Color table extraction neural network model construction, color table extraction method and system

Abstract

The present application provides a color table extraction neural network model construction, a color table extraction method and system, belonging to the technical field of computer image processing and data visualization, comprising: obtaining a single scalar field visualization image; inputting the single scalar field visualization image into a pre-trained color table extraction neural network model, the color table extraction neural network model processing the input image to extract scalar data and a color control point sequence; interpolating the color control point sequence to generate a continuous color table, and then mapping the continuous color table and the extracted scalar data to generate a reconstructed image; performing self-supervised fine-tuning in the inference stage of the color table extraction neural network model to extract the final continuous color table.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of computer image processing and data visualization, and particularly relates to a color table extraction neural network model construction, a color table extraction method and a system. BACKGROUND

[0002] The statements in this section merely provide background information related to the present application and do not necessarily constitute prior art.

[0003] A continuous color map refers to a color mapping scheme used for visualizing continuous data, which maps a numerical range to a series of smoothly transitioning colors. In scientific visualization and image analysis, continuous color maps are widely used in the visualization of scalar field data, mapping different colors to different data values to make the spatial distribution characteristics and numerical changes of data more intuitive.

[0004] However, a large number of visualization images (such as academic publications, web screenshots, report charts, etc.) in practice lack clear legends or color map information, making it difficult for machines or humans to accurately understand, reuse or redesign the data information in these images.

[0005] Among them, the legend is a key component in data visualization to explain the meaning of graphical elements, and the legend design can significantly improve the readability of the chart. Existing color map extraction methods rely on legends to map colors and data.

[0006] In the prior art, color and data mapping is often performed in the process of implementing color table extraction, and OCR (Optical Character Recognition) and image processing techniques are often combined to obtain color-value pairs in the legend. The color-value pairs in the legend are important elements in visualization for explaining the correspondence between color coding and data values. However, in the absence of legends, such methods cannot be applied, and the extracted color maps lack accuracy and interpretability.

[0007] In addition, some supervised learning methods rely on paired training samples (image and color map pairs) for model training, which have weak generalization ability and are difficult to adapt to real scenarios with diverse image distributions in different fields. SUMMARY

[0008] To overcome the shortcomings of the prior art, the present application provides a self-supervised color table extraction method and system for scalar data visualization images, which can accurately recover the continuous color map and potential scalar data used in a single visualization image without paired color map information or legends, improving the analysis and reuse ability of visualization images.

[0009] To achieve the above purpose, one or more embodiments of the present application provide the following technical solutions:

[0010] In a first aspect, a color table extraction neural network model construction method is disclosed, comprising:

[0011] Preprocessing a single scalar field visualization image into a standard format image;

[0012] Inputting the preprocessed standard format image into a visualization image encoder to extract deep features of the image;

[0013] Dividing the extracted deep features into two branches for separate processing: one branch generates grayscale scalar data corresponding to the image, and the other branch generates a color control point sequence through a multi-layer perception network, and the color control point sequence generates a continuous color table after spline interpolation;

[0014] Performing color mapping on the extracted scalar data and continuous color table to generate a reconstructed image;

[0015] Comparing the reconstructed image with the original single scalar field visualization image, comparing the continuous color table and scalar data with the true continuous color table and true data, respectively calculating a plurality of self-supervised loss functions, and continuously updating neural network parameters and color control points through back propagation for self-supervised fine-tuning to obtain a color table extraction neural network model.

[0016] As a further technical solution, the color table extraction neural network model comprises:

[0017] A visualization image encoder, a data decoder, and a color decoder;

[0018] The visualization image encoder is used to extract the latent representation of the input single scalar field visualization image;

[0019] The data decoder restores the scalar data based on the latent representation of the image;

[0020] The color decoder outputs a fixed number of color control point sequences to construct a continuous color table through spline interpolation.

[0021] As a further technical solution, self-supervised fine-tuning is performed through a plurality of self-supervised loss functions, and the loss functions specifically include:

[0022] A reconstruction loss function and a color order loss function between the input image and the reconstructed image;

[0023] A data loss function and a color loss function between the scalar data and the continuous color table generated in the initial iteration and the scalar data and the continuous color table generated in each subsequent iteration.

[0024] As a further technical solution, the reconstruction loss function between the input image and the reconstructed image is used to measure the pixel difference between the input image and the reconstructed image.

[0025] As a further technical solution, the color order loss function between the input image and the reconstructed image is used to punish the color order abnormality between the color control point sequence.

[0026] As a further technical solution, the data loss function between the scalar data and the continuous color table generated in the initial iteration and the scalar data and the continuous color table generated in each subsequent iteration is used to measure the difference between the extracted scalar data and the true data.

[0027] As a further technical solution, the color loss function between the scalar data and the continuous color table generated in the initial iteration and the scalar data and the continuous color table generated in each subsequent iteration is used to measure the difference between the extracted color table and the original color table.

[0028] In a second aspect, a color table extraction method is disclosed, comprising:

[0029] Obtaining a single scalar field visualization image to be processed;

[0030] Inputting the single scalar field visualization image into a pre-trained color table extraction neural network model, the color table extraction neural network model processing the input image to extract scalar data and a color control point sequence;

[0031] Interpolating the color control point sequence to generate a continuous color table, and mapping the continuous color table and the extracted scalar data to generate a reconstructed image;

[0032] In the inference stage of the color table extraction neural network model, self-supervised fine-tuning is performed, including: obtaining the difference between the input single scalar field visualization image and the reconstructed image, and iteratively optimizing the color control point sequence and the network parameters of the related color table extraction neural network model based on the obtained difference through a back propagation mechanism to extract a final continuous color table.

[0033] In a third aspect, a color table extraction system is disclosed, comprising:

[0034] An image acquisition module configured to acquire a single scalar field visualization image;

[0035] An image processing module configured to input the single scalar field visualization image into a pre-trained color table extraction neural network model, the color table extraction neural network model processing the input image to extract scalar data and a color control point sequence;

[0036] The reconstructed image generation module is configured to interpolate the color control point sequence to generate a continuous color table, and map the continuous color table and the extracted scalar data to generate a reconstructed image.

[0037] The self-supervised fine-tuning module is configured to perform self-supervised fine-tuning in the color table extraction neural network model inference stage, including obtaining the difference between the input single scalar field visualization image and the reconstructed image, iteratively optimizing the color control point sequence and the network parameters of the related color table extraction neural network model based on the obtained difference through a back propagation mechanism, and extracting a final continuous color table.

[0038] The above one or more technical solutions have the following beneficial effects:

[0039] In the training model process in the technical solution of the application, the difference between the input single scalar field visualization image and the reconstructed image is obtained first, the network parameters of the color control point sequence and the related color table extraction neural network model are iteratively optimized based on the obtained difference through a back propagation mechanism, and a final continuous color table is extracted. The above-mentioned means optimizes the network in the inference stage through a self-supervised fine-tuning mechanism, so that the system has the adaptability to images with large distribution differences, and the robustness and generalization ability of the model in a real complex scene are significantly improved. This method can not only be applied to color table inversion and image recoloring, but also can be extended to color table migration technology.

[0040] In the process of self-supervised fine-tuning of the color table extraction neural network model in the inference stage, multiple self-supervised loss functions are used for self-supervised fine-tuning. The loss function introduces multiple constraint mechanisms such as color fidelity loss, data reconstruction loss and color sequence loss, which effectively improves the accuracy and visual consistency of color table extraction. Especially using the spline interpolation method based on control points, a continuous color table with smooth transition and consistent sequence can be generated, avoiding the problem of color band jumping in traditional linear interpolation.

[0041] In the technical solution of the application, the single scalar field visualization image is input into the pre-trained color table extraction neural network model, the color table extraction neural network model processes the input image to extract scalar data and color control point sequence, and the above-mentioned means can be applied to both cases with legends and cases without legends by accurately extracting the color table and implicit scalar data used in the single visualization image. The color control point sequence is interpolated to generate a continuous color table, and the continuous color table and the extracted scalar data are mapped to generate a reconstructed image, realizing effective separation and reconstruction of color information and data information in the image, and improving the interpretability, usability and automatic processing capability of the visualization image.

[0042] Advantages of the present application's additional aspects will become apparent in the description below. To the extent that certain details are not described, those of ordinary skill in the art will appreciate that BRIEF DESCRIPTION OF DRAWINGS

[0043] The accompanying drawings, which form a part of the specification, are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification. The illustrations are shown to explain the present application and are not intended to limit the present application in any manner.

[0044] Figure 1 A flowchart of the self-supervised color table extraction method for a scalar field visualization image according to an embodiment of the present application;

[0045] Figure 2 An architecture diagram of the color extraction neural network according to an embodiment of the present application;

[0046] Figure 3 Qualitative results under synthetic data sets and real-world data sets according to an embodiment of the present application. DETAILED DESCRIPTION

[0047] It should be noted that the following detailed description is merely exemplary and is intended to provide further description of the application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the application pertains.

[0048] It should be noted that the terms used herein are merely for the purpose of describing specific embodiments and are not intended to limit exemplary embodiments according to the present application.

[0049] In the case of no conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.

[0050] Embodiment one

[0051] The present embodiment discloses a color table extraction neural network model construction method, comprising:

[0052] A single scalar field visualization image is preprocessed into a standard format image;

[0053] The preprocessed standard format image is input into a visualization image encoder to extract the deep features of the image;

[0054] The extracted deep features are divided into two branches for separate processing: one branch generates the grayscale scalar data corresponding to the image, and the other branch generates a color control point sequence through a multi-layer perception network. The color control point sequence is interpolated through a spline to generate a continuous color table;

[0055] The extracted scalar data and continuous color table are color mapped to generate a reconstructed image;

[0056] The reconstructed image is compared with the original single scalar field visualization image, the continuous color table and the scalar data are compared with the real continuous color table and the real data, a plurality of self-supervised loss functions are calculated respectively, and the neural network parameters and the color control points are continuously updated through back propagation, self-supervised fine-tuning is performed, and a color table extraction neural network model is obtained.

[0057] More specifically, the color table extraction neural network model is based on the training of the color extraction neural network, and the specific training process is as follows:

[0058] Taking the training process of a single picture as an example, a scalar field visualization image I and a neural network N are input to the model, the image is first scaled to a uniform size and normalized to a standard format, and then input to a visualization image encoder to extract deep features z. The extracted deep features are then divided into two path branches for processing: one path generates the gray scalar data D(I) corresponding to the image through the U-Net structure, and the other path generates the vector P(I) representing the color control points through a multi-layer perceptron network. These control points generate a continuous color table C(I) after three B-spline interpolations. Then, the model performs color mapping using the extracted scalar data and the continuous color table to generate a reconstructed image R(I). Then, the model compares the reconstructed image with the original image, compares C(I) and D(I) with the real continuous color table and the real data, calculates a plurality of self-supervised loss functions such as image reconstruction loss, color order loss, data loss and color loss, and continuously updates the neural network parameters and the color control points through back propagation, self-supervised fine-tuning, and gradually improves the extraction ability and output quality of the model.

[0059] When constructing the training data, the present application first collects two types of 2D scalar field data, including 179 meteorological data from the ECMWF data set, and 228 scalar field data generated by different mathematical functions (such as linear gradient change, etc.). In addition, the present application collects 179 color tables from the Python matplotlib library and academic papers. Based on the collected scalar field data and color tables, the present application generates corresponding scalar field visualization images using linear mapping, which includes a total of 72853 visualization images. The training set includes scalar field visualization images, color tables, and scalar field data.

[0060] In order to make the above-mentioned model have better processing performance on the input image, in the inference stage, the self-supervised fine-tuning is performed through the reconstruction loss and the color order loss between the input scalar field visualization image I and the reconstructed image R(I), the scalar data and the continuous color table generated in the initial iteration and the scalar data and the continuous color table generated in each iteration (i is the iteration number) multiple self-supervised loss functions, and constantly update the neural network parameters and color control points through backpropagation, perform self-supervised fine-tuning, and gradually improve the extraction ability and output quality of the model.

[0061] More specifically, the loss function includes the following four, for joint training and self-supervised fine-tuning:

[0062] Image reconstruction loss , used to measure the pixel difference between the scalar field visualization image and the reconstructed scalar field visualization image, defined by the mean square error (MSE) formula as:

[0063]

[0064] wherein, W and H are the width and height of the input scalar field visualization, is the color value of the original scalar field visualization image at the two-dimensional image space , and is the color value of the reconstructed scalar field visualization at the two-dimensional image space .

[0065] Data loss , used to measure the difference between the extracted scalar data and the true data, defined by the formula as:

[0066]

[0067] wherein, is the normalized scalar value (range from 0 to 1) of the original scalar field data at the two-dimensional data space (h, w), is the normalized scalar value of the extracted scalar field data at the two-dimensional data space (h, w), W and H are the width and height of the input scalar field data, consistent with the width and height of the scalar field visualization image.

[0068] Color loss , used to measure the difference between the extracted color table and the true color table, defined by the formula as:

[0069]

[0070] wherein, is the i-th color in the original color mapping, is the color at the parameter position t(i) in the extracted color mapping, and m is the number of colors in the color mapping.

[0071] Color order loss , to penalize the color ordering abnormality between control points, the formula is defined as:

[0072]

[0073] where, and are the color values at the parameter positions t(i) and t(j) in the extracted colormap, i and j are the indices of colors, ranging from 1 to m.

[0074] The above image reconstruction loss function evaluates the quality of colormap extraction from a global perspective by remapping the extracted colormap and data into an image and comparing it with the original scalar field image. This loss function provides an end-to-end optimization target that ensures overall consistency; it can also serve as a self-supervised signal in unsupervised or semi-supervised scenarios to enhance the network's generalization ability.

[0075] The data loss function is mainly used to measure the difference between the scalar field data and the real data. This loss function helps to improve the restoration accuracy of the scalar field data, thereby strengthening the consistency of color and data decoupling.

[0076] The color loss function mainly measures the difference between the predicted colormap and the real colormap, which helps to improve the extraction accuracy of the colormap and ensures that the predicted colormap is as consistent as possible with the real colormap.

[0077] The color order loss function mainly ensures that the colors in the recovered colormap are arranged in a certain order, consistent with human perception. This loss function helps to ensure the coherence and interpretability of color changes, avoiding perceptual ambiguity caused by color disorder, especially when the input lacks a legend (color table) visualization image. This loss function can improve the perceptual rationality of the predicted colormap.

[0078] In the model training and inference stage, the above losses are proportionally combined to form the total loss function, defined as:

[0079]

[0080] where, , , are adjustable weight coefficients, all set to 1 by default.

[0081] The advantage of the above training scheme is that it simultaneously decouples the color table information and the grayscale scalar data information in the scalar field visualization image, integrates B-spline curves and color table order constraints, and introduces a differentiable colormap mapping module. Not only does it construct an end-to-end colormap extraction solution, but it also enhances the generalization ability of the colormap extraction method in a self-supervised fine-tuning manner.

[0082] In another embodiment, a color table extraction method is disclosed, comprising:

[0083] Step one: obtaining a single scalar field visualization image;

[0084] Step two: inputting the image into a pre-trained color table extraction neural network model to obtain extracted scalar data and a color control point sequence;

[0085] Step three: interpolating the control point sequence output by the color table decoder through a cubic B-spline function to generate a continuous color table curve, and generating a reconstructed image through a differentiable color table mapping module with the obtained scalar data;

[0086] Step four: performing self-supervised fine-tuning in the model inference stage, based on the difference between the input image and the reconstructed image, iteratively optimizing the control points and related network parameters through a backpropagation mechanism to extract a continuous color table;

[0087] Among them, the training process of the color table extraction neural network model defines and jointly uses multiple loss functions, including reconstruction loss function, data loss function, color loss function and color sorting loss function, to optimize the model.

[0088] In the technical solution of the embodiment, the color table extraction neural network model processes the input image to extract scalar data and a color control point sequence. The above method can be used in both cases with legends and without legends by accurately extracting the color table and implicit scalar data from a single visualization image. The color control point sequence is interpolated to generate a continuous color table, and the continuous color table and the extracted scalar data are mapped to generate a reconstructed image, effectively separating and reconstructing the color information and data information in the image, improving the interpretability, usability and automatic processing capability of the visualization image, and solving the problem that the prior art cannot accurately extract the color table and potential data in the visualization image when lacking legend information.

[0089] In step one, a single scalar field visualization image to be processed is obtained, and the obtained visualization image is preprocessed in image size, with the image size set to 256x256, as the input image of the neural network model.

[0090] In one embodiment, step two inputs the image into a pre-trained color table extraction neural network model to obtain extracted scalar data and a color control point sequence. The color table extraction neural network model is as follows: Figure 2As shown, the visual image encoder, the data decoder, and the color decoder are composed. The visual image encoder is composed of a convolutional layer and six down-sampling modules, each module is composed of 1 residual block and 1 down-sampling (stride = 2), and the residual block contains two 3x3 convolutional layers. The visual image encoder is used to extract the latent representation of the input image. The data decoder adopts the U-Net structure and is composed of six up-sampling modules and a jump connection, each up-sampling module contains a nearest neighbor up-sampling (scale = 2) and a residual block, the residual block contains 2 3x3 convolution, the jump connection is symmetrically corresponding to the encoder layer, and the data decoder is used for decoupling processing of the input to restore the scalar data. The color decoder adopts a multi-layer perceptron (MLP) containing 6 layers of fully connected neural networks, and the color decoder is used to output a fixed number of control points for constructing a continuous color table through cubic B-spline interpolation. The above model can extract the color table and the latent data of the visual image from a single visual image.

[0091] In step three, specifically, a differentiable color table mapping module is used to receive the scalar data and the continuous color table decoupled by the neural network decoder, the neural network decoder includes the data decoder and the color table decoder, and the mapping is performed according to the size of the color values in the continuous color table and the gray values in the scalar data, so that the reconstructed image is obtained.

[0092] The gray values in the scalar data output by the neural network are one-to-one mapped with the colors in the continuous differentiable color table, that is, the range of the normalized gray values, each gray value corresponds to the color value of a position in the color table to realize the conversion of the scalar to the RGB color, and the gray values from small to large correspond to the color values of the continuous color table from left to right.

[0093] The network is responsible for extracting features and outputting scalar data and color control points. The differentiable color mapping module takes the output of the main network as input to complete the final color image reconstruction. The reconstructed image is subjected to loss calculation (such as MSE) with the original image, and the loss can be back propagated to the color table and the main network. The main loop of training calls the main network and the color mapping module to obtain the final output and calculate the loss. The differentiable color table mapping module converts the traditional discrete color mapping into a differentiable continuous operation, so that the entire color conversion process can be learned and optimized through the neural network, and an end-to-end trainable mapping from scalar data to color image is realized.

[0094] In order to prove the effectiveness of the scheme described in the application, specific experimental verification is carried out in this embodiment:

[0095] As shown in Table 1, the quantitative comparison of the method of the embodiment and the baseline method in the RGB space on the synthetic dataset and the real-world dataset is shown. Three different evaluation indicators are used, including image reconstruction error (MSE), peak signal-to-noise ratio (PSNR), and structural similarity (SSIM), including two cases of whether to consider the color table extraction direction. It can be seen that the method of the embodiment is superior to the baseline method. This result shows that in terms of image reconstruction capability, the method of the embodiment is superior to the traditional baseline method in the indicators of MSE, PSNR and SSIM, indicating that the model can more accurately restore the color distribution and structural information of the input image.

[0096] Table 1

[0097]

[0098] In Figure 3 , the qualitative comparison of the method of the embodiment and the baseline method on the extracted color table results of the synthetic dataset and the real-world dataset is shown, and six samples with corresponding true values are shown for each dataset. Input is the input visualization image, GT is the color table true value, Ours is the method of the embodiment, and Yuan et al is the baseline method. It can be seen that the results of the method of the embodiment are closer to the true value than the results of the baseline method on the two datasets, and can extract higher-quality color tables.

[0099] Embodiment Two

[0100] The purpose of the embodiment is to provide a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to realize the steps of the above method.

[0101] Embodiment Three

[0102] The purpose of the embodiment is to provide a computer-readable storage medium.

[0103] A computer-readable storage medium has a computer program stored thereon, and the program is executed by a processor to perform the steps of the above method.

[0104] Embodiment Four

[0105] The purpose of the embodiment is to provide a self-supervised color table extraction system for scalar data visualization images, including:

[0106] An image acquisition module configured to acquire a single scalar field visualization image;

[0107] The image processing module is configured to input the single scalar field visualization image into a pre-trained color table extraction neural network model, and the color table extraction neural network model processes the input image to extract scalar data and a color control point sequence;

[0108] The reconstructed image generation module is configured to interpolate the color control point sequence to generate a continuous color table, and map the continuous color table and the extracted scalar data to generate a reconstructed image;

[0109] The self-supervised fine-tuning module is configured to perform self-supervised fine-tuning in the inference stage of the color table extraction neural network model, including: obtaining a difference between the input single scalar field visualization image and the reconstructed image, iteratively optimizing the color control point sequence and network parameters of the color table extraction neural network model based on the obtained difference through a back propagation mechanism, and extracting a final continuous color table.

[0110] Embodiment five

[0111] The purpose of the embodiment is to provide a computer program product containing instructions which, when run on a computer, cause the computer to perform the method and functions involved in any of the above embodiments.

[0112] The steps and method embodiments of the above embodiments correspond to embodiment one, and the specific embodiments can refer to the related description part of embodiment one. The term "computer readable storage medium" should be understood to include a single medium or multiple media of one or more instruction sets; it should also be understood to include any medium that can store, encode or carry instruction sets for execution by a processor and cause the processor to perform any method in the present application.

[0113] Those skilled in the art should understand that the above-mentioned modules or steps of the present application can be realized by a general computer device, alternatively, they can be realized by program codes executable by a computing device, so that they can be stored in a storage device for execution by a computing device, or they can be made into individual integrated circuit modules, or a plurality of modules or steps among them can be made into a single integrated circuit module. The present application is not limited to any specific combination of hardware and software.

[0114] The above describes the specific embodiments of the present application in combination with the accompanying drawings, but is not a limitation on the scope of protection of the present application. Those skilled in the art should understand that various modifications or variations made by those skilled in the art on the basis of the technical solutions of the present application without creative labor are still within the scope of protection of the present application.

Claims

1. A method for constructing a neural network model for color table extraction, characterized by: include: The single scalar field visualization image is preprocessed into a standard format image; The image, preprocessed to a standard format, is input into a visual image encoder to extract the latent deep features of the image. The extracted deep features are divided into two branches for processing: one branch generates grayscale scalar data corresponding to the image, and the other branch generates a sequence of color control points through a multilayer perceptron network. The sequence of color control points is then interpolated by splines to generate a continuous color table. Color mapping is performed on the extracted scalar data and the continuous color table to generate a reconstructed image; The reconstructed image is compared with the original single scalar field visualization image. The continuous color table and scalar data are compared with the real continuous color table and real data. Multiple self-supervised loss functions are calculated respectively. The neural network parameters and color control points are continuously updated through backpropagation to perform self-supervised fine-tuning and obtain the color table extraction neural network model. Self-supervised fine-tuning is performed using multiple self-supervised loss functions, which specifically include: Reconstruction loss function and color order loss function between input image and reconstructed image; The reconstruction loss function between the input image and the reconstructed image is used to measure the pixel differences between the input image and the reconstructed image; The color order loss function between the input image and the reconstructed image is used to penalize color ordering anomalies between color control point sequences; The data loss function and color loss function between the scalar data and continuous color table generated in the initial iteration and the scalar data and continuous color table generated in each subsequent iteration; The data loss function between the scalar data and continuous color table generated in the initial iteration and the scalar data and continuous color table generated in each subsequent iteration is used to measure the difference between the extracted scalar data and the real data. The color loss function between the scalar data and continuous color table generated in the initial iteration and the scalar data and continuous color table generated in each subsequent iteration is used to measure the difference between the extracted color table and the original color table; The color order loss The formula is defined as: in, and These are the color values ​​at positions t(i) and t(j) in the extracted color table, where i and j are the color indices, ranging from 1 to m, and m is the number of colors in the color map.

2. The method for constructing a neural network model for color table extraction as described in claim 1, characterized in that, The color table extraction neural network model includes: Visual image encoder, data decoder, and color decoder; The visualization image encoder is used to extract the latent representation of the input single scalar field visualization image; The data decoder restores scalar data based on the latent representation of the image; The color decoder outputs a fixed number of color control point sequences, which are then used to construct a continuous color table through spline interpolation.

3. A color table extraction method, characterized in that, The color table extraction neural network model is implemented based on the color table extraction neural network model construction method according to any one of claims 1-2, the method comprising: Acquire a single scalar field visualization image to be processed; A single scalar field visualization image is input into a pre-trained color table extraction neural network model, which processes the input image to extract scalar data and color control point sequences. The color control point sequence is interpolated to generate a continuous color table, and then the continuous color table is mapped to the extracted scalar data to generate a reconstructed image. Self-supervised fine-tuning is performed during the inference stage of the color table extraction neural network model, including: obtaining the difference between the input single scalar field visualization image and the reconstructed image; iteratively optimizing the network parameters of the color control point sequence and the related color table extraction neural network model through backpropagation based on the obtained difference, and extracting the final continuous color table.

4. A color table extraction system, characterized in that, The system implements a color table extraction neural network model based on the color table extraction neural network model construction method according to any one of claims 1-2, and the system includes: The image acquisition module is configured to acquire a single scalar field visualization image; The image processing module is configured to: input a single scalar field visualization image into a pre-trained color table extraction neural network model, wherein the color table extraction neural network model processes the input image to extract scalar data and color control point sequences; The reconstructed image generation module is configured to: interpolate the color control point sequence to generate a continuous color table, and then map the continuous color table to the extracted scalar data to generate a reconstructed image; The self-supervised fine-tuning module is configured to perform self-supervised fine-tuning during the inference stage of the color table extraction neural network model, including: acquiring the difference between the input single scalar field visualization image and the reconstructed image, and iteratively optimizing the network parameters of the color control point sequence and the related color table extraction neural network model through backpropagation based on the acquired difference, so as to extract the final continuous color table.

5. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 3.

6. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method described in any one of claims 1-3.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it performs the steps of the method described in any one of claims 1-3 above.

Images