Image processing method and device, equipment and medium

Abstract

The invention provides an image processing method and device, equipment and a medium. The method comprises the following steps: acquiring a target image; the target image is input into the image segmentation model, an object segmentation result output by the image segmentation model is obtained, and the object segmentation result is used for representing contour information of an object in the target image; the image segmentation model is a model which is obtained by taking a feature map determined according to the sample image and an object contour in the sample image as input and through machine learning training and is used for segmenting the object contour in the image; the image segmentation model comprises an image feature analysis layer, and the image feature analysis layer is used for performing feature extraction on the target image based on a plurality of preset feature extraction requirement types, and obtaining contour information of an object corresponding to the feature extraction requirement type according to the extracted features. According to the method, the data volume of image analysis can be reduced, and the segmentation efficiency of the object on the image is improved.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to an image processing method, apparatus, device and medium. Background Art

[0002] Image processing primarily involves segmenting and annotating objects within the image. Object segmentation and annotation are tailored to different scenarios based on user needs. For example, segmenting and annotating obstacles (people or objects) in road images facilitates the development of autonomous driving for intelligent vehicles. Similarly, segmenting and annotating problematic areas (such as cracks and stains) in product images facilitates the development of automated inspection for intelligent devices.

[0003] Analyzing the semantic features of the entire image will result in a large amount of analysis, which reduces the efficiency of object segmentation and labeling on the image. Summary of the Invention

[0004] In view of the problems existing in the prior art, the present invention provides an image processing method, device, equipment and medium.

[0005] The present invention provides an image processing method, comprising: Acquire the target image; Inputting the target image into an image segmentation model to obtain an object segmentation result output by the image segmentation model, wherein the object segmentation result is used to represent contour information of the object in the target image; The image segmentation model is a model that is obtained through machine learning training by taking a feature map determined based on a sample image and an object contour in the sample image as input, and is used to segment the object contour in the image; The image segmentation model includes an image feature parsing layer, which is used to extract features from a target image based on a plurality of preset feature extraction requirement types, and obtain contour information of an object corresponding to the feature extraction requirement type according to the extracted features.

[0006] According to an image processing method provided by the present invention, the feature extraction requirement types include global type, local type and regional type, and the image feature parsing layer includes global parsing layer, local parsing layer and regional parsing layer. Accordingly, the target image is input into an image segmentation model to obtain an object segmentation result output by the image segmentation model, including: Obtain the semantic features of the target image and input the semantic features of the target image into the global parsing layer, the local parsing layer and the regional parsing layer; The global parsing layer extracts features from semantic features to obtain global feature information, and obtains contour information of global class objects on the target image based on the global feature information; The local parsing layer extracts features from the semantic features to obtain local feature information, and obtains contour information of the local class object on the target image based on the local feature information; The region parsing layer extracts the semantic features to obtain region feature information, and obtains the contour information of the region class object on the target image based on the region feature information; The contour information of the global class object, the contour information of the local class object and the contour information of the regional class object are fused on the target image.

[0007] According to an image processing method provided by the present invention, the image segmentation model further includes an image feature alignment layer, which is used to fuse contour information of an object corresponding to the required type of feature extraction onto the target image.

[0008] According to an image processing method provided by the present invention, the image feature alignment layer includes a first alignment layer, a second alignment layer, and a third alignment layer. Accordingly, the contour information of the global class object, the contour information of the local class object, and the contour information of the regional class object are fused on the target image, including: The first alignment layer aligns the contour information of the global class object at the initial position of the target image; The second alignment layer aligns the contour information of the local class object to the initial position of the target image; The third alignment layer aligns the contour information of the regional object to the initial position of the target image; Among them, the first alignment layer, the second alignment layer and the third alignment layer respectively use dilated convolution layers with different expansion rates.

[0009] According to an image processing method provided by the present invention, the image segmentation model further includes a downsampling layer, and the downsampling layer is used to extract features of the target image to obtain semantic features.

[0010] According to an image processing method provided by the present invention, the semantic features are input into the first alignment layer, the second alignment layer, and the third alignment layer. Accordingly, the first alignment layer aligns the contour information of the global class object at the initial position of the target image based on the semantic features; The second alignment layer aligns the contour information of the local class object at the initial position of the target image based on the semantic features; The third alignment layer aligns the contour information of the region-type object at the initial position of the target image based on the semantic features.

[0011] According to an image processing method provided by the present invention, the image segmentation model also includes an upsampling layer, which is used to restore the image that is a fusion of contour information of global class objects, contour information of local class objects, and contour information of regional class objects to the resolution of the target image.

[0012] The present invention also provides an image processing device, comprising: An acquisition module, used to acquire a target image; a processing module, configured to input the target image into an image segmentation model to obtain an object segmentation result output by the image segmentation model, wherein the object segmentation result is used to represent contour information of the object in the target image; The image segmentation model is a model that is obtained through machine learning training by taking a feature map determined based on a sample image and an object contour in the sample image as input, and is used to segment the object contour in the image; The image segmentation model includes an image feature parsing layer, which is used to extract features from a target image based on a plurality of preset feature extraction requirement types, and obtain contour information of an object corresponding to the feature extraction requirement type according to the extracted features.

[0013] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements any of the above-mentioned image processing methods when executing the program.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which implements any of the above-mentioned image processing methods when executed by a processor.

[0015] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any one of the above-mentioned image processing methods.

[0016] The present invention provides an image processing method, apparatus, device and medium. By inputting a target image into an image segmentation model, the image segmentation model extracts features from the target image based on a plurality of preset feature extraction requirement types, and obtains contour information of an object corresponding to the feature extraction requirement type based on the extracted features, the amount of data required for image analysis can be reduced, and the efficiency of segmentation of objects on the image can be accelerated. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. 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 It is a flowchart of the image processing method provided by the present invention.

[0019] Figure 2 It is a structural schematic diagram of the image processing device provided by the present invention.

[0020] Figure 3 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0021] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0022] The following combination Figure 1-Figure 3 The image processing method, apparatus, device and medium of the present invention are described.

[0023] Figure 1 A schematic diagram of the flow chart of an image processing method provided by the present invention is shown. Figure 1 , the method comprises the following steps: Step 11: Acquire the target image; Step 12: Input the target image into the image segmentation model to obtain an object segmentation result output by the image segmentation model, and the object segmentation result is used to characterize the contour information of the object in the target image; wherein the image segmentation model is a model that takes the feature map determined according to the sample image and the object contour in the sample image as input, is obtained through machine learning training, and is used to segment the object contour in the image; the image segmentation model includes an image feature parsing layer, and the image feature parsing layer is used to extract features of the target image based on a plurality of preset feature extraction requirement types, and obtain the contour information of the object corresponding to the feature extraction requirement type according to the extracted features.

[0024] Regarding steps 11 and 12, it should be noted that the image processing methods described herein primarily segment and label objects within the image. Object segmentation and labeling vary depending on user needs and are tailored to different scenarios. For example, segmenting and labeling obstacles (such as people or objects) in road images facilitates the development of autonomous driving for intelligent vehicles. Similarly, segmenting and labeling problematic areas (such as cracks and stains) in product images facilitates the development of automated detection for intelligent devices.

[0025] Analyzing the semantic features of the entire image results in a large amount of analysis, which reduces the efficiency of segmenting and labeling objects in the image. To this end, the present invention provides an image processing method that can classify objects in an image into different types of requirements and perform targeted object detection and recognition based on different types of requirements. This can reduce the amount of data required for image analysis and speed up the efficiency of segmenting and labeling objects in the image.

[0026] In the present invention, a target image is captured. This target image can be a static image or a dynamic image. The target image is input into an image segmentation model, and an object segmentation result is output by the image segmentation model. The object segmentation result is used to represent the contour information of the object in the target image. In other words, the image output by the image segmentation model contains the contour information of the object.

[0027] In the present invention, the image segmentation model is a model that takes a feature map determined according to a sample image and an object contour in the sample image as input, is obtained through machine learning training, and is used to segment the object contour in the image.

[0028] For example, to segment and label obstacles in road images during vehicle driving, a large number of road image samples can be collected, and then objects of different types can be labeled on the road image samples. Then, an initial neural network model is trained and verified based on the road image samples to obtain an image segmentation model.

[0029] During the training process of the model, the loss function is used to optimize the parameters of the model. OHEM (Online Hard Example Mining) is used to sort all samples from high to low according to the loss value based on the calculation of the cross-entropy loss function (Cross-Entropy Loss), and select the top samples. k The samples with the highest loss are back-propagated, where k is the proportion of samples with the highest loss.

[0030]

[0031] Refers to the cross entropy loss function value, It is The true label vector of the sample element. The probability distribution of the model prediction for samples elements.

[0032] For example, the demand for road images can be divided into global, local, and regional requirements. A road image represents the scene ahead of a vehicle, which may include distant buildings, near-field roads, and near-field obstacles (people, objects, etc.). In this driving scenario, the vehicle must navigate the road. Therefore, the road ahead serves as a crucial area for detecting obstacles, so image segmentation of the road is considered regional object segmentation. Obstacles on the road are local objects requiring special attention, so image segmentation of obstacles is considered local object segmentation. Objects in the distance are unlikely to impact driving, so image segmentation of distant scenes is considered global object segmentation.

[0033] Due to the analysis of the above-mentioned image processing scenarios, the image segmentation model includes an image feature parsing layer, which is used to extract features of the target image based on a plurality of preset feature extraction requirement types, and obtain the contour information of the object corresponding to the feature extraction requirement type according to the extracted features.

[0034] The feature extraction requirements here correspond to different types of objects in an image. In different usage scenarios, the model is trained on a large number of image samples from each scenario to accurately process the captured images, segmenting and labeling objects of different types.

[0035] The image processing method provided by the present invention inputs the target image into an image segmentation model, and the image segmentation model extracts features of the target image based on a plurality of preset feature extraction requirement types, and obtains contour information of the object corresponding to the feature extraction requirement type according to the extracted features, thereby reducing the amount of data for image analysis and speeding up the efficiency of segmentation of objects on the image.

[0036] In a further method of the above method, the feature extraction requirement types required by the present invention include global type, local type and regional type. Based on the type classification, the image feature parsing layer includes a global parsing layer, a local parsing layer and a regional parsing layer. At this time, the target image is input into the image segmentation model to obtain the object segmentation result output by the image segmentation model, which can be specifically: Obtain the semantic features of the target image and input the semantic features of the target image into the global parsing layer, the local parsing layer and the regional parsing layer; The global parsing layer extracts the semantic features to obtain global feature information, and based on the global feature information, obtains the contour information of the global class object on the target image; The local parsing layer extracts the semantic features to obtain local feature information, and based on the local feature information, obtains the contour information of the local class object on the target image; The regional parsing layer extracts the semantic features to obtain regional feature information, and then obtains the contour information of the regional object on the target image based on the regional feature information; The contour information of the global class object, the contour information of the local class object and the contour information of the regional class object are fused on the target image.

[0037] In this regard, it is important to note that we can first perform feature extraction on the target image to obtain its overall semantic features. Image semantics are divided into three layers: the visual layer, the object layer, and the conceptual layer. The visual layer is the underlying layer of understanding, namely color, texture, shape, etc. These features are referred to as underlying feature semantics; the object layer is the intermediate layer, containing attribute features, such as the state of an object at a given moment; the conceptual layer is the high-level layer, representing what the image conveys that is closest to human understanding. To put it simply, for example, if a picture contains sand, blue sky, and seawater, the visual layer distinguishes the blocks, the object layer represents the sand, blue sky, and seawater, and the conceptual layer represents the beach. This is the semantic information expressed by the image.

[0038] In the present invention, objects of different types need to be segmented. To this end, the semantic features of the target image are input into the global parsing layer, the local parsing layer and the regional parsing layer.

[0039] It should be noted that the regional parsing layer can be constructed by the large-core deep convolutional layer, the local parsing layer can be constructed by the n*n convolutional layer, and the global parsing layer can be constructed by the frequency-domain attention convolutional layer.

[0040] The local parsing layer extracts the semantic features to obtain local feature information, and then marks the contour information of the local class objects on the target image based on the local feature information.

[0041] The global parsing layer extracts semantic features to obtain global feature information, and then labels the contour information of global class objects on the target image based on the global feature information.

[0042] The regional parsing layer extracts the semantic features to obtain regional feature information, and then marks the contour information of the regional objects on the target image based on the regional feature information.

[0043] Since the convolution algorithms used in each parsing layer are different, the contour information of the global class object, the contour information of the local class object and the contour information of the regional class object must be fused on the image to form the required object segmentation result.

[0044] In the present invention, a global parsing layer is constructed by the convolutional layer of frequency domain attention, and the calculation formula is as follows:

[0045] in, , , , They represent Fourier convolution, inverse Fourier convolution, element-wise multiplication, and global average pooling, respectively. . is the output feature after frequency domain attention.

[0046] A further method of the present invention uses different convolutional layers to segment and label different objects, thereby accelerating the speed of image segmentation.

[0047] In a further method of the above method, the image segmentation model also includes an image feature alignment layer, which is used to fuse contour information of objects corresponding to the feature extraction requirement type onto the target image.

[0048] The image feature alignment layer includes the first alignment layer, the second alignment layer, and the third alignment layer. Accordingly, the contour information of the global class object, the contour information of the local class object, and the contour information of the regional class object are fused on the target image, including: The first alignment layer aligns the contour information of the global class object to the initial position of the target image; The second alignment layer aligns the contour information of the local class object to the initial position of the target image; The third alignment layer aligns the contour information of the regional object to the initial position of the target image; Among them, the first alignment layer, the second alignment layer and the third alignment layer respectively use dilated convolution layers with different expansion rates.

[0049] The contour information of the global class object, the contour information of the local class object and the contour information of the regional class object are fused at the initial position of the object on the image by adding them element by element.

[0050] For example, the fusion calculation method can be as follows:

[0051] , , and Represent input features, output features and 1×1 convolution respectively. Represents different specific operations: large kernel depth convolution, frequency domain attention and 1×1 convolution.

[0052] The present invention aims to set three groups of dilated convolutions with different expansion rates, and use different receptive fields to achieve alignment of features between different levels on the image.

[0053] In a further embodiment of the above method, the image segmentation model further includes a downsampling layer, which is used to extract features from the target image to obtain semantic features. The semantic features are input into the global parsing layer, the local parsing layer, and the regional parsing layer.

[0054] The downsampling layers consist of the backbone network in the ResNet-34 network.

[0055]

[0056] in, is the original input image, For the encoder, Represents the features obtained by the encoder.

[0057] The semantic features are also input into the first alignment layer, the second alignment layer and the third alignment layer. Accordingly, the first alignment layer aligns the contour information of the global class object at the initial position of the target image based on the semantic features; The second alignment layer aligns the contour information of the local class object to the initial position of the target image based on the semantic features; The third alignment layer aligns the contour information of the region-type object to the initial position of the target image based on semantic features.

[0058] Due to the semantic characteristics of the overall concept of the image, it can help serve as reference information and achieve alignment of features between different levels on the image with the help of different receptive fields.

[0059] In a further embodiment of the above method, the image segmentation model further includes an upsampling layer, which is configured to restore the image containing the contour information of the global class object, the contour information of the local class object, and the contour information of the regional class object to the resolution of the target image, so that the image output by the model is equal to the original target image in pixel size. The upsampling layer is implemented using a bilinear interpolation method.

[0060] The image processing device provided by the present invention is described below. The image processing device described below and the image processing method described above can be referenced to each other.

[0061] Figure 2 A schematic diagram of the process of an image processing device provided by the present invention is shown. Figure 2 The device includes a determination module 21 and a processing module 22, wherein: An acquisition module 21 is used to acquire a target image; A processing module 22 is configured to input the target image into an image segmentation model to obtain an object segmentation result output by the image segmentation model, wherein the object segmentation result is used to represent contour information of the object in the target image; The image segmentation model is a model that takes a feature map determined based on a sample image and an object contour in the sample image as input, is trained through machine learning, and is used to segment the object contour in the image; The image segmentation model includes an image feature parsing layer, which is used to extract features from the target image based on a plurality of preset feature extraction requirement types, and obtain contour information of the object corresponding to the feature extraction requirement type according to the extracted features.

[0062] Since the principles of the apparatus of the embodiment of the present invention are the same as those of the method of the above embodiment, more detailed explanations are omitted here.

[0063] It should be noted that, in the embodiment of the present invention, relevant functional modules may be implemented by a hardware processor.

[0064] The image processing device provided by the present invention inputs the target image into the image segmentation model, and the image segmentation model extracts features of the target image based on multiple preset feature extraction requirement types, and obtains contour information of the object corresponding to the feature extraction requirement type based on the extracted features. This can reduce the amount of data for image analysis and speed up the efficiency of segmentation of objects on the image.

[0065] Figure 3 An example of a physical structure diagram of an electronic device is shown below. Figure 3As shown, the electronic device may include: a processor 31 (processor), a communication interface 32 (Communications Interface), a memory 33 (memory) and a communication bus 34, wherein the processor 31, the communication interface 32, and the memory 33 communicate with each other via the communication bus 34. The processor 31 may call the logic instructions in the memory 53 to execute the image processing method, which includes: acquiring a target image; inputting the target image into an image segmentation model to obtain an object segmentation result output by the image segmentation model, wherein the object segmentation result is used to characterize the contour information of the object in the target image; wherein the image segmentation model is a model that is obtained by machine learning training using a feature map determined according to a sample image and the object contour in the sample image as input, and is used to segment the object contour in the image; the image segmentation model includes an image feature parsing layer, which is used to extract features from the target image based on a plurality of preset feature extraction requirement types, and obtain the contour information of the object corresponding to the feature extraction requirement type based on the extracted features.

[0066] Furthermore, the logic instructions in the aforementioned memory 33 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include various media capable of storing program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0067] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the image processing method provided by the above-mentioned methods, which includes: obtaining a target image; inputting the target image into an image segmentation model to obtain an object segmentation result output by the image segmentation model, and the object segmentation result is used to characterize the contour information of the object in the target image; wherein the image segmentation model is a model that takes a feature map determined according to a sample image and an object contour in the sample image as input, is obtained through machine learning training, and is used to segment the contour of the object in the image; the image segmentation model includes an image feature parsing layer, which is used to extract features of the target image based on a preset plurality of feature extraction requirement types, and obtain the contour information of the object corresponding to the feature extraction requirement type based on the extracted features.

[0068] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the image processing method provided by the above-mentioned methods, the method comprising: acquiring a target image; inputting the target image into an image segmentation model to obtain an object segmentation result output by the image segmentation model, the object segmentation result being used to characterize the contour information of the object in the target image; wherein the image segmentation model is a model obtained by machine learning training using a feature map determined according to a sample image and an object contour in the sample image as input, and is used to segment the contour of the object in the image; the image segmentation model comprises an image feature parsing layer, the image feature parsing layer being used to perform feature extraction on the target image based on a preset plurality of feature extraction requirement types, and obtain contour information of the object corresponding to the feature extraction requirement type based on the extracted features.

[0069] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0070] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. An image processing method, characterized in that: include: Get the target image; Inputting the target image into an image segmentation model to obtain an object segmentation result output by the image segmentation model, wherein the object segmentation result is used to characterize contour information of an object in the target image; The image segmentation model is a model that uses a feature map determined according to a sample image and an object contour in the sample image as input, is obtained through machine learning training, and is used to segment the object contour in the image; The image segmentation model includes an image feature analysis layer, which is used to extract features from a target image based on a plurality of preset feature extraction requirement types, and obtain contour information of an object corresponding to the feature extraction requirement type according to the extracted features.

2. The image processing method according to claim 1, characterized in that: The feature extraction requirement types include global type, local type and regional type, the image feature parsing layer includes global parsing layer, local parsing layer and regional parsing layer, and accordingly, the target image is input into the image segmentation model to obtain the object segmentation result output by the image segmentation model, including: Obtaining the semantic features of the target image, and inputting the semantic features of the target image into the global parsing layer, the local parsing layer, and the regional parsing layer; The global parsing layer extracts features from semantic features to obtain global feature information, and obtains contour information of global class objects on the target image based on the global feature information; The local parsing layer extracts features from semantic features to obtain local feature information, and obtains contour information of local class objects on the target image based on the local feature information; The regional analysis layer extracts the semantic features to obtain regional feature information, and obtains the contour information of the regional object on the target image based on the regional feature information; The contour information of the global class object, the contour information of the local class object and the contour information of the regional class object are fused on the target image.

3. The image processing method according to claim 2, characterized in that: The image segmentation model also includes an image feature alignment layer, which is used to fuse contour information of an object corresponding to the feature extraction requirement type onto the target image.

4. The image processing method according to claim 3, characterized in that: The image feature alignment layer includes a first alignment layer, a second alignment layer and a third alignment layer. Accordingly, the contour information of the global class object, the contour information of the local class object and the contour information of the regional class object are fused on the target image, including: The first alignment layer aligns the contour information of the global class object at the initial position of the target image; The second alignment layer aligns the contour information of the local class object at the initial position of the target image; The third alignment layer aligns the contour information of the region object at the initial position of the target image; Among them, the first alignment layer, the second alignment layer and the third alignment layer respectively use dilated convolution layers with different expansion rates.

5. The image processing method according to claim 4, characterized in that: The image segmentation model also includes a downsampling layer, which is used to extract features from the target image to obtain semantic features.

6. The image processing method according to claim 5, characterized in that: The semantic features are input into the first alignment layer, the second alignment layer and the third alignment layer, and accordingly, the first alignment layer aligns the contour information of the global class object at the initial position of the target image based on the semantic features; The second alignment layer aligns the contour information of the local class object at the initial position of the target image based on the semantic features; The third alignment layer aligns the contour information of the region type object at the initial position of the target image based on the semantic features.

7. The image processing method according to claim 6, characterized in that: The image segmentation model also includes an upsampling layer, which is used to restore the image that is fused with contour information of global class objects, contour information of local class objects, and contour information of regional class objects to the resolution of the target image.

8. An image processing device, characterized in that: include: An acquisition module, used for acquiring a target image; A processing module, used for inputting the target image into an image segmentation model to obtain an object segmentation result output by the image segmentation model, wherein the object segmentation result is used for representing contour information of an object in the target image; The image segmentation model is a model that uses a feature map determined according to a sample image and an object contour in the sample image as input, is obtained through machine learning training, and is used to segment the object contour in the image; The image segmentation model includes an image feature analysis layer, which is used to extract features from a target image based on a plurality of preset feature extraction requirement types, and obtain contour information of an object corresponding to the feature extraction requirement type according to the extracted features.

9. An electronic 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 program, the image processing method according to any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the image processing method according to any one of claims 1 to 7 is implemented.

Images