Image editing method and device, computer device and storage medium

Abstract

The application provides an image editing method and device, computer equipment and a storage medium. The method acquires a single certificate image to be edited and corresponding insurance type information in a preset claim settlement system; detects a claim settlement related area in the single certificate image according to the insurance type information; detects a privacy area in the single certificate image; determines a region identification type according to a region contour corresponding to the claim settlement related area; the region identification type includes one or more of a circular identification, a square identification and an arrow identification; acquires color distribution information corresponding to the privacy area in the single certificate image; the color distribution information includes a color corresponding to each pixel point in the privacy area; generates mosaic information corresponding to the privacy area according to the color distribution information; the mosaic information includes a superimposed color corresponding to each pixel point in the privacy area; performs color superimposition on each pixel point in the privacy area according to the mosaic information, marks the region contour according to the identification type, and completes editing of the single certificate image.

Description

Technical Field

[0001] This application relates to the field of image processing, with applications in the financial sector, and particularly to an image editing method, apparatus, computer device, and storage medium. Background Technology

[0002] In the financial sector, after recording information during the claims process, claims adjusters typically upload photos of supporting documents as verification evidence. These photos usually refer to images of various documents and evidence related to the claims application. However, there are many types of supporting documents, often requiring annotations of details and relevant information. The traditional method involves using specialized image editing software to process the photos before uploading, which is cumbersome and time-consuming, resulting in a poor user experience and significantly impacting the progress of the claims process.

[0003] Therefore, there is an urgent need for an image editing method for document photos in the financial sector, especially in the field of insurance claims, to improve the processing efficiency of claims personnel. Summary of the Invention

[0004] This application provides an image editing method, apparatus, computer device, and storage medium, aiming to solve the problem that in the financial claims process, due to the wide variety of document photos, it is often necessary to annotate some details and add relevant information. The traditional method is to use special image editing software to process the photos and then upload them, which is cumbersome and time-consuming, resulting in a poor working experience for operators and significantly affecting the progress of the claims process.

[0005] In a first aspect, this application provides an image editing method, including:

[0006] Obtain the document image to be edited and the corresponding insurance information from the pre-set claims system;

[0007] Based on the insurance type information, detect the claim-related areas in the document images;

[0008] Privacy areas were detected in document images outside of the claims-related areas.

[0009] The region identifier type is determined based on the region outline corresponding to the relevant claim area in the document image; wherein, the region identifier type includes at least one or more of the following: circular identifier, square identifier, and arrow identifier;

[0010] Obtain the color distribution information corresponding to the privacy region in the document image; the color distribution information includes the color corresponding to each pixel in the privacy region.

[0011] Mosaic information corresponding to the privacy region is generated based on the color distribution information; the mosaic information includes the overlay color corresponding to each pixel in the privacy region.

[0012] Based on the mosaic information, each pixel in the privacy area is overlaid with color, and the region outline is marked according to the identifier type, thus completing the editing of the document image.

[0013] In some embodiments, detecting claim-related regions in a document image based on insurance type information includes: acquiring a pre-built image recognition model library, which includes multiple image recognition models and recognition categories corresponding to each image recognition model; matching the insurance type information with the recognition categories to acquire the image recognition model corresponding to the successfully matched recognition category; inputting the document image into the image recognition model, which then parses the document image and outputs the claim-related regions.

[0014] For example, before acquiring the pre-built image recognition model library, the method further includes: acquiring multiple historical document images of claims already processed and the corresponding insurance type information and historical claim area for each historical document image from the claims system; classifying the multiple historical document images according to the insurance type information to obtain multiple historical image groups, each historical image group including at least one historical document image; generating multiple image recognition models to be trained based on the number of historical image groups; inputting the historical document images of each historical image group into an image recognition model, the image recognition model parsing each historical document image and outputting the claim prediction area corresponding to each historical document image; training each corresponding image recognition model based on the claim prediction area and historical claim area corresponding to each historical document image; and generating the recognition category of the corresponding image recognition model based on the insurance type information corresponding to each historical image group.

[0015] In some embodiments, detecting a privacy region in a document image outside the claim-related area includes: performing optical character recognition in the other areas outside the claim-related area to obtain text information corresponding to the document image and location information corresponding to the text information; parsing the text information based on natural language processing technology to obtain semantic information corresponding to the text information; and matching the semantic information with a preset privacy type to generate a privacy region from the location information corresponding to the successfully matched semantic information.

[0016] In some embodiments, determining the region identifier type based on the region contour corresponding to the claim-related region in the document image includes: parsing the region contour and obtaining the geometric feature information corresponding to the region contour; the geometric feature information includes at least one or more of aspect ratio, roundness, and corner features; if the regularity of the region contour is determined to be greater than a preset regularity based on the geometric feature information, the identifier type is determined to be a circular identifier; if the regularity of the region contour is determined to be less than or equal to the preset regularity based on the geometric feature information, the identifier type is determined to be a square identifier; if the area of ​​the region contour is determined to be less than a preset area based on the geometric feature information, the identifier type is determined to be an arrow identifier.

[0017] In some embodiments, before editing the document image, the method further includes: obtaining the shooting location information corresponding to the document image; detecting the claim-related area and obtaining the element type corresponding to the claim-related area; generating watermark information based on the element type, shooting location information and insurance type information; adding the watermark information to the document image; and completing the editing of the document image.

[0018] In some embodiments, color overlay is applied to each pixel in the privacy area based on mosaic information, and the region outline is marked according to the identifier type to complete the editing of the document image, including: obtaining an initialized preset container; converting the document image into a preset document object model format; setting the converted document image as a background image in the preset container; color overlaying is applied to each pixel in the privacy area corresponding to the background image based on mosaic information in the preset container, and the region outline corresponding to the background image is marked according to the identifier type to complete the editing of the background image; and the edited background image is uploaded to the claims system.

[0019] Secondly, this application provides an image editing apparatus, comprising:

[0020] The information acquisition unit is used to acquire the document image to be edited and the corresponding insurance information from the preset claims system;

[0021] The region detection unit is used to detect claim-related regions in the document image based on insurance type information;

[0022] A privacy detection unit is used to detect privacy areas in document images outside of the claims-related areas;

[0023] The identifier determination unit is used to determine the region identifier type based on the region outline corresponding to the claim-related region in the document image; wherein, the region identifier type includes at least one or more of the following: text identifier, circular identifier, square identifier, and arrow identifier;

[0024] The distribution determination unit is used to obtain color distribution information corresponding to the privacy region in the document image; the color distribution information includes the color corresponding to each pixel in the privacy region;

[0025] The overlay determination unit is used to generate mosaic information corresponding to the privacy region based on the color distribution information; the mosaic information includes the overlay color corresponding to each pixel in the privacy region.

[0026] The editing completion unit is used to overlay the color of each pixel in the privacy area according to the mosaic information, mark the outline of the area according to the identification type, and complete the editing of the document image.

[0027] Thirdly, this application also provides a computer device, comprising:

[0028] Memory and processor;

[0029] The memory is used to store computer programs;

[0030] The processor is configured to execute the computer program and, in executing the computer program, implement the steps of the image editing method as described in the first aspect above.

[0031] Fourthly, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the steps of the image editing method described in the first aspect above.

[0032] This application provides an image editing method, apparatus, computer device, and storage medium. The method acquires the document image to be edited from a preset claims system. Based on the acquired insurance information, the document image is classified, and claim-related areas are identified. Claim-related areas are automatically detected in the image. Areas outside the claim-related areas (i.e., privacy areas) that may contain private information are identified. The outline of the claim-related areas is then determined, and their area identification types are defined, using identification styles such as circles, squares, or arrows to mark the areas. Color distribution information within the privacy areas is acquired, and the color of each pixel is analyzed. Based on this information, a mosaic effect is generated to conceal the privacy information. The generated mosaic information is applied to the privacy areas, and color overlay processing is performed on each pixel. The final edited document image includes the identified claim-related areas and effectively masks the privacy information. In summary, this method improves image processing efficiency and privacy protection in the financial claims process through automated and intelligent image editing technology, making the overall operation smoother and more secure.

[0033] The method provided thus has the following beneficial effects:

[0034] 1. Improve work efficiency: Automated area detection and labeling reduce manual operations and improve the efficiency of claims review and processing.

[0035] 2. User experience optimization: The operation process has been simplified, eliminating the need for additional image editing software and significantly improving the operator's experience.

[0036] 3. Privacy Protection: Effectively protects sensitive information that may be involved in document images by automatically adding mosaic effects to privacy areas to ensure information security.

[0037] 4. Accuracy and consistency: Based on insurance type information and preset rules, relevant areas are detected and labeled, which improves the accuracy and consistency of image processing.

[0038] 5. Flexibility and scalability: It provides multiple identifier types, which makes the method highly flexible and able to adapt to different claims needs and insurance characteristics.

[0039] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0040] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a schematic flowchart illustrating the steps of an image editing method provided in an embodiment of this application;

[0042] Figure 2 This is a schematic flowchart illustrating the steps of a claims-related area detection method according to an embodiment of this application;

[0043] Figure 3 This is a schematic flowchart illustrating the steps of a privacy region detection method provided in an embodiment of this application;

[0044] Figure 4 This is a schematic diagram of the structure of an image editing device provided in an embodiment of this application;

[0045] Figure 5 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application.

[0046] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Detailed Implementation

[0047] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0048] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0049] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" are not necessarily different.

[0050] It should 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 scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0051] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0052] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0053] In the financial sector, after recording information during the claims process, claims adjusters typically upload photos of supporting documents as verification evidence. These photos usually refer to images of various documents and evidence related to the claims application. However, there are many types of supporting documents, often requiring annotations of details and relevant information. The traditional method involves using specialized image editing software to process the photos before uploading, which is cumbersome and time-consuming, resulting in a poor user experience and significantly impacting the progress of the claims process.

[0054] Therefore, there is an urgent need for an image editing method for document photos in the financial sector, especially in the field of insurance claims, to improve the processing efficiency of claims personnel.

[0055] To resolve the above issues, please refer to [link / reference]. Figure 1 , Figure 1 This is a schematic flowchart of an image editing method provided in one embodiment of this application. The image editing method can be implemented by a computer device, which can be deployed on a single server or a server cluster. It can also be deployed on a handheld terminal, laptop, wearable device, or robot, etc.

[0056] It should be noted that the acquisition of relevant information involved in the process of obtaining any information, such as insurance information, using the provided method is in compliance with relevant regulations and is carried out with the user's consent. It will not infringe on the user's privacy or violate relevant laws and regulations.

[0057] Specifically, such as Figure 1 As shown, the provided image editing method includes steps S101 to S107, which are detailed below:

[0058] Step S101. Obtain the document image to be edited and the corresponding insurance information from the preset claims system.

[0059] Specifically, within the established claims system, images of documents to be edited are retrieved via API interfaces or database queries. These images are typically stored in formats such as JPEG and PNG, and the documents to be processed may include various types of insurance files, such as policy images and accident scene photos. In addition to acquiring images, the method also needs to retrieve relevant insurance information, such as insurance type (auto insurance, property insurance, health insurance, etc.), policy number, policyholder's name, and contact information. For example, the computer retrieves a customer's auto insurance policy photo and related information from the database (Insurance type: Auto insurance, Policy number: 123456, Policyholder: Zhang San). This ensures consistency between the processed images and data, reducing data entry errors. It also increases processing speed, allowing claims processing to quickly proceed to subsequent steps and improving operational efficiency.

[0060] Step S102. Detect the claim-related area in the document image based on the insurance type information.

[0061] Specifically, computer vision technology, combined with deep learning algorithms (such as YOLO and Faster R-CNN), is used to identify and locate claim-related regions in document images. This step involves semantic segmentation and object detection, training models to identify key information required for insurance claims. Claim-related regions may include photos of the damage, expense details, accident descriptions, etc. For example, in a car insurance claim scene photo, the system identifies the collided vehicles, damaged parts, and the location of the accident. Automated region recognition reduces errors from manual annotation and improves processing accuracy. Furthermore, it can process multiple images simultaneously, accelerating the claims process.

[0062] In some embodiments, such as Figure 2 As shown, the claim-related area is detected in the document image based on the insurance type information, including steps S102a to S102c.

[0063] Step S102a. Obtain a pre-built image recognition model library, which includes multiple image recognition models and the recognition category corresponding to each image recognition model.

[0064] Step S102b. Match the insurance type information with the identification category to obtain the image recognition model corresponding to the successfully matched identification category.

[0065] Step S102c. Input the document image into the image recognition model. The image recognition model parses the document image and outputs the claim-related area.

[0066] Image recognition model libraries are collections of pre-trained models, each optimized for a specific recognition category. These models may be built using techniques such as Convolutional Neural Networks (CNNs) and deep learning. Each model corresponds to one or more recognition categories, potentially covering image features in different insurance scenarios, such as vehicle damage, medical bills, and accident scenes. The process involves matching the provided insurance information (e.g., car insurance, health insurance) with the recognition categories in the model library. This matching process may utilize information retrieval techniques or simple rule matching to ensure the selection of the image recognition model best suited to the current insurance type. Based on the matching results, a suitable image recognition model is selected for processing the current document image. This process optimizes model selection efficiency and prevents unnecessary computational overhead. The document image is then input into the selected recognition model. The model parses the image content and outputs the claim-related area. The parsing process involves steps such as feature extraction, feature matching, and boundary detection.

[0067] When processing auto insurance claims, a photo containing a damaged vehicle is identified. An image recognition model library is accessed, containing a model specifically designed for auto insurance damage detection. The insurance type "auto insurance" is matched, and a recognition model relevant to damage detection is selected. The vehicle photo is then input into the model. The model outputs information about the damaged areas of the vehicle in the image, which is used for labeling and further processing.

[0068] Correspondingly, each image recognition model is optimized for a specific category, providing higher accuracy and precision, which is crucial in the claims process. Through automatic model selection and image parsing, manual intervention is reduced, and claims-related areas are identified quickly and in real time, accelerating the entire claims process. Thanks to the model library structure, new models can be continuously added to support more insurance types and different types of claims needs. A flexible matching mechanism allows the system to adapt to various scenarios, helping to maintain its advanced nature and competitiveness. The automated selection and recognition process reduces human decision-making and operational complexity, improving user experience and processing consistency.

[0069] Through the image recognition mechanism described in the above embodiments, the insurance system can handle various claims-related issues in a more intelligent and automated manner, ensuring efficient operation and user privacy protection. This method not only improves the system's processing capabilities but also significantly enhances user satisfaction and trust.

[0070] For example, before acquiring the pre-built image recognition model library, the method further includes: acquiring multiple historical document images of claims already processed and the corresponding insurance type information and historical claim area for each historical document image from the claims system; classifying the multiple historical document images according to the insurance type information to obtain multiple historical image groups, each historical image group including at least one historical document image; generating multiple image recognition models to be trained based on the number of historical image groups; inputting the historical document images of each historical image group into an image recognition model, the image recognition model parsing each historical document image and outputting the claim prediction area corresponding to each historical document image; training each corresponding image recognition model based on the claim prediction area and historical claim area corresponding to each historical document image; and generating the recognition category of the corresponding image recognition model based on the insurance type information corresponding to each historical image group.

[0071] This involves extracting multiple historical document images of completed claims from the claims system, along with the corresponding insurance type information and accurately labeled historical claim areas for each image. This step requires database operations and data extraction techniques. Based on the extracted insurance type information, the multiple historical document images are categorized into multiple historical image groups. Images within each group belong to the same insurance type or are related to the same type of claim. The classification step can use insurance type tags or document metadata for grouping. For each historical image group, an image recognition model to be trained is initialized. The initial structure and parameters of these models can be set based on deep learning frameworks (such as TensorFlow or PyTorch).

[0072] Images from each historical image group are input into the corresponding image recognition model. The image content is parsed, and a claim prediction region is output. The historical claim regions are used as labels, compared with the predicted regions, and adjusted accordingly to train the model and improve its recognition accuracy. Loss functions from supervised learning (such as cross-entropy loss) are used to optimize the model parameters. Based on the insurance type information corresponding to each historical image group, the model's recognition category is defined. This process ensures that each model focuses on a specific recognition task, improving processing efficiency and accuracy.

[0073] For example, insurance companies possess a large number of auto insurance and health insurance claims. Historical images of auto insurance claims and health insurance claims are extracted from the system. These are then grouped separately into auto insurance and health insurance groups. A recognition model is initialized for each group, and the models are trained separately for the auto insurance and health insurance groups. Auto insurance group images are input into the model for training, and the model learns to recognize vehicle damage areas; health insurance images are input into another model to learn to recognize medical expense details. During training, the models are corrected and optimized using labeled data from historical claims areas. Ultimately, two recognition models are generated, one focusing on auto insurance claims area recognition and the other on health insurance claims area recognition.

[0074] By leveraging historical data, the system can quickly build and train multiple recognition models, significantly shortening development and deployment cycles. The trained models exhibit high specificity and accuracy for particular insurance types. Simultaneously, comparing historical labeled data with predicted outputs allows for precise adjustment of model parameters, improving the accuracy and consistency of recognition results. With a flexible model building and training mechanism, the system can easily expand its functionality by adding new historical datasets and corresponding models to adapt to more insurance types. High-precision automated recognition reduces the workload of manual inspection and labeling, accelerates the claims process, and improves overall processing efficiency. Transforming large-scale, usable historical data into training resources enhances the accuracy and practicality of existing technologies, further improving the economic benefits of data utilization. Through the technical path described above, insurance companies can effectively utilize historical claims data, not only improving the accuracy and efficiency of image recognition but also optimizing the overall claims process and enhancing the value of data acquisition.

[0075] Step S103. Detect privacy areas in the document image outside the claim-related area.

[0076] Specifically, image processing algorithms (such as edge detection-based algorithms or image segmentation techniques) are used to identify and mark privacy-sensitive areas in document images. These privacy-sensitive areas generally include personal identification information (such as ID card numbers, bank account information), addresses, and contact information. OCR (Optical Character Recognition) technology is used to automatically identify text information in the image to help determine the location of privacy-sensitive areas. For example, in car insurance policy images, the insured's name, ID card number, and other information can be identified and marked as privacy-sensitive areas requiring protection. By clearly defining the scope of personal privacy information, data security is ensured. This complies with current data protection laws and regulations (such as GDPR, CCPA, etc.) and maintains the company's image and reputation.

[0077] In some embodiments, such as Figure 3 As shown, privacy areas are detected in the document image outside the claim-related area, including steps S103a to S103c.

[0078] Step S103a. Perform optical character recognition in areas other than the claim-related area to obtain the text information corresponding to the document image and the location information corresponding to the text information.

[0079] Step S103b. Parse the text information based on natural language processing technology to obtain the semantic information corresponding to the text information.

[0080] Step S103c. Match semantic information with preset privacy types, and generate a privacy region from the location information corresponding to the successfully matched semantic information.

[0081] OCR technology is applied to non-claims-related areas of document images to extract text information and corresponding location information. OCR technology involves image preprocessing (such as noise reduction and binarization) and character recognition (such as using OCR engines like Tesseract). The processing results include text content and the coordinates or boundary information of each text block in the image. The extracted text information is then parsed using NLP technology to obtain semantic information. NLP technology may involve word segmentation, part-of-speech tagging, named entity recognition (NER), and syntactic analysis in this process. The parsing results form a set of semantic tags, which may identify the specific meaning of the text (such as name, address, ID number, etc.). The obtained semantic information is then matched against a preset privacy type. The privacy type can be predefined, such as keywords or patterns containing sensitive information like name, ID number, address, and contact information. Based on the matching results, it is determined which text semantics meet the privacy type criteria, identifying the privacy information that needs protection. For successfully matched privacy information, a privacy region is generated based on its location information. This step typically involves defining the identified sensitive information location as the area to be masked or processed. These areas can then be blurred or pixelated to ensure that the information cannot be visually identified.

[0082] For example, when processing a customer's insurance application, it's desirable to identify potential personal information. OCR is applied to detect and extract text information from the application image, such as "Name: Zhang San, Phone: 1234567890". NLP analysis is used to identify the semantics of "Zhang San" and "1234567890" as "name" and "phone number," respectively. The semantics of "name" and "phone number" are matched against preset privacy types to confirm that these two pieces of information belong to the privacy content to be protected. Based on the matching results, their positions in the image are determined, and regions for subsequent privacy processing are generated.

[0083] By leveraging OCR and NLP technologies, privacy information can be automatically detected and identified, eliminating the need for manual item-by-item checking and improving efficiency and accuracy. Semantic-based matching is more flexible and adaptable than keyword-based matching, better handling diverse textual expressions and linguistic variations. Automated identification and annotation of privacy areas ensures sensitive data is not leaked, helping companies comply with privacy regulations such as GDPR and reducing legal risks. Rapid identification and processing of privacy information gives customers greater confidence when providing materials, enhancing trust between customers and businesses. This technology is not limited to the insurance sector but can also be applied to industries that require processing large amounts of personal information, such as banking, healthcare, and government agencies, demonstrating broad applicability. Through these examples, insurance companies and related enterprises can more effectively and intelligently protect customer privacy information, improve processing efficiency, reduce the risk of human error, and provide strong technical support for meeting regulatory compliance requirements.

[0084] Step S104. Determine the region identifier type based on the region outline corresponding to the claim-related region in the document image; wherein, the region identifier type includes at least one or more of the following: circular identifier, square identifier, and arrow identifier.

[0085] Specifically, the most suitable label type is determined based on the shape and characteristics of the detected claim-related areas. This step primarily relies on geometric feature analysis of the image region, with the label type selection determined by a rule engine and model. Multiple label types are available, including circles, squares, and arrows, allowing for adaptability to various usage scenarios. For example, for vehicle damage photos, square labels are used to surround the damaged area, and arrows are used in the accident description area to point to the cause of the accident. This flexible labeling method allows users to clearly identify important information and improves image readability. Increased labeling accuracy helps claims reviewers quickly understand the accident details and the extent of the damage.

[0086] Step S105. Obtain the color distribution information corresponding to the privacy region in the document image; the color distribution information includes the color corresponding to each pixel in the privacy region.

[0087] Specifically, pixel-level processing algorithms are used to obtain the color information of each pixel within the privacy region, primarily by traversing all pixels in that region. This information provides the foundational data for subsequent mosaic generation. Methods such as color histograms can be used to analyze and summarize the color distribution within the privacy region. For example, the RGB values ​​of each pixel in the customer's ID number area can be extracted to create a detailed color data table. By providing the necessary color information for processing the privacy region, the naturalness and consistency of the mosaic effect are ensured, avoiding visual abruptness after processing and making the final result more aesthetically pleasing.

[0088] Step S106. Generate mosaic information corresponding to the privacy region based on the color distribution information; the mosaic information includes the superimposed color corresponding to each pixel in the privacy region.

[0089] Specifically, the color distribution information of the privacy area is used to generate a corresponding mosaic effect. This is typically achieved by overlaying colors onto each pixel, ensuring that the privacy information cannot be identified. Blurring or block-based techniques are employed to render the privacy area as a mosaic pattern composed of small color blocks, effectively obscuring its true information. For example, by applying a mosaic to the area of ​​a customer's ID card, a blurred area is generated based on the pixel color distribution, concealing sensitive information within. This effectively prevents the leakage of sensitive information and protects customer privacy. It also ensures that the digitally processed image maintains good visual consistency and does not affect the document's basic functionality.

[0090] Step S107. Apply color overlay to each pixel in the privacy area based on the mosaic information, mark the area outline according to the identifier type, and complete the editing of the document image.

[0091] Specifically, using the mosaic information generated in the previous steps, each pixel in the privacy area is overlaid with a corresponding color. Based on the identification type (circle, square, arrow, etc.), identification and explanatory text are added to the claim-related areas. This ultimately creates a new, edited document image, ready to be uploaded to the claims system or sent to relevant reviewers. For example, the computer device processes the customer's ID card area as a mosaic, while adding square identification and the descriptive text "Accident Damage" around the vehicle damage area, generating the final edited image. Automated editing significantly improves the efficiency of claims processing and reduces the probability of manual changes. The professionalism and refinement of the finished image improves customer and reviewer satisfaction with the claims process and enhances the user experience of the entire claims system.

[0092] Through the above steps, the provided image editing method greatly improves the efficiency of image processing in the financial claims process, protects customer privacy, and ensures the accuracy of claims information, thereby promoting the optimization of the entire claims process.

[0093] In some embodiments, determining the region identifier type based on the region contour corresponding to the claim-related region in the document image includes: parsing the region contour and obtaining the geometric feature information corresponding to the region contour; the geometric feature information includes at least one or more of aspect ratio, roundness, and corner features; if the regularity of the region contour is determined to be greater than a preset regularity based on the geometric feature information, the identifier type is determined to be a circular identifier; if the regularity of the region contour is determined to be less than or equal to the preset regularity based on the geometric feature information, the identifier type is determined to be a square identifier; if the area of ​​the region contour is determined to be less than a preset area based on the geometric feature information, the identifier type is determined to be an arrow identifier.

[0094] Contour detection is performed on claim-related areas in document images. This may utilize image processing techniques such as edge detection algorithms (e.g., Canny edge detection) and contour extraction techniques (e.g., findContours in OpenCV). The detection results provide boundary information for the regions, facilitating further analysis. The geometric features of the contours are analyzed, including: Aspect ratio: calculating the ratio of the contour's width to its height; Roundness: calculating the degree of deviation from a perfect circle, possibly approximating it based on the ratio of area to perimeter; Corner features: analyzing whether the contour exhibits distinct angles, such as the acute angles of a rectangle.

[0095] Regularity is determined by geometric features, such as judging the symmetry and uniformity of a region. A pre-set regularity threshold is used as a comparison; if the regularity is greater than this threshold, the region is considered regular (e.g., circular); if it is less than or equal to this threshold, it is considered irregular (e.g., square). If the regularity is greater than the pre-set value, the region is marked as circular. This usually indicates high symmetry and uniform boundaries. If the regularity is less than or equal to the pre-set value, the region is marked as square, indicating relative irregularity but still possessing certain boundary characteristics. If the region's area is less than the pre-set value, it is marked as an arrow, typically used to indicate direction or emphasize certain areas.

[0096] For example, analyzing images of marked damage areas in car insurance claims can automatically determine the label type. This is achieved by acquiring and analyzing the contours of the damage area images. The aspect ratio, roundness, and corner features are calculated. For instance, if a region's aspect ratio is close to 1 and its roundness is high, it is considered circular. If the regularity is greater than a preset threshold, the region is labeled as circular. If another region has low regularity and a more square boundary, it is labeled as square. If certain small, protruding marks conform to arrow features, they are labeled as arrows to indicate the start or direction of the damage. By automatically identifying different label types based on image features, human error and subjective judgment bias are reduced. Different shaped labels can be used to visualize different information, aiding in the understanding and analysis of the content in document images.

[0097] Automated geometric analysis significantly reduces processing time, making it suitable for large-scale batch processing of document images. Pre-identification of identifier types provides a clear starting point and conditions for subsequent processing (such as data extraction, analysis, and pattern recognition). The technology is not limited to insurance claims but can be extended to other applications requiring image recognition and analysis, such as map annotation and architectural scheme recognition.

[0098] The above embodiments provide an efficient and scalable method for identifying identifier types by analyzing the geometric features of image contours and combining preset regularity and area standards. This not only improves the accuracy of image parsing but also provides a foundation for information visualization and data mining, contributing to improved overall system performance and user experience.

[0099] In some embodiments, before editing the document image, the method further includes: obtaining the shooting location information corresponding to the document image; detecting the claim-related area and obtaining the element type corresponding to the claim-related area; generating watermark information based on the element type, shooting location information and insurance type information; adding the watermark information to the document image; and completing the editing of the document image.

[0100] Extracting the shooting location information of the document image is typically done through image metadata (such as EXI F information), including geographical information such as latitude and longitude. The image is then analyzed to detect claim-related areas and identify element types within those areas. This may involve using image recognition technology to determine different object or text types, such as license plates, damaged areas, etc. Specific watermark information is generated based on element type, shooting location information, and insurance type information. The watermark can be in text or graphic form and may include: element type description (e.g., "license plate" or "damage image"), precise geographical location (e.g., city name, region), and relevant insurance type information (e.g., "car insurance" or "health insurance"). Watermark generation may involve formatting and encoding steps to ensure end-user readability and machine processing compatibility. The generated watermark information is then added to the document image. This step can be achieved using image processing tools, ensuring the watermark does not affect the readability of the main content. The layout, transparency, and position of the watermark can be customized according to specific application scenarios to achieve both anti-counterfeiting and non-interference with user reading.

[0101] For example, when processing car insurance claims, ensure the authenticity and integrity of submitted images by examining photo archives. This involves extracting the shooting location from the vehicle damage photos in the claim, such as "Chaoyang District, Beijing." The primary element in the image is detected as the license plate, confirming the insurance type as car insurance. A watermark is generated: "License Plate - Chaoyang District, Beijing - Car Insurance." This watermark is added semi-transparently to the lower right corner of the image, ensuring it doesn't obscure key damaged areas.

[0102] By adding watermarks containing location, time, and type of insurance, false information or tampering is prevented, improving the authenticity and traceability of evidence. The item information is clearly readable from the image, facilitating quick confirmation and processing by reviewers. The clearly visible watermark increases user trust in the application while providing the system with verification references. Furthermore, in case of disputes or doubts, the detailed information in the watermark provides additional supporting evidence, helping to reduce potential fraud risks. This simplifies the review and verification process while ensuring information reliability and rapid retrieval.

[0103] Through the above embodiments, insurance companies or related organizations can introduce precise metadata and element descriptions at each stage of image information processing to prevent tampering, while simultaneously improving the transparency and security of data processing. This not only significantly optimizes internal operational processes but also provides strong protection in terms of privacy and compliance requirements.

[0104] In some embodiments, color overlay is applied to each pixel in the privacy area based on mosaic information, and the region outline is marked according to the identifier type to complete the editing of the document image, including: obtaining an initialized preset container; converting the document image into a preset document object model format; setting the converted document image as a background image in the preset container; color overlaying is applied to each pixel in the privacy area corresponding to the background image based on mosaic information in the preset container, and the region outline corresponding to the background image is marked according to the identifier type to complete the editing of the background image; and the edited background image is uploaded to the claims system.

[0105] First, a pre-defined container is created to hold and process the image. This container can be memory space within the image processing environment or a virtual canvas in which subsequent image processing operations are performed. The document image is then converted to a format suitable for the Document Object Model (DOM). This step involves transforming traditional image data into a more structured format for flexible subsequent operations. The DOM format facilitates access to and manipulation of different image elements, such as background images, foreground objects, and text, throughout the workflow. The converted document image is then set as the background image within the pre-defined container. This is equivalent to laying a base image on the canvas, providing a foundation for subsequent overlay and labeling operations. Mosaic information is then used to color-overlay pixels in the privacy regions of the background image. This typically involves applying blurring or pixelation to the privacy regions, interfering with their clarity and making them unrecognizable. This step protects privacy information without affecting the overall structural logic of the image.

[0106] Outline markings are applied to relevant areas in the background image based on the type of marker (e.g., circle, square, or arrow). Different marker types require different marking methods. Marking steps may include drawing boundary lines or color filling to enhance clarity of identification and information delivery. The edited background image is then uploaded to the relevant claims system for further review and processing.

[0107] If an insurance company needs to perform privacy processing on vehicle insurance claims documents and mark the type or location of damage, it can do so by creating a pre-defined processing container. The vehicle damage document image is converted to DOM format for easier manipulation. The image is set as the background image of the container. Mosaic processing is applied to license plates or personal information privacy areas using color overlay to ensure they are unrecognizable. Boundary markers are drawn for the damage locations on the image based on the damage type, for example, using a red circle to mark the impact area. After editing, it is uploaded to the internal claims system.

[0108] By employing color overlay and mosaic processing, privacy information within the image is effectively protected, enhancing the security of image data usage. Different label types are used to clearly mark areas, increasing information visualization and facilitating rapid communication and decision-making. Simultaneously, automated processes reduce manual intervention, improving image processing efficiency and making it suitable for batch document processing scenarios. By directly uploading processed images to the claims system, seamless integration from image acquisition to review is achieved, improving overall operational efficiency.

[0109] The above embodiments, through a reasonable image processing workflow combined with privacy protection and information tagging technologies, provide an efficient, secure, and intelligent method for image management and review in industries such as insurance, greatly improving ease of operation and user trust.

[0110] The method obtains the document images to be edited from a pre-set claims system. Based on the obtained insurance information, the document images are classified, and claim-related areas are identified. Claim-related areas are automatically detected within the image. Outside the claim-related areas, areas potentially containing private information (i.e., privacy areas) are identified. The outlines of the claim-related areas are then determined, and their region identifier types are defined, using identifier styles such as circles, squares, or arrows to mark the areas. Color distribution information within the privacy areas is obtained, and the color of each pixel is analyzed. Based on this information, a mosaic effect is generated to conceal the privacy information. The generated mosaic information is applied to the privacy areas, and color overlay processing is performed on each pixel. The final edited document image includes the identified claim-related areas and effectively masks the privacy information. In summary, this method improves image processing efficiency and privacy protection in the financial claims process through automated and intelligent image editing technology, making the overall operation smoother and more secure.

[0111] The method provided thus has the following beneficial effects:

[0112] 1. Improve work efficiency: Automated area detection and labeling reduce manual operations and improve the efficiency of claims review and processing.

[0113] 2. User experience optimization: The operation process has been simplified, eliminating the need for additional image editing software and significantly improving the operator's experience.

[0114] 3. Privacy Protection: Effectively protects sensitive information that may be involved in document images by automatically adding mosaic effects to privacy areas to ensure information security.

[0115] 4. Accuracy and consistency: Based on insurance type information and preset rules, relevant areas are detected and labeled, which improves the accuracy and consistency of image processing.

[0116] 5. Flexibility and scalability: It provides multiple identifier types, which makes the method highly flexible and able to adapt to different claims needs and insurance characteristics.

[0117] Please see Figure 4 As shown, Figure 4 This is a schematic diagram of the structure of an image editing device 200 provided in an embodiment of this application. The image editing device 200 is used to execute the steps of the image editing methods shown in the above embodiments. The image editing device 200 can be a single server or a server cluster, or it can be a terminal, such as a handheld terminal, a laptop computer, a wearable device, or a robot.

[0118] like Figure 4 As shown, the image editing device 200 includes:

[0119] The information acquisition unit 201 is used to acquire the document image to be edited and the insurance information corresponding to the document image from the preset claims system;

[0120] The region detection unit 202 is used to detect claim-related regions in the document image based on insurance type information;

[0121] Privacy detection unit 203 is used to detect privacy areas in document images outside the claim-related area;

[0122] The identifier determination unit 204 is used to determine the region identifier type based on the region outline corresponding to the claim-related region in the document image; wherein, the region identifier type includes at least one or more of the following: text identifier, circular identifier, square identifier, and arrow identifier;

[0123] The distribution determination unit 205 is used to obtain color distribution information corresponding to the privacy region in the document image; the color distribution information includes the color corresponding to each pixel in the privacy region;

[0124] The overlay determination unit 206 is used to generate mosaic information corresponding to the privacy region based on the color distribution information; the mosaic information includes the overlay color corresponding to each pixel in the privacy region.

[0125] The editing unit 207 is used to overlay the color of each pixel in the privacy area according to the mosaic information, mark the outline of the area according to the identification type, and complete the editing of the document image.

[0126] It should be noted that those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the image editing device and its modules described above can be referred to the corresponding processes in the image editing method embodiments described above, and will not be repeated here.

[0127] The image editing method described above can be implemented as a computer program, which can be used in, for example... Figure 4 It runs on the device shown.

[0128] Please see Figure 5 , Figure 5 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application. The computer device includes a processor, a memory, and a network interface connected via a device bus, wherein the memory may include a storage medium and internal memory.

[0129] The storage medium can store operating devices and computer programs. The computer program includes program instructions that, when executed, cause the processor to perform any image editing method.

[0130] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0131] Internal memory provides an environment for the execution of computer programs stored in non-volatile storage media. When these computer programs are executed by the processor, the processor can perform any image editing method.

[0132] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the terminal to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0133] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0134] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps:

[0135] Obtain the document image to be edited and the corresponding insurance information from the pre-set claims system;

[0136] Based on the insurance type information, detect the claim-related areas in the document images;

[0137] Privacy areas were detected in document images outside of the claims-related areas.

[0138] The region identifier type is determined based on the region outline corresponding to the relevant claim area in the document image; wherein, the region identifier type includes at least one or more of the following: circular identifier, square identifier, and arrow identifier;

[0139] Obtain the color distribution information corresponding to the privacy region in the document image; the color distribution information includes the color corresponding to each pixel in the privacy region.

[0140] Mosaic information corresponding to the privacy region is generated based on the color distribution information; the mosaic information includes the overlay color corresponding to each pixel in the privacy region.

[0141] Based on the mosaic information, each pixel in the privacy area is overlaid with color, and the region outline is marked according to the identifier type, thus completing the editing of the document image.

[0142] In some embodiments, detecting claim-related regions in a document image based on insurance type information includes: acquiring a pre-built image recognition model library, which includes multiple image recognition models and recognition categories corresponding to each image recognition model; matching the insurance type information with the recognition categories to acquire the image recognition model corresponding to the successfully matched recognition category; inputting the document image into the image recognition model, which then parses the document image and outputs the claim-related regions.

[0143] For example, before acquiring the pre-built image recognition model library, the method further includes: acquiring multiple historical document images of claims already processed and the corresponding insurance type information and historical claim area for each historical document image from the claims system; classifying the multiple historical document images according to the insurance type information to obtain multiple historical image groups, each historical image group including at least one historical document image; generating multiple image recognition models to be trained based on the number of historical image groups; inputting the historical document images of each historical image group into an image recognition model, the image recognition model parsing each historical document image and outputting the claim prediction area corresponding to each historical document image; training each corresponding image recognition model based on the claim prediction area and historical claim area corresponding to each historical document image; and generating the recognition category of the corresponding image recognition model based on the insurance type information corresponding to each historical image group.

[0144] In some embodiments, detecting a privacy region in a document image outside the claim-related area includes: performing optical character recognition in the other areas outside the claim-related area to obtain text information corresponding to the document image and location information corresponding to the text information; parsing the text information based on natural language processing technology to obtain semantic information corresponding to the text information; and matching the semantic information with a preset privacy type to generate a privacy region from the location information corresponding to the successfully matched semantic information.

[0145] In some embodiments, determining the region identifier type based on the region contour corresponding to the claim-related region in the document image includes: parsing the region contour and obtaining the geometric feature information corresponding to the region contour; the geometric feature information includes at least one or more of aspect ratio, roundness, and corner features; if the regularity of the region contour is determined to be greater than a preset regularity based on the geometric feature information, the identifier type is determined to be a circular identifier; if the regularity of the region contour is determined to be less than or equal to the preset regularity based on the geometric feature information, the identifier type is determined to be a square identifier; if the area of ​​the region contour is determined to be less than a preset area based on the geometric feature information, the identifier type is determined to be an arrow identifier.

[0146] In some embodiments, before editing the document image, the method further includes: obtaining the shooting location information corresponding to the document image; detecting the claim-related area and obtaining the element type corresponding to the claim-related area; generating watermark information based on the element type, shooting location information and insurance type information; adding the watermark information to the document image; and completing the editing of the document image.

[0147] In some embodiments, color overlay is applied to each pixel in the privacy area based on mosaic information, and the region outline is marked according to the identifier type to complete the editing of the document image, including: obtaining an initialized preset container; converting the document image into a preset document object model format; setting the converted document image as a background image in the preset container; color overlaying is applied to each pixel in the privacy area corresponding to the background image based on mosaic information in the preset container, and the region outline corresponding to the background image is marked according to the identifier type to complete the editing of the background image; and the edited background image is uploaded to the claims system.

[0148] The embodiments of this application also provide a computer-readable storage medium storing a computer program, the computer program including program instructions, and the processor executing the program instructions to implement the steps of the image editing method provided in the above embodiments of this application.

[0149] The computer-readable storage medium can be an internal storage unit of the computer device described in the foregoing embodiments, such as the hard disk or memory of the computer device. The computer-readable storage medium can also be an external storage device of the computer device, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device.

[0150] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An image editing method, characterized in that, include: Obtain the document image to be edited and the insurance type information corresponding to the document image from the preset claims system; Detecting the claim-related area in the document image based on the insurance type information includes: acquiring a pre-built image recognition model library, which includes multiple image recognition models and a recognition category corresponding to each image recognition model; matching the insurance type information with the recognition category to acquire the image recognition model corresponding to the successfully matched recognition category; inputting the document image into the image recognition model, which parses the document image and outputs the claim-related area. Privacy areas were detected in the document image outside the claims-related area. The region identifier type is determined based on the region outline corresponding to the claim-related region in the document image; wherein, the region identifier type includes at least a circular identifier, a square identifier, and an arrow identifier; Obtain color distribution information corresponding to the privacy region from the document image; the color distribution information includes the color corresponding to each pixel in the privacy region. Mosaic information corresponding to the privacy region is generated based on the color distribution information; the mosaic information includes the overlay color corresponding to each pixel in the privacy region. The color of each pixel in the privacy area is overlaid according to the mosaic information, and the outline of the area is marked according to the identifier type, thus completing the editing of the document image; The step of determining the region identifier type based on the region contour corresponding to the claim-related region in the document image includes: parsing the region contour to obtain the geometric feature information corresponding to the region contour; the geometric feature information includes at least one or more of aspect ratio, roundness, and corner features; if the regularity of the region contour is determined to be greater than a preset regularity based on the geometric feature information, the identifier type is determined to be the circular identifier; if the regularity of the region contour is determined to be less than or equal to the preset regularity based on the geometric feature information, the identifier type is determined to be the square identifier; if the area of ​​the region contour is determined to be less than a preset area based on the geometric feature information, the identifier type is determined to be the arrow identifier.

2. The method according to claim 1, characterized in that, Before obtaining the pre-built image recognition model library, the following is also included: The claims system retrieves multiple historical document images of claims that have been processed, as well as the insurance type information and historical claims area corresponding to each historical document image. Based on the insurance type information, multiple historical document images are classified to obtain multiple historical image groups, each of which includes at least one historical document image. Multiple image recognition models to be trained are generated based on the number of historical image groups; Each of the historical document images in each of the historical image groups is input into an image recognition model. The image recognition model parses each of the historical document images and outputs the claim prediction area corresponding to each of the historical document images. Training of each image recognition model is completed based on the claim prediction area and historical claim area corresponding to each historical document image; The recognition category of the corresponding image recognition model is generated based on the insurance information corresponding to each of the historical image groups.

3. The method according to claim 1, characterized in that, The detection of privacy areas in the document image outside the claim-related area includes: Optical character recognition is performed in areas other than the claims-related area to obtain the text information corresponding to the document image and the location information corresponding to the text information. The text information is parsed using natural language processing technology to obtain the semantic information corresponding to the text information; The privacy region is generated by matching the semantic information with the preset privacy type and the location information corresponding to the successfully matched semantic information.

4. The method according to claim 1, characterized in that, Before completing the editing of the document image, the process also includes: Obtain the shooting location information corresponding to the document image; The claim-related area is detected to obtain the element type corresponding to the claim-related area; Watermark information is generated based on the element type, shooting location information, and insurance type information. The watermark information is then added to the document image to complete the editing of the document image.

5. The method according to claim 1, characterized in that, The step of color-overlaying each pixel in the privacy region based on the mosaic information, marking the region outline according to the identifier type, and completing the editing of the document image includes: Get the initial preset container; Convert the document image into a preset document object model format; The converted document image is set as the background image in the preset container; In the preset container, each pixel in the privacy area corresponding to the background image is color-overlaid according to the mosaic information, and the outline of the area corresponding to the background image is marked according to the identifier type to complete the editing of the background image. The edited background image is then uploaded to the claims system.

6. An image editing device, characterized in that, include: The information acquisition unit is used to acquire the document image to be edited and the insurance type information corresponding to the document image from the preset claims system; A region detection unit is used to detect claim-related regions in the document image based on the insurance type information, including: acquiring a pre-built image recognition model library, the image recognition model library including multiple image recognition models and a recognition category corresponding to each image recognition model; matching the insurance type information with the recognition category to acquire the image recognition model corresponding to the successfully matched recognition category; inputting the document image into the image recognition model, the image recognition model parsing the document image and outputting the claim-related region; A privacy detection unit is used to detect privacy areas in the document image outside the claim-related area; The identifier determination unit is used to determine the region identifier type based on the region outline corresponding to the claim-related region in the document image; wherein, the region identifier type includes at least text identifier, circular identifier, square identifier, and arrow identifier; A distribution determination unit is used to obtain color distribution information corresponding to the privacy region in the document image; the color distribution information includes the color corresponding to each pixel in the privacy region; The overlay determination unit is used to generate mosaic information corresponding to the privacy region based on the color distribution information; the mosaic information includes the overlay color corresponding to each pixel in the privacy region. The editing completion unit is used to overlay the color of each pixel in the privacy area according to the mosaic information, mark the outline of the area according to the identifier type, and complete the editing of the document image; The step of determining the region identifier type based on the region contour corresponding to the claim-related region in the document image includes: parsing the region contour to obtain the geometric feature information corresponding to the region contour; the geometric feature information includes at least one or more of aspect ratio, roundness, and corner features; if the regularity of the region contour is determined to be greater than a preset regularity based on the geometric feature information, the identifier type is determined to be the circular identifier; if the regularity of the region contour is determined to be less than or equal to the preset regularity based on the geometric feature information, the identifier type is determined to be the square identifier; if the area of ​​the region contour is determined to be less than a preset area based on the geometric feature information, the identifier type is determined to be the arrow identifier.

7. A computer device, characterized in that, The computer device includes a memory and a processor; The memory is used to store computer programs; The processor is configured to execute the computer program and, in executing the computer program, implement the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, causes the processor to implement the method as described in any one of claims 1 to 5.

Images