semi-supervised training is performed on the first model; The second model is a teacher model of the first model; the second model is used to generate a feature representation of the input feature.

7. The method according to any one of claims 1 to 5, characterized in that The first model is an independent feature extraction model, or the first model is a sub-model of the third model; The third model is used to process image tasks; wherein the image tasks include one or more of the following: image classification, image segmentation, image recognition, target detection, and feature extraction.

8. The method according to claim 1, characterized in that The step of obtaining an image to be processed includes: Get the original image; Segmenting the original image to obtain a plurality of image sub-blocks; Randomly masking the plurality of image sub-blocks to obtain uncovered image sub-blocks; The uncovered image sub-block is used as an image to be processed.

9. An image processing device, characterized in that: include: An acquisition module, used for acquiring an image to be processed; A processing module is used to process the image to be processed using a pre-trained first model to obtain image features output by the first model; wherein the first model is used to extract quantum features corresponding to input features from multiple different scales, and to perform residual fusion of the input features and the quantum features and output them.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: The processor runs the computer program to implement the method according to any one of claims 1 to 8.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the method according to any one of claims 1 to 8.

12. A computer program product, characterized in that The invention comprises a computer-readable code, or a non-volatile computer-readable storage medium carrying the computer-readable code, wherein when the computer-readable code runs in a processor of an electronic device, the processor in the electronic device executes the method according to any one of claims 1 to 8.

Images