AI-based intelligent image preprocessing method and system

Abstract

The application discloses an AI-based intelligent image preprocessing method and system, which comprises the following steps: obtaining image data to be processed, separating a main image area and a background area, generating a main area mask and a background area mask, and determining an image type label; based on the main area mask, the main image is divided into sub-areas with independent geometric features, and pixel-level image preprocessing is performed on each sub-area; by calculating the visual correlation weight between the sub-areas, a sub-area correlation graph is constructed and feature propagation is performed to obtain consistency correction parameters between the sub-areas; the pixels of each sub-area are processed to make the adjacent areas smoothly transition, and finally the optimized main image is synthesized; based on the background area mask, the background color temperature of the optimized main image is migrated, and the background blurriness is adjusted. The application has the advantages that by means of image segmentation, sub-area division and graph convolution network feature propagation, the image quality is optimized, and the natural transition between the main area and the background of the image is realized.

Description

Technical Field

[0001] This invention relates to AI image processing, and particularly to an AI-based intelligent image preprocessing method and system. Background Technology

[0002] With the widespread application of digital images, especially in fields such as medical imaging, satellite remote sensing, security monitoring, and autonomous driving, image preprocessing technology has received considerable attention. In medical image processing, image preprocessing helps doctors see lesions more clearly; in autonomous driving, preprocessing can enhance image data acquired by cameras, improving recognition accuracy. The purpose of image preprocessing is to improve the analyzability and quality of images, ensuring the effectiveness and accuracy of subsequent algorithms such as pattern recognition and machine learning, making it an indispensable technology in modern intelligent systems.

[0003] Current image preprocessing methods and systems rely on manually designed feature extraction and simple image processing techniques, such as fixed filtering algorithms, color space conversion, and noise suppression. These methods struggle to adapt to complex image content and cannot dynamically adjust processing strategies to handle different image features. Particularly in image region segmentation, traditional methods often use thresholding or edge detection-based algorithms, which are easily affected by image noise and complex backgrounds, leading to insufficient segmentation accuracy. When dealing with the visual correlation and consistency between different sub-regions, traditional methods typically lack a global perspective, failing to consider the mutual influence and overall structure between sub-regions. This results in inconsistent image processing effects and loss of detail, especially in complex image scenes where boundary transitions appear abrupt and are difficult to achieve smoothly. Summary of the Invention

[0004] To improve existing methods and systems, this paper provides an AI-based intelligent image preprocessing method and system. This method achieves high-quality image optimization through precise image region separation, pixel-level optimization processing, and consistency correction, improving image details, clarity, and natural transitions. It is suitable for scenarios requiring high-precision image processing.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] AI-based intelligent image preprocessing methods include:

[0007] The process involves acquiring image data to be processed, inputting pre-trained key information to extract a neural network model, separating the main image region and background image region in the image, generating main region masks and background region masks, and obtaining image type labels.

[0008] Based on the main region mask, the main image is divided into geometrically independent sub-regions, each containing a continuous set of pixels;

[0009] Perform pixel-level image processing operations on each sub-region, including adaptive filtering, color space conversion, and noise suppression;

[0010] Based on image type labels, the visual association weights between sub-regions are calculated to obtain the image association of each sub-region. A sub-region association graph is constructed, and feature propagation is performed on the sub-region association graph through a graph convolutional network to obtain consistency correction parameters between sub-regions.

[0011] The pixels of each sub-region are processed based on the consistency correction parameter to make the transition between adjacent sub-regions smooth. The processed sub-regions are then combined to obtain the optimized main image.

[0012] Based on the background region mask, the color histogram of the optimized main image is obtained for background color temperature shifting, and the background blur is adjusted according to the boundary distance of the main region.

[0013] Preferably, the steps of acquiring the image data to be processed, inputting pre-trained key information to extract a neural network model, separating the main image region and background image region in the image, generating a main region mask and a background region mask, and acquiring the image type label specifically include:

[0014] The image data to be processed is acquired and input into a pre-trained key information extraction neural network model for forward propagation.

[0015] Image segmentation is performed using a key information extraction neural network model to obtain the main image region and background image region of the image;

[0016] Based on the output image segmentation results, a main region mask is generated based on the main image region, and a background region mask is generated based on the background image region. The pixel value of the main region is 1, and the pixel value of the background region is 0.

[0017] Based on a key information extraction neural network model, image classification is performed simultaneously with image segmentation, and type labels are added to each image.

[0018] Preferably, the step of dividing the main image into geometrically independent sub-regions based on the main region mask, with each sub-region containing a continuous set of pixels, specifically includes:

[0019] Connectivity analysis is performed based on the obtained main region mask, and the continuous pixel portion in the main region mask is obtained according to the 8-neighbor connectivity.

[0020] Independence detection is performed on all the obtained connected components to divide the region into sub-regions, and the shape features of each sub-region are obtained through geometric feature extraction.

[0021] Regions are merged and segmented based on the shape features of each sub-region, thereby enhancing the independence of the shape features of each sub-region;

[0022] Based on the segmented sub-regions, obtain the set of continuous pixels within each sub-region.

[0023] Preferably, the pixel-level image processing operations performed on each sub-region, including adaptive filtering, color space conversion, and noise suppression, specifically include:

[0024] The appropriate filter window size is obtained based on the local characteristics of each sub-region, and the statistical characteristics of the sub-region are calculated in the neighborhood of each pixel, and the filter parameters are dynamically adjusted.

[0025] Based on the acquired local characteristics of each sub-region, color space conversion processing is performed to convert the color information of the sub-region from the original color space to the target color space, including RGB, HSV, and Lab.

[0026] Various types of noise are suppressed by mean filtering, Gaussian filtering, and median filtering.

[0027] Preferably, the step of calculating the visual association weights between sub-regions based on image type labels, obtaining the image association of each sub-region, constructing a sub-region association graph, and performing feature propagation on the sub-region association graph through a graph convolutional network to obtain consistency correction parameters between sub-regions specifically includes:

[0028] Based on the image type labels obtained by the key information extraction neural network model, feature extraction is performed on the geometric and visual features of each sub-region.

[0029] The similarity between sub-regions is calculated based on the geometric and visual features of the acquired sub-regions, and the association weight is set based on the similarity.

[0030] Based on the image correlation of each sub-region, a sub-region correlation graph is constructed. Each sub-region is regarded as a node in the graph, the visual correlation between sub-regions is regarded as the edge between nodes, and the correlation weight is regarded as the weight of the edge.

[0031] Sub-region feature propagation is performed based on graph convolutional networks to enhance the expression of each node's relationship with its surrounding environment and obtain the final representation of each sub-region;

[0032] Based on the final representation of each sub-region, the differences between the features of the sub-regions are calculated to obtain the consistency correction parameter.

[0033] Preferably, the step of processing the pixels of each sub-region based on the consistency correction parameter to achieve a smooth transition between adjacent sub-regions, and then combining the processed sub-regions to obtain the optimized main image specifically includes:

[0034] Based on the obtained consistency correction parameters, the pixel features of each sub-region are adjusted, including fine-tuning the color and shape geometry of the sub-region boundaries;

[0035] Based on the adjusted sub-regions, Gaussian smoothing is used to smoothly transition between adjacent sub-regions;

[0036] The sub-regions after smooth transition processing are merged, and the optimized main image region is obtained through global consistency adjustment.

[0037] Preferably, the step of obtaining the color histogram of the optimized main image based on the background region mask, performing background color temperature shifting, and adjusting the background blur according to the boundary distance of the main region specifically includes:

[0038] Based on the background region, obtain the color histogram of the optimized main image, and perform color temperature shifting on the background region by adjusting the color channels of the background region based on the color histogram;

[0039] The pixel blur is defined by calculating the distance between each pixel in the background region and the boundary of the main region, and the background region is then blurred.

[0040] The background area, after color temperature shifting and blur adjustment, is merged with the main image area to obtain a complete image.

[0041] Furthermore, an AI-based intelligent image preprocessing system is proposed, including:

[0042] Image preprocessing module: The image preprocessing module is used to input the image data to be processed into a pre-trained key information extraction neural network model, separate the main image region and background region in the image, generate a mask and perform image classification;

[0043] Region segmentation module: The region segmentation module is based on the main region mask. This module performs connected component analysis to divide the main image region into geometrically independent sub-regions and extracts the shape features of each sub-region.

[0044] Image processing module: The image processing module performs pixel-level processing on each sub-region, including adaptive filtering, color space conversion and noise suppression, to optimize the local characteristics of the sub-region;

[0045] The correlation calculation module obtains the correlation weights between each sub-region, constructs a sub-region correlation graph and performs feature propagation, and obtains consistency correction parameters between sub-regions.

[0046] Sub-region synthesis module: The sub-region synthesis module adjusts the pixels of the sub-region based on the consistency correction parameter and performs smooth transition processing to finally synthesize the optimized main image region;

[0047] Background area adjustment and image synthesis module: The background area adjustment and image synthesis module obtains the optimized main image color histogram based on the background area mask, performs background color temperature shifting and adjusts the background blur, and finally synthesizes a complete image;

[0048] Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

[0049] Compared with the prior art, the advantages of the present invention are:

[0050] By combining a pre-trained key information extraction neural network model, the main image region and background region can be accurately separated, providing a clear image structure for subsequent processing. The sub-region division and pixel-level processing operations in this method allow each sub-region to be optimized independently, exhibiting strong adaptability and improving the clarity and realism of image details. Graph convolutional networks propagate visual relationships between sub-regions, ensuring seamless connections between regions and avoiding the unnatural transitions common in traditional methods. Furthermore, background region color temperature shifting and blur adjustment make the optimized image more natural and realistic, with higher visual perception quality. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the AI-based intelligent image preprocessing method proposed in this invention;

[0052] Figure 2 This is a schematic diagram of the region mask proposed in this invention;

[0053] Figure 3 This is a schematic diagram of the sub-region division proposed in this invention;

[0054] Figure 4 This is a schematic diagram of the sub-region image processing proposed in this invention;

[0055] Figure 5 This is a schematic diagram of the consistency correction parameters proposed in this invention;

[0056] Figure 6 This is a schematic diagram of the optimized main images proposed in this invention;

[0057] Figure 7 This is a schematic diagram of the background area processing and merging proposed in this invention. Detailed Implementation

[0058] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0059] AI-based intelligent image preprocessing systems include:

[0060] Image preprocessing module: The image preprocessing module is used to input the image data to be processed into a pre-trained key information extraction neural network model, separate the main image region and background region in the image, generate a mask and perform image classification;

[0061] Region segmentation module: The region segmentation module is based on the main region mask. This module performs connected component analysis to divide the main image region into geometrically independent sub-regions and extracts the shape features of each sub-region.

[0062] Image processing module: The image processing module performs pixel-level processing on each sub-region, including adaptive filtering, color space conversion and noise suppression, to optimize the local characteristics of the sub-region;

[0063] The correlation calculation module obtains the correlation weights between each sub-region, constructs a sub-region correlation graph and performs feature propagation, and obtains consistency correction parameters between sub-regions.

[0064] Sub-region synthesis module: The sub-region synthesis module adjusts the pixels of the sub-region based on the consistency correction parameter and performs smooth transition processing to finally synthesize the optimized main image region;

[0065] Background area adjustment and image synthesis module: The background area adjustment and image synthesis module obtains the optimized main image color histogram based on the background area mask, performs background color temperature shifting and adjusts the background blur, and finally synthesizes a complete image;

[0066] Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

[0067] See Figure 1 As shown, the AI-based intelligent image preprocessing method includes:

[0068] Step 1: Obtain the image data to be processed, input the key information of the pre-trained neural network model, separate the main image region and the background image region in the image, generate the main region mask and the background region mask, and obtain the image type label;

[0069] Step 2: Based on the main region mask, divide the main image into geometrically independent sub-regions, each containing a continuous set of pixels;

[0070] Step 3: Perform pixel-level image processing operations on each sub-region, including adaptive filtering, color space conversion, and noise suppression;

[0071] Step 4: Based on image type labels, calculate the visual association weights between sub-regions, obtain the image association of each sub-region, construct a sub-region association graph, and perform feature propagation on the sub-region association graph through a graph convolutional network to obtain consistency correction parameters between sub-regions;

[0072] Step 5: Process the pixels of each sub-region based on the consistency correction parameter to make the transition between adjacent sub-regions smooth, and combine the processed sub-regions to obtain the optimized main image;

[0073] Step 6: Based on the background region mask, obtain the color histogram of the optimized main image, perform background color temperature shifting, and adjust the background blur according to the boundary distance of the main region.

[0074] See Figure 2 As shown, the process involves acquiring the image data to be processed, inputting pre-trained key information to extract a neural network model, separating the main image region and background image region in the image, generating main region masks and background region masks, and obtaining image type labels, specifically including:

[0075] The image data to be processed is acquired and input into a pre-trained key information extraction neural network model for forward propagation.

[0076] Image segmentation is performed using a key information extraction neural network model to obtain the main image region and background image region of the image;

[0077] Based on the output image segmentation results, a main region mask is generated based on the main image region, and a background region mask is generated based on the background image region. The pixel value of the main region is 1, and the pixel value of the background region is 0.

[0078] Based on a key information extraction neural network model, image classification is performed simultaneously with image segmentation, and type labels are added to each image.

[0079] Specifically, the image data to be processed is loaded and converted into a tensor format suitable for neural network input. The images may come from files, cameras, or data streams and need to be preprocessed to match the training data distribution of the model.

[0080] Image tensors are input into a pre-trained neural network model to extract key information. Forward propagation is then performed, and the model simultaneously handles segmentation and classification tasks, outputting a segmentation probability map and a classification probability vector. Based on the output segmentation probability map, a threshold is applied to classify the probabilities. Figure 2 Value-based quantization separates the main image region from the background image region, and thresholding is used to determine pixel assignment.

[0081] Based on the segmentation results, a binary mask tensor is generated. In the main region mask, the pixel value of the main region is 1, and the pixel value of the background region is 0. In the background region mask, the pixel value of the background region is 1, and the pixel value of the main region is 0. The background region mask can be directly derived from the main region mask.

[0082] During forward propagation, the model simultaneously outputs a classification probability vector. Based on this vector, the predicted category of the image is determined by taking the maximum index, and a type label is added.

[0083] See Figure 3 As shown, based on the main region mask, the main image is divided into geometrically independent sub-regions, each containing a continuous set of pixels, specifically including:

[0084] Connectivity analysis is performed based on the obtained main region mask, and the continuous pixel portion in the main region mask is obtained according to the 8-neighbor connectivity.

[0085] Independence detection is performed on all the obtained connected components to divide the region into sub-regions, and the shape features of each sub-region are obtained through geometric feature extraction.

[0086] Regions are merged and segmented based on the shape features of each sub-region, thereby enhancing the independence of the shape features of each sub-region;

[0087] Based on the segmented sub-regions, obtain the set of continuous pixels within each sub-region.

[0088] Specifically, given pixel position Its 8-neighborhood is defined as: ;

[0089] For each pixel in the mask image, if its value is 1 and it has not been labeled, a breadth-first search or depth-first search is performed starting from that pixel, and all connected pixels are labeled as the same connected component. Each connected component will be assigned a unique label.

[0090] The goal of independence detection is to determine the boundaries between connected components and ensure that they do not overlap or touch. The detection method can be achieved by checking the adjacency of each connected component. If the boundary of one connected component touches or overlaps with the boundary of another connected component, then they belong to the same group of regions.

[0091] The purpose of geometric feature extraction is to analyze the properties of sub-regions by calculating their shape features. Common shape features include area, perimeter, shape moments, and aspect ratio.

[0092] Based on geometric features, clustering algorithms are used to divide connected regions, ensuring that regions with similar shapes are grouped together. For example, if two sub-regions have similar aspect ratios and areas, they may belong to the same type of region.

[0093] If two or more sub-regions are geometrically very close and are considered to belong to the same object through independence detection, they can be merged into a larger region. The area, perimeter, and other shape features of the merged region will be updated based on its constituent parts. If a sub-region is too complex or contains multiple objects, it can be further divided into multiple sub-regions using shape features, using bounding boxes and shape moment features to determine the appropriate segmentation method.

[0094] Based on the segmented sub-regions, obtain the set of continuous pixels within each sub-region. For each sub-region, find its connected set of pixels, so that these pixels remain connected in space.

[0095] See Figure 4 As shown, pixel-level image processing operations are performed on each sub-region, including adaptive filtering, color space conversion, and noise suppression. Specifically, these include:

[0096] The appropriate filter window size is obtained based on the local characteristics of each sub-region, and the statistical characteristics of the sub-region are calculated in the neighborhood of each pixel, and the filter parameters are dynamically adjusted.

[0097] Based on the acquired local characteristics of each sub-region, color space conversion processing is performed to convert the color information of the sub-region from the original color space to the target color space, including RGB, HSV, and Lab.

[0098] Various types of noise are suppressed by mean filtering, Gaussian filtering, and median filtering.

[0099] See Figure 5 As shown, based on image type labels, visual association weights between sub-regions are calculated to obtain the image association between each sub-region, and a sub-region association graph is constructed. A graph convolutional network is then used to propagate features from the sub-region association graph to obtain consistency correction parameters between sub-regions. Specifically, these parameters include:

[0100] Based on the image type labels obtained by the key information extraction neural network model, feature extraction is performed on the geometric and visual features of each sub-region.

[0101] The similarity between sub-regions is calculated based on the geometric and visual features of the acquired sub-regions, and the association weight is set based on the similarity.

[0102] Based on the image correlation of each sub-region, a sub-region correlation graph is constructed. Each sub-region is regarded as a node in the graph, the visual correlation between sub-regions is regarded as the edge between nodes, and the correlation weight is regarded as the weight of the edge.

[0103] Sub-region feature propagation is performed based on graph convolutional networks to enhance the expression of each node's relationship with its surrounding environment and obtain the final representation of each sub-region;

[0104] Based on the final representation of each sub-region, the differences between the features of the sub-regions are calculated to obtain the consistency correction parameter.

[0105] Specifically, after extracting geometric and visual features, we can calculate the similarity between each sub-region. We use cosine similarity to measure the similarity of visual and geometric features between sub-regions. Based on the calculated similarity, we set association weights to represent the visual association between sub-regions. If the similarity between sub-regions is high, the association weight will be large, indicating that they have a strong connection on the edges of the graph.

[0106] Based on the similarity and association weights calculated above, an association graph of sub-regions is constructed. Each sub-region in the graph is a node, and the edges between nodes represent the association relationships between sub-regions. A weighted graph can be constructed.

[0107] Graph Convolutional Networks (GCNs) are used to propagate and aggregate node features in a graph. By propagating node features through multiple layers of convolution operations, the features of each sub-region not only contain its own information but also the information of its neighboring sub-regions. Finally, after propagation through multiple layers of GCNs, the final representation of each sub-region is obtained.

[0108] The consistency correction parameter is obtained by calculating the difference in the final feature representations between sub-regions. The difference between the features of sub-regions can be calculated using Euclidean distance.

[0109] See Figure 6 As shown, the pixels of each sub-region are processed based on the consistency correction parameter to achieve a smooth transition between adjacent sub-regions. The processed sub-regions are then combined to obtain the optimized main image. Specifically, this includes:

[0110] Based on the obtained consistency correction parameters, the pixel features of each sub-region are adjusted, including fine-tuning the color and shape geometry of the sub-region boundaries;

[0111] Based on the adjusted sub-regions, Gaussian smoothing is used to smoothly transition between adjacent sub-regions;

[0112] The sub-regions after smooth transition processing are merged, and the optimized main image region is obtained through global consistency adjustment.

[0113] Specifically, based on the previously calculated consistency correction parameters between sub-regions, we adjust the pixel features of each sub-region. These adjustments include fine-tuning the color and shape geometry of the sub-region boundaries. For color fine-tuning, the consistency correction parameters can be adjusted using the following formula:

[0114]

[0115] in, sub-region The color, The color after fine-tuning. The weighting coefficient represents the degree of influence of the consistency correction parameter on color adjustment. sub-region Consistency correction parameters;

[0116] For each pixel in the sub-region, adjustments are made based on the corrected geometric and visual features;

[0117] After completing the consistency correction, we need to smooth the transition between adjacent sub-regions to ensure that the boundaries and transition parts of the image are more natural. Gaussian smoothing is a common smoothing technique used to smooth image or pixel data, reduce noise and optimize the transition effect.

[0118] After the smooth transition, we will merge the sub-regions based on global consistency adjustment to obtain the optimized main image region. The sub-regions are merged based on the following criteria, including geometric consistency, visual similarity and similarity measurement.

[0119] Global consistency adjustment ensures that merging all sub-regions does not lead to the loss or distortion of important features in the image, resulting in an optimized main image region with higher visual consistency, geometric accuracy, and spatial coherence.

[0120] See Figure 7 As shown, based on the background region mask, the color histogram of the optimized main image is obtained for background color temperature shifting, and the background blur is adjusted according to the boundary distance of the main region. Specifically, this includes:

[0121] Based on the background region, obtain the color histogram of the optimized main image, and perform color temperature shifting on the background region by adjusting the color channels of the background region based on the color histogram;

[0122] The pixel blur is defined by calculating the distance between each pixel in the background region and the boundary of the main region, and the background region is then blurred.

[0123] The background area, after color temperature shifting and blur adjustment, is merged with the main image area to obtain a complete image.

[0124] Specifically, color channel adjustments are made by calculating the color histogram difference between the background region and the main region; the blur of the background region is defined by calculating the distance between each pixel and the boundary of the main region. For each pixel in the background region, its distance to the boundary of the main region is calculated, and the blur of the pixel is defined by a Gaussian function as a distance-weighted blur, with pixels that are farther away having smaller blur.

[0125] After color temperature shifting and blurring, we merge the adjusted background area with the main area to obtain a complete image.

[0126] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0127] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0128] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An AI-based intelligent image preprocessing method, characterized in that, include: The process involves acquiring image data to be processed, inputting pre-trained key information to extract a neural network model, separating the main image region and background image region in the image, generating main region masks and background region masks, and obtaining image type labels. Based on the main region mask, the main image is divided into geometrically independent sub-regions, each containing a continuous set of pixels; Perform pixel-level image processing operations on each sub-region, including adaptive filtering, color space conversion, and noise suppression; Based on the image type labels obtained by the key information extraction neural network model, feature extraction is performed on the geometric and visual features of each sub-region. The similarity between sub-regions is calculated based on the geometric and visual features of the acquired sub-regions, and the association weight is set based on the similarity. Based on the image correlation of each sub-region, a sub-region correlation graph is constructed. Each sub-region is regarded as a node in the graph, the visual correlation between sub-regions is regarded as the edge between nodes, and the correlation weight is regarded as the weight of the edge. Sub-region feature propagation is performed based on graph convolutional networks to enhance the expression of each node's relationship with its surrounding environment and obtain the final representation of each sub-region; Based on the final representation of each sub-region, the differences between the features of the sub-regions are calculated to obtain the consistency correction parameters; The pixels of each sub-region are processed based on the consistency correction parameter to make the transition between adjacent sub-regions smooth. The processed sub-regions are then combined to obtain the optimized main image. Based on the background region mask, the color histogram of the optimized main image is obtained for background color temperature shifting, and the background blur is adjusted according to the boundary distance of the main region.

2. The AI-based intelligent image preprocessing method according to claim 1, characterized in that, The process of acquiring the image data to be processed, inputting pre-trained key information to extract a neural network model, separating the main image region and background image region in the image, generating a main region mask and a background region mask, and obtaining the image type label specifically includes: The image data to be processed is acquired and input into a pre-trained key information extraction neural network model for forward propagation. Image segmentation is performed using a key information extraction neural network model to obtain the main image region and background image region of the image; Based on the output image segmentation results, a main region mask is generated based on the main image region, and a background region mask is generated based on the background image region. The pixel value of the main region is 1, and the pixel value of the background region is 0. Based on a key information extraction neural network model, image classification is performed simultaneously with image segmentation, and type labels are added to each image.

3. The AI-based intelligent image preprocessing method according to claim 1, characterized in that, The process of dividing the main image into geometrically independent sub-regions based on the main region mask, with each sub-region containing a continuous set of pixels, specifically includes: Connectivity analysis is performed based on the obtained main region mask, and the continuous pixel portion in the main region mask is obtained according to the 8-neighbor connectivity. Independence detection is performed on all the obtained connected components to divide the region into sub-regions, and the shape features of each sub-region are obtained through geometric feature extraction. Regions are merged and segmented based on the shape features of each sub-region, thereby enhancing the independence of the shape features of each sub-region; Based on the segmented sub-regions, obtain the set of continuous pixels within each sub-region.

4. The AI-based intelligent image preprocessing method according to claim 1, characterized in that, The pixel-level image processing operations performed on each sub-region, including adaptive filtering, color space conversion, and noise suppression, specifically include: The appropriate filter window size is obtained based on the local characteristics of each sub-region, and the statistical characteristics of the sub-region are calculated in the neighborhood of each pixel, and the filter parameters are dynamically adjusted. Based on the acquired local characteristics of each sub-region, color space conversion processing is performed to convert the color information of the sub-region from the original color space to the target color space, including RGB, HSV, and Lab. Various types of noise are suppressed by mean filtering, Gaussian filtering, and median filtering.

5. The AI-based intelligent image preprocessing method according to claim 1, characterized in that, The process of processing pixels in each sub-region based on consistency correction parameters to achieve a smooth transition between adjacent sub-regions, and then combining the processed sub-regions to obtain the optimized main image, specifically includes: Based on the obtained consistency correction parameters, the pixel features of each sub-region are adjusted, including fine-tuning the color and shape geometry of the sub-region boundaries; Based on the adjusted sub-regions, Gaussian smoothing is used to smoothly transition between adjacent sub-regions; The sub-regions after smooth transition processing are merged, and the optimized main image region is obtained through global consistency adjustment.

6. The AI-based intelligent image preprocessing method according to claim 1, characterized in that, The process of obtaining the optimized color histogram of the main image based on the background region mask, performing background color temperature shifting, and adjusting the background blur according to the boundary distance of the main region specifically includes: Based on the background region, obtain the color histogram of the optimized main image, and perform color temperature shifting on the background region by adjusting the color channels of the background region based on the color histogram; The pixel blur is defined by calculating the distance between each pixel in the background region and the boundary of the main region, and the background region is then blurred. The background area, after color temperature shifting and blur adjustment, is merged with the main image area to obtain a complete image.

7. An AI-based intelligent image preprocessing system, used to implement the AI-based intelligent image preprocessing method as described in any one of claims 1-6, characterized in that, include: Image preprocessing module: The image preprocessing module is used to input the image data to be processed into a pre-trained key information extraction neural network model, separate the main image region and background region in the image, generate a mask and perform image classification; Region segmentation module: The region segmentation module is based on the main region mask. This module performs connected component analysis to divide the main image region into geometrically independent sub-regions and extracts the shape features of each sub-region. Image processing module: The image processing module performs pixel-level processing on each sub-region, including adaptive filtering, color space conversion and noise suppression, to optimize the local characteristics of the sub-region; The correlation calculation module obtains the correlation weights between each sub-region, constructs a sub-region correlation graph and performs feature propagation, and obtains consistency correction parameters between sub-regions. Sub-region synthesis module: The sub-region synthesis module adjusts the pixels of the sub-region based on the consistency correction parameter and performs smooth transition processing to finally synthesize the optimized main image region; Background area adjustment and image synthesis module: The background area adjustment and image synthesis module obtains the optimized main image color histogram based on the background area mask, performs background color temperature shifting and adjusts the background blur, and finally synthesizes a complete image; Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

Images