Intelligent historical cultural relic image texture enhancement method and system

Abstract

The application discloses a kind of intelligent historical cultural relic image texture enhancement method and system, belong to digital image processing technical field.The method includes multi-source data acquisition, defect material joint modeling, cross-domain consistent texture recovery and historical cultural relic image texture enhancement step;Defect material joint modeling is carried out based on multi-source image data set, and defect probability distribution diagram and material texture statistical parameter are generated;Through multi-source image structure direction consistency, phase consistency and high-frequency energy proportion stability analysis, cross-domain consistent texture seed region is constructed;According to the weight of credible area and direction matching relationship, the high-frequency texture component is controlled gain adjusted, and the enhancement reconstruction is realized in combination with safety upper limit control;The application can effectively avoid the misenhancement of crack, stain and reflective region, improve the authenticity, safety and controllability of texture enhancement result, and is suitable for historical cultural relic digitization protection and image restoration application.

Description

Technical Field

[0001] This invention relates to the field of digital image processing technology, specifically to an intelligent method and system for enhancing the texture of historical artifact images. Background Technology

[0002] The intelligent historical artifact image texture enhancement method refers to a digital image processing method that uses multi-source image data consistency analysis and defect region constraint mechanism to perform controlled enhancement of the real structural texture in historical artifact images while suppressing the enhancement of defect regions. This method extracts structural texture information that is stable under different imaging conditions by uniformly processing and jointly analyzing multi-source images of the same artifact. The enhancement intensity is controlled according to the degree of texture consistency and defect distribution, so that the enhancement process only acts on reliable texture regions. This improves the clarity of artifact image texture while avoiding the false enhancement of cracks, stains or peeling areas.

[0003] Existing image texture enhancement methods mostly employ sharpening filters, high-frequency enhancement, or contrast enhancement to achieve texture strengthening. However, images of historical artifacts have unique characteristics that distinguish them from ordinary natural images. Their surfaces often contain both genuine material textures and non-structural defect areas, such as cracks, corrosion marks, stains, or reflective areas. These areas also exhibit high-frequency information in the frequency domain. Traditional enhancement methods struggle to distinguish between genuine textures and defect information, and tend to simultaneously enhance defective areas during the enhancement process, resulting in enhancement results that deviate from the true structural characteristics of the artifact.

[0004] In addition, images of cultural relics are usually acquired under different lighting conditions, shooting angles or imaging bands. There are differences in lighting, noise and local pseudo-high frequency information between images. Existing methods usually enhance only a single image and lack a mechanism to comprehensively utilize the consistent structure between multiple source images. This makes it easy to mistakenly enhance pseudo-high frequency information caused by noise or changes in imaging conditions.

[0005] Meanwhile, most existing enhancement methods lack a safety control mechanism for the enhancement process, and the enhancement intensity is difficult to adjust according to the reliability of the texture, which can easily lead to over-enhancement in local areas, affecting the authenticity and reliability of the enhancement results.

[0006] Therefore, how to effectively distinguish between real material textures and defect areas during the process of enhancing the texture of historical artifact images, and how to achieve controlled enhancement based on the structural consistency of multi-source images, are technical problems that urgently need to be solved in this field. Summary of the Invention

[0007] To address the above issues and overcome the shortcomings of existing technologies, this invention provides an intelligent method and system for enhancing the texture of historical artifact images. The technical solution adopted by this invention is as follows: This invention provides an intelligent method for enhancing the texture of historical artifact images, which includes the following steps:

[0008] Step S1: Multi-source data acquisition;

[0009] Step S2: Joint modeling of defective materials;

[0010] Step S3: Cross-domain consistent texture restoration;

[0011] Step S4: Enhance the texture of historical artifact images.

[0012] Furthermore, in step S1, the multi-source data acquisition is used to construct a basic dataset of historical artifact images that is spatiotemporally consistent, radiometrically consistent, and uniform in resolution, providing a stable input basis for subsequent defect identification and texture restoration; specifically, it involves acquiring image data of the same artifact under different lighting conditions, different shooting angles, or different imaging bands, and performing geometric registration, exposure normalization, noise level estimation, and resolution unification processing on the acquired data to obtain a multi-source image data set with structural alignment and controlled lighting differences;

[0013] The multi-source image data set specifically includes: geometrically aligned image data, radiometrically normalized image data, noise estimation parameter data, and uniform resolution image data.

[0014] Furthermore, in step S2, the joint modeling of defect materials is used to distinguish between the real material texture of cultural relics and non-structural defect information, so as to avoid the mis-magnification of cracks, stains or reflective areas during the subsequent enhancement process; specifically, based on the multi-source image data set, a defect saliency distribution model and a material texture statistical model are constructed. Through the joint analysis of spatial features and frequency domain features, a defect probability distribution map and material prototype matching parameters are generated, and a structural mutual exclusion constraint relationship is formed to obtain the defect material joint modeling result data.

[0015] The joint modeling results of the defect materials specifically include: defect probability distribution map data, material texture statistical parameter data, and defect texture mutual exclusion constraint parameter data;

[0016] The defect probability distribution map data is used to characterize the spatial distribution of cracks, stains, reflective areas, or peeling areas.

[0017] The material texture statistical parameter data is used to characterize the spectral distribution characteristics, directional characteristics, and contrast distribution characteristics of the current cultural relic surface material;

[0018] The defect texture mutual exclusion constraint parameter data is used to limit the upper limit of high-frequency gain of the defect region during the enhancement stage.

[0019] Further, in step S3, the cross-domain consistent texture restoration is used to extract structural texture evidence that is stable under different imaging conditions from multiple source images and to suppress pseudo-high-frequency information that appears only in a single image. Specifically, based on the multi-source image dataset and the joint modeling result data of the defective material, a cross-domain consistent texture field is constructed, and a texture screening method combined with multi-dimensional feature consistency improvement is used to perform consistency verification and energy screening on the local structural responses in images from different sources, generating reliable texture seed regions, and obtaining cross-domain consistent texture restoration result data, including the following steps:

[0020] Step S31: Local structure orientation extraction. Gradient calculation and local structure orientation analysis are performed on each image in the multi-source image dataset to generate the structure orientation field corresponding to each image. Combined with the noise estimation parameter data, the preset orientation threshold is adaptively adjusted. The structure orientation of different images at the same spatial location is compared pairwise to calculate the orientation consistency index. When the orientation consistency index is greater than the preset orientation threshold, the location is marked as an orientation stable region, and the orientation consistency judgment result data is output.

[0021] Step S32: Local frequency domain phase consistency calculation: Perform local frequency domain transformation on each image in the multi-source image dataset to extract the corresponding local phase information; perform consistency superposition calculation on the local phases of different images at the same spatial location to generate a phase consistency index; when the phase consistency index is greater than a preset phase threshold, mark the location as a phase stable region and output the phase consistency judgment result data.

[0022] Step S33: High-frequency energy ratio stability detection: Perform frequency domain decomposition on each image in the multi-source image dataset to extract high-frequency energy components; calculate the high-frequency energy ratio stability index between different images at the same spatial location; when the high-frequency energy ratio stability index is greater than a preset energy ratio threshold, mark the location as an energy stable region and output the energy ratio stability data.

[0023] Step S34: Constructing a cross-domain consistency comprehensive score. Based on the direction consistency judgment result data, the phase consistency judgment result data, and the energy ratio stability data, a cross-domain consistency comprehensive score model is constructed; a comprehensive consistency score value is generated for each spatial location; the defect probability distribution map data in the defect material joint modeling result data is introduced for a one-vote veto filtering: if the defect probability of a certain spatial location is higher than the preset defect safety threshold, the comprehensive consistency score value of that location is directly set to zero or marked as a non-enhanced area to eliminate pseudo-high frequency interference; when the comprehensive consistency score value is greater than the preset comprehensive threshold, the location is determined as a reliable texture seed region.

[0024] Step S35: Generate a reliable high-frequency candidate region. Based on the reliable texture seed region, filter the original high-frequency components and retain only the high-frequency information within the reliable texture seed region. Generate cross-domain consistent texture seed map data, texture consistency score data, cross-domain consistent structure orientation field data, and reliable high-frequency candidate region data to obtain cross-domain consistent texture recovery result data.

[0025] The cross-domain consistent texture recovery result data specifically includes: cross-domain consistent texture seed map data, texture consistency score data, cross-domain consistent structure orientation field data, and reliable high-frequency candidate region data;

[0026] The cross-domain consistent texture seed map data is used to characterize structural regions that exist stably in multiple imaging domains;

[0027] The texture consistency score data is used to characterize the consistency intensity of each local region;

[0028] The cross-domain consistent structural orientation field data is used to characterize the cross-domain stable texture dominant orientation, serving as a benchmark for subsequent orientation matching enhancement.

[0029] The reliable high-frequency candidate region data is used as the priority gain region for subsequent enhancement.

[0030] Further, in step S4, the historical artifact image texture enhancement is used to perform controlled enhancement of reliable texture regions and gain suppression of defective regions while ensuring that the true structural features of the artifact are not altered. Specifically, based on the cross-domain consistent texture restoration result data and the joint modeling result data of the defective material, a reliable region-controlled optimized texture enhancement method is used to perform direction-selective gain adjustment on high-frequency components to obtain the historical artifact image texture enhancement result data, including the following steps:

[0031] Step S41: Frequency band separation processing: smoothing and filtering the original cultural relic image to extract the low-frequency basic structure image; based on the difference between the original image and the low-frequency basic structure image, obtain the high-frequency texture component data; output the low-frequency structure data and the high-frequency texture component data.

[0032] Step S42: Constructing Trusted Region Weights. Based on the cross-domain consistent texture seed map data and texture consistency score data in the cross-domain consistent texture recovery result data, and the defect probability distribution map data in the defect material joint modeling result data, construct region trust weight data. The region trust weight data is used to characterize the enhancement permission level of each spatial location. When a region simultaneously satisfies the trust seed label and the consistency score is higher than a preset threshold, a higher enhancement weight is assigned. When the defect probability is higher than the preset threshold, the enhancement weight is suppressed.

[0033] Step S43: Direction matching enhancement control, performing gradient direction analysis on the high-frequency texture component data to extract local structural direction information; matching the local structural direction information with the main direction information determined in the cross-domain consistent texture restoration stage; allowing enhancement when the direction matching degree meets the preset conditions; reducing the enhancement intensity of the corresponding region when the direction deviation exceeds the preset threshold; generating direction control coefficient data;

[0034] Step S44: Safety upper limit control. Based on the regional confidence weight data and the direction control coefficient data, calculate the local enhancement gain value at each spatial location; set a global safety gain upper limit parameter. When the local enhancement gain value exceeds the safety gain upper limit parameter, it is truncated and limited; generate high-frequency gain mapping data and safety threshold parameter data.

[0035] Step S45: Enhanced reconstruction generation: Adjust the high-frequency texture component data and the corresponding local enhancement gain value to obtain enhanced high-frequency component data; reconstruct the enhanced high-frequency component data with low-frequency structure data to generate enhanced cultural relic image data; generate enhancement contribution mapping data according to the change ratio of high-frequency components before and after enhancement; output the texture enhancement result data of historical cultural relic image.

[0036] The enhanced texture result data of the historical artifact image specifically includes: enhanced artifact image data, high-frequency gain mapping data, enhancement contribution mapping data, and safety threshold parameter data.

[0037] The present invention provides an intelligent image texture enhancement system for historical artifacts, comprising a data acquisition module, a defect modeling module, and a texture enhancement module;

[0038] The data acquisition module is used for multi-source data acquisition, obtaining a multi-source image data set through multi-source data acquisition, and sending the multi-source image data set to the defect modeling module;

[0039] The defect modeling module is used for joint modeling of defect materials. Through joint modeling of defect materials, it obtains joint modeling result data of defect materials and sends the joint modeling result data of defect materials to the texture enhancement module.

[0040] The texture enhancement module is used for cross-domain consistent texture restoration and texture enhancement of historical artifact images. Through cross-domain consistent texture restoration and texture enhancement of historical artifact images, the texture enhancement result data of historical artifact images is obtained.

[0041] The beneficial effects achieved by adopting the above solution are as follows:

[0042] (1) In the existing methods for enhancing the texture of historical artifact images, there is a lack of effective distinction between non-structural defects such as cracks, stains, reflections or peeling areas and real material textures. This leads to the defects being misjudged as real textures and simultaneously amplified during the enhancement process, thereby destroying the authenticity of the artifacts. For example, when there are natural oxidation spots and corrosion cracks on the surface of bronzes, traditional sharpening or high-frequency enhancement algorithms often mistakenly enhance the edges of corrosion cracks, making the crack boundaries present an over-sharpened pseudo-texture that does not conform to the original appearance of the artifact. This solution adopts joint modeling of defect materials. By constructing a defect probability distribution map and material texture statistical parameters, and forming a mutually exclusive constraint relationship between defect textures, the enhancement stage can restrict the enhancement area based on the defect probability distribution, thereby avoiding mistaken enhancement of defect areas, realizing effective distinction between real material textures and defect areas, and ensuring that the enhancement results conform to the real structural characteristics of the artifacts.

[0043] (2) In the existing methods for enhancing the texture of historical artifact images, there is a problem that enhancement is based solely on the high-frequency information of a single image, which cannot effectively distinguish between real structural textures and pseudo high-frequency information caused by noise, compression artifacts, or changes in illumination. This results in noise amplification or pseudo-structures in the enhanced image. For example, in mural images taken under low illumination conditions, random high-frequency fluctuations exist in local areas due to sensor noise. Traditional enhancement methods will mistakenly enhance this random noise into texture structure. This solution adopts cross-domain consistent texture restoration. By jointly verifying the structural direction consistency, phase consistency, and high-frequency energy ratio stability of images under different illumination conditions or different imaging bands, and combining the defect probability for veto filtering, only structural regions that are stable in multiple imaging domains are retained as reliable texture seed regions. This effectively suppresses pseudo high-frequency information that only appears in a single image and improves the authenticity and stability of the enhancement results.

[0044] (3) In view of the existing methods for enhancing the texture of historical artifact images, there is a lack of control mechanism for the texture direction characteristics in the enhancement process, which can easily destroy the original texture direction structure during the enhancement process, resulting in disordered texture direction or structural distortion after enhancement. For example, in the image of ancient brocade artifacts with obvious fabric texture direction, the traditional isotropic enhancement method will produce unreasonable enhancement in non-principal directions, making the fabric texture direction blurred or distorted. This scheme adopts direction matching enhancement control. By extracting local structural direction information and matching it with cross-domain consistent structural direction field, enhancement is only performed on the area with consistent direction and enhancement is suppressed on the area with deviated direction, thereby ensuring that the enhancement process is carried out along the real texture principal direction, so that the enhanced texture structure is consistent with the original structural direction.

[0045] (4) In view of the problems that existing methods for enhancing the texture of historical artifact images lack security control and the enhancement process is not traceable, it is easy to have over-enhancement in local areas or lack of reliability verification basis for enhancement results. For example, in images of stone carving artifacts with dense high-frequency textures, traditional enhancement methods may produce excessively high gains in local areas, making the texture present an unreal enhancement effect, and it is impossible to determine the reliability of the source of the enhancement area. This scheme adopts a texture enhancement mechanism that combines the construction of credible region weights with the control of security upper limit. By constructing regional credibility weight data, setting global security gain upper limit parameters and generating enhancement contribution mapping data, the enhancement process has the technical effect of controlled intensity and traceable enhancement source, thereby avoiding over-enhancement and improving the security and credibility of the results. Attached Figure Description

[0046] Figure 1 A flowchart illustrating an intelligent method for enhancing the texture of historical artifact images provided by this invention;

[0047] Figure 2 A schematic diagram of an intelligent historical artifact image texture enhancement system provided by the present invention;

[0048] Figure 3 This is a flowchart illustrating the process of cross-domain consistent texture restoration in step S3.

[0049] Figure 4 This is a flowchart illustrating the process of enhancing the texture of historical artifact images in step S4.

[0050] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0051] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0052] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0053] Example 1, see Figure 1 The present invention provides an intelligent method for enhancing the texture of historical artifact images, the method comprising the following steps:

[0054] Step S1: Multi-source data acquisition;

[0055] Step S2: Joint modeling of defective materials;

[0056] Step S3: Cross-domain consistent texture restoration;

[0057] Step S4: Enhance the texture of historical artifact images.

[0058] Example 2, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S1, the multi-source data acquisition is used to construct a basic dataset of historical cultural relics images that are spatiotemporally consistent, radiometrically consistent, and have uniform resolution, so as to provide a stable input basis for subsequent defect identification and texture restoration. Specifically, it involves acquiring image data of the same cultural relic object under different lighting conditions, different shooting angles, or different imaging bands, and performing geometric registration, exposure normalization, noise level estimation, and resolution unification processing on the acquired data to obtain a multi-source image data set with structural alignment and controlled lighting differences.

[0059] The multi-source image data set specifically includes: geometrically aligned image data, radiometrically normalized image data, noise estimation parameter data, and uniform resolution image data;

[0060] The geometrically aligned image data is used to characterize the spatial consistency of images from different sources;

[0061] The radiometrically normalized image data is used to eliminate brightness deviations caused by differences in exposure and illumination.

[0062] The noise estimation parameter data is used to characterize the noise level and compression artifact intensity of each source image;

[0063] The uniform resolution image data is used as the standard input tensor for subsequent joint modeling.

[0064] Example 3, see Figure 1 , Figure 2This embodiment is based on the above embodiment. In step S2, the joint modeling of defect materials is used to distinguish between the real material texture of cultural relics and non-structural defect information, so as to avoid the mis-enlargement of cracks, stains or reflective areas in the subsequent enhancement process. Specifically, based on the multi-source image data set, a defect saliency distribution model and a material texture statistical model are constructed. Through the joint analysis of spatial features and frequency domain features, a defect probability distribution map and material prototype matching parameters are generated, and a structural mutual exclusion constraint relationship is formed to obtain the defect material joint modeling result data.

[0065] The joint modeling results of the defect materials specifically include: defect probability distribution map data, material texture statistical parameter data, and defect texture mutual exclusion constraint parameter data;

[0066] The defect probability distribution map data is used to characterize the spatial distribution of cracks, stains, reflective areas, or peeling areas.

[0067] The material texture statistical parameter data is used to characterize the spectral distribution characteristics, directional characteristics, and contrast distribution characteristics of the current cultural relic surface material;

[0068] The defect texture mutual exclusion constraint parameter data is used to limit the upper limit of high-frequency gain of the defect region during the enhancement stage.

[0069] By performing the above operations, this method addresses the problem in existing methods for enhancing the texture of historical artifact images that lack effective differentiation between non-structural defects such as cracks, stains, reflections, or peeling areas and the actual material texture. This leads to defects being misjudged as real textures and magnified during the enhancement process, thus compromising the authenticity of the artifact. For example, when natural oxidation spots and corrosion cracks exist on the surface of bronze artifacts, traditional sharpening or high-frequency enhancement algorithms often mistakenly enhance the edges of corrosion cracks, resulting in over-sharpened pseudo-textures that do not conform to the original appearance of the artifact. This solution adopts joint modeling of defects and materials. By constructing a defect probability distribution map and material texture statistical parameters, and forming a mutually exclusive constraint relationship between defects and textures, the enhancement stage can restrict the enhancement area based on the defect probability distribution, thereby avoiding mistaken enhancement of defect areas and achieving effective differentiation between real material textures and defect areas. This ensures that the enhancement result conforms to the real structural characteristics of the artifact.

[0070] Example 4, see Figure 1 , Figure 2 and Figure 3This embodiment is based on the above embodiment. In step S3, the cross-domain consistent texture restoration is used to extract structural texture evidence that is stable under different imaging conditions from multiple source images and to suppress pseudo-high-frequency information that only appears in a single image. Specifically, based on the multi-source image dataset and the joint modeling result data of the defective material, a cross-domain consistent texture field is constructed, and a texture screening method combined with multi-dimensional feature consistency improvement is used to perform consistency verification and energy screening on the local structural responses in images from different sources, generating a reliable texture seed region, and obtaining the cross-domain consistent texture restoration result data, including the following steps:

[0071] Step S31: Local structure orientation extraction. Gradient calculation and local structure orientation analysis are performed on each image in the multi-source image dataset to generate the structure orientation field corresponding to each image. Combined with the noise estimation parameter data, the preset orientation threshold is adaptively adjusted. The structure orientation of different images at the same spatial location is compared pairwise to calculate the orientation consistency index. When the orientation consistency index is greater than the preset orientation threshold, the location is marked as an orientation stable region, and the orientation consistency judgment result data is output.

[0072] In this embodiment, gradient calculation is first performed on each image in the multi-source image dataset to extract the gradient responses in the horizontal and vertical directions, and local structural orientation information is calculated based on the gradient responses. Preferably, the local structural orientation is calculated using a structural tensor or the Sobel operator, and the formula for calculating the orientation consistency index between different images at the same location is as follows:

[0073] ;

[0074] In the formula, This is an index of directional consistency between different images at the same location. x is the X-axis coordinate index, y is the Y-axis coordinate index, K is the total number of images, and k is the image index. It is the k-th image at position The structural direction at that location, It is the average structural direction;

[0075] In a preferred embodiment:

[0076] when When this occurs, mark the location as a region with stable orientation;

[0077] when At that time, the location is marked as an unstable region.

[0078] Simultaneously, the directional consistency index is corrected by combining the noise estimation parameter data obtained in step S1. When the local noise level is high, the directional consistency score is appropriately attenuated.

[0079] Step S32: Local frequency domain phase consistency calculation: Perform local frequency domain transformation on each image in the multi-source image dataset to extract the corresponding local phase information; perform consistency superposition calculation on the local phases of different images at the same spatial location to generate a phase consistency index; when the phase consistency index is greater than a preset phase threshold, mark the location as a phase stable region and output the phase consistency judgment result data.

[0080] In this embodiment, a local frequency domain transformation is performed on each image in the multi-source image dataset to extract the corresponding phase information; preferably, the frequency domain transformation uses either a Log-Gabor filter or a short-time Fourier transform, and the specific calculation formula for the phase consistency index is as follows:

[0081] ;

[0082] In the formula, It is a phase consistency index. It is the k-th image at position Local phase at;

[0083] In a preferred embodiment:

[0084] when At that time, mark this location as the phase-stable region;

[0085] when At that time, the location is marked as a phase-unstable region;

[0086] Since real texture structures typically maintain similar phase characteristics under different lighting or imaging band conditions, while random noise or compression artifacts often lack stable phase structures, phase consistency analysis can effectively distinguish between real texture and pseudo-structure information; by performing the above operations, phase consistency determination result data is obtained.

[0087] Step S33: High-frequency energy ratio stability detection: Perform frequency domain decomposition on each image in the multi-source image dataset to extract high-frequency energy components; calculate the high-frequency energy ratio stability index between different images at the same spatial location; when the high-frequency energy ratio stability index is greater than a preset energy ratio threshold, mark the location as an energy stable region and output the energy ratio stability data.

[0088] In this embodiment, frequency domain decomposition is performed on the multi-source images to extract the high-frequency energy components from each image. Preferably, the frequency domain decomposition employs wavelet decomposition or Laplacian pyramid decomposition. The formula for calculating the high-frequency energy proportion stability index is as follows:

[0089] ;

[0090] In the formula, It is the high-frequency energy ratio stability index. It is the k-th image at position The high-frequency energy at that location, where Var is the variance calculation function. It is the mean calculation function. It is a tiny constant that prevents the denominator from being zero;

[0091] In a preferred embodiment:

[0092] when At that time, mark this location as an energy-stable region;

[0093] when At that time, the location is marked as an energy-unstable region;

[0094] By performing the above operations, energy ratio stability data is generated;

[0095] Step S34: Constructing a cross-domain consistency comprehensive score. Based on the direction consistency judgment result data, the phase consistency judgment result data, and the energy ratio stability data, a cross-domain consistency comprehensive score model is constructed; a comprehensive consistency score value is generated for each spatial location; the defect probability distribution map data in the defect material joint modeling result data is introduced for a one-vote veto filtering: if the defect probability of a certain spatial location is higher than the preset defect safety threshold, the comprehensive consistency score value of that location is directly set to zero or marked as a non-enhanced area to eliminate pseudo-high frequency interference; when the comprehensive consistency score value is greater than the preset comprehensive threshold, the location is determined as a reliable texture seed region.

[0096] Preferably, the formula for calculating the overall consistency score is as follows:

[0097] ;

[0098] In the formula, It is the overall consistency score. This is the directional weight, with a preferred value of 0.35. It is an indicator of directional consistency. This is the phase weight, with a preferred value of 0.40. It is a phase consistency index. This is the energy stability weight, with a preferred value of 0.25. It is the high-frequency energy ratio stability index;

[0099] As a further improvement to this embodiment, the defect probability distribution map data obtained in step S2 is introduced for constraint, when the following conditions are met... At that time, enhancement suppression processing is performed at the position (x,y), where, It is defect probability distribution data. This is the defect safety threshold, preferably set to 0.40, when the following conditions are met. and When this occurs, the location is identified as a reliable texture region;

[0100] Step S35: Generate a reliable high-frequency candidate region. Based on the reliable texture seed region, filter the original high-frequency components and retain only the high-frequency information within the reliable texture seed region. Generate cross-domain consistent texture seed map data, texture consistency score data, cross-domain consistent structure orientation field data, and reliable high-frequency candidate region data to obtain cross-domain consistent texture recovery result data.

[0101] In this embodiment, a reliable texture seed map is constructed based on the comprehensive consistency score data obtained in step S34, and spatial locations that meet the reliability conditions are marked as texture seed regions. Subsequently, neighborhood connectivity analysis and morphological smoothing are performed on the texture seed regions to eliminate isolated pixels and enhance region continuity; preferably, the neighborhood analysis adopts an 8-neighborhood connectivity strategy, and the texture seed regions are smoothed through 3×3 morphological closing operations to generate continuous reliable texture regions;

[0102] The cross-domain consistent texture recovery result data specifically includes: cross-domain consistent texture seed map data, texture consistency score data, cross-domain consistent structure orientation field data, and reliable high-frequency candidate region data;

[0103] The cross-domain consistent texture seed map data is used to characterize structural regions that exist stably in multiple imaging domains;

[0104] The texture consistency score data is used to characterize the consistency intensity of each local region;

[0105] The cross-domain consistent structural orientation field data is used to characterize the cross-domain stable texture dominant orientation, serving as a benchmark for subsequent orientation matching enhancement.

[0106] The reliable high-frequency candidate region data is used as the priority gain region for subsequent enhancement.

[0107] By performing the above operations, this method addresses the problem in existing methods for enhancing the texture of historical artifact images. These methods rely solely on high-frequency information from a single image, failing to effectively distinguish between genuine structural textures and pseudo-high-frequency information caused by noise, compression artifacts, or lighting variations. This leads to noise amplification or pseudo-structure in the enhanced image. For example, in mural images taken under low-light conditions, sensor noise causes random high-frequency fluctuations in local areas. Traditional enhancement methods mistakenly enhance this random noise as texture structure. This solution employs cross-domain consistent texture restoration. By jointly verifying the structural orientation consistency, phase consistency, and high-frequency energy ratio stability of images under different lighting conditions or imaging bands, and combining this with defect probability for veto filtering, only structural regions that are stable in multiple imaging domains are retained as reliable texture seed regions. This effectively suppresses pseudo-high-frequency information appearing only in a single image, improving the realism and stability of the enhancement results.

[0108] Example 5, see Figure 1 , Figure 2 and Figure 4 This embodiment is based on the above embodiment. In step S4, the historical artifact image texture enhancement is used to perform controlled enhancement on reliable texture regions and gain suppression on defective regions while ensuring that the true structural features of the artifact are not changed. Specifically, based on the cross-domain consistent texture restoration result data and the joint modeling result data of the defective material, a reliable region controlled optimized texture enhancement method is used to perform direction-selective gain adjustment on high-frequency components to obtain historical artifact image texture enhancement result data, including the following steps:

[0109] Step S41: Frequency band separation processing: smoothing and filtering the original cultural relic image to extract the low-frequency basic structure image; based on the difference between the original image and the low-frequency basic structure image, obtain the high-frequency texture component data; output the low-frequency structure data and the high-frequency texture component data.

[0110] In this embodiment, the original artifact image I is first acquired, and low-frequency structure is extracted from it using edge-preserving smoothing filtering to obtain the low-frequency basic structure image I. L Preferably, the edge-preserving smoothing filter employs either guided filtering or bilateral filtering, wherein:

[0111] When the surface of an artifact image has obvious undulations in brightness but the edge contour is relatively stable, guided filtering is preferred.

[0112] When there is local noise in the image of an artifact and it is necessary to preserve the transition boundaries of the blocks, bilateral filtering is preferred.

[0113] In a preferred embodiment, the guiding filter radius is set to 5–9 pixels, and the regularization parameter is set to 10. −3~10 −2 The spatial domain standard deviation of the bilateral filter is 3 to 7, and the gray domain standard deviation is 0.08 to 0.15.

[0114] After obtaining the low-frequency basic structure image I L Subsequently, high-frequency texture component data I was extracted by the difference between the original artifact image and the low-frequency basic structure image. H , is represented as:

[0115] I H =I−I L ;

[0116] The low-frequency basic structure image is used to characterize the overall outline, layered background, and large-scale material distribution in the cultural relic image; the high-frequency texture component data is used to characterize the fine texture, scratch edges, fabric texture, or brushstroke structure in the cultural relic image.

[0117] Preferably, to avoid local extreme noise being directly included in the high-frequency texture component, an amplitude normalization process is further performed after high-frequency extraction to limit the high-frequency texture component to a preset amplitude range, which can be set to 8% to 20% of the dynamic range of the original image;

[0118] By performing the above operations, the subsequent enhancement processing is focused on the texture detail layer, rather than directly changing the overall structural brightness and basic outline of the artifact image, thus providing a clear target for the enhancement of the reliable region control.

[0119] Step S42: Constructing Trusted Region Weights. Based on the cross-domain consistent texture seed map data and texture consistency score data in the cross-domain consistent texture recovery result data, and the defect probability distribution map data in the defect material joint modeling result data, construct region trust weight data. The region trust weight data is used to characterize the enhancement permission level of each spatial location. When a region simultaneously satisfies the trust seed label and the consistency score is higher than a preset threshold, a higher enhancement weight is assigned. When the defect probability is higher than the preset threshold, the enhancement weight is suppressed.

[0120] In this embodiment, the cross-domain consistent texture seed map data M in the cross-domain consistent texture recovery result data is used. s Texture consistency score data S c And the defect probability distribution data P in the joint modeling result data of the defective material. d Construct regional credibility weights W c The region confidence weight is used to measure whether a certain spatial location is suitable as a texture enhancement region, and the degree to which it is suitable for enhancement.

[0121] Preferably, the regional credibility weight can be expressed as:

[0122] ;

[0123] In the formula, It is the regional credibility weight. It is cross-domain consistent texture seed map data. It is texture consistency score data. It is a defect probability distribution map data;

[0124] In a preferred embodiment, a higher enhancement license weight is assigned to the corresponding region when the following conditions are met: the cross-domain consistent texture seed is marked as 1; the texture consistency score Sc is greater than 0.65; and the defect probability Pd is less than 0.30.

[0125] When the defect probability P d When the value exceeds a preset defect suppression threshold, the enhancement weight is rapidly suppressed; preferably, the defect suppression threshold is between 0.40 and 0.55; more preferably, when P d When the value is greater than 0.50, the weight of the corresponding region will be directly reduced to less than 20% of the original value to avoid false enhancement of crack boundaries, stained areas or peeling transition areas.

[0126] Furthermore, to prevent discrete jumps in the weight map, it is preferable to use a 3×3 or 5×5 neighborhood smoothing method to locally continuousize the regional confidence weight data, generating a continuous confidence weight map. After this processing, the transition of the enhanced region boundary is smoother, which helps to avoid blocky abrupt changes in the enhancement result.

[0127] By performing the above operations, this step converts the credible texture evidence obtained in the previous steps into a spatial weight distribution that can directly participate in gain scheduling, so that subsequent enhancement is no longer a uniform amplification of all high-frequency regions, but rather differentiated processing based on the strength of texture evidence and the level of defect risk.

[0128] Step S43: Direction matching enhancement control, performing gradient direction analysis on the high-frequency texture component data to extract local structural direction information; matching the local structural direction information with the main direction information determined in the cross-domain consistent texture restoration stage; allowing enhancement when the direction matching degree meets the preset conditions; reducing the enhancement intensity of the corresponding region when the direction deviation exceeds the preset threshold; generating direction control coefficient data;

[0129] In this embodiment, the high-frequency texture component data I is first processed. H Perform local gradient direction analysis to extract local structural orientation information at each spatial location in the current image. Preferably, local structural orientation information is obtained through the Sobel operator, the Scharr operator, or structural tensor analysis; in images of textiles and murals with obvious texture orientation, structural tensor analysis is preferred to improve the stability of main orientation extraction.

[0130] Then, the local structural orientation information Cross-domain consistent structure orientation field obtained from the cross-domain consistent texture recovery stage Perform a matching determination and calculate the directional matching coefficient W. o Preferably, the direction matching coefficient is constructed using the absolute value of the direction difference, and the direction difference is defined as:

[0131] ;

[0132] In the formula, It's a difference in direction. It is local structural orientation information. It is a cross-domain consistent structural orientation field;

[0133] In practical implementation, considering that the texture direction can have 180° symmetry, the direction difference can be further uniformly converted to the range of 0°~90°;

[0134] In a preferred embodiment:

[0135] when When the region is in the height direction matching region, a higher directional control coefficient is assigned.

[0136] when When the region is identified as a medium matching region, the corresponding enhancement coefficient is reduced proportionally.

[0137] when When this occurs, it is determined to be a region of directional deviation, significantly reducing the reinforcement strength;

[0138] when In this case, the corresponding region is preferably not enhanced or only retains the minimum level of security enhancement;

[0139] Preferably, the directional control coefficient W o A piecewise attenuation function can be used to construct the function, with its value ranging from 0.2 to 1.0, so as to ensure that even if the local direction is not perfectly matched, the texture will not be broken due to excessive suppression.

[0140] Furthermore, in an optional embodiment, a lightweight orientation discrimination network can be introduced to assist in correcting the orientation control results. The lightweight orientation discrimination network can adopt a MobileNetV3-small or ShuffleNetV2 structure, performing principal orientation stability discrimination on 16×16 or 32×32 local texture blocks, outputting orientation reliability coefficients, and multiplicatively fusing them with the aforementioned orientation control coefficients. This approach maintains the overall model's lightweight design while improving the robustness of orientation matching in local weak texture regions.

[0141] By performing the above operations, the enhancement process is mainly carried out along the true texture principal direction indicated by the cross-domain consistent structure direction field, thereby effectively avoiding unreasonable high-frequency amplification in non-principal directions by traditional isotropic enhancement.

[0142] Step S44: Safety upper limit control. Based on the regional confidence weight data and the direction control coefficient data, calculate the local enhancement gain value at each spatial location; set a global safety gain upper limit parameter. When the local enhancement gain value exceeds the safety gain upper limit parameter, it is truncated and limited; generate high-frequency gain mapping data and safety threshold parameter data.

[0143] In this embodiment, based on the regional credibility weight data W c With the direction control coefficient data W o A local enhancement gain value G is constructed; preferably, the local enhancement gain value is determined by coupling the base gain and the control factor, and can be expressed as:

[0144] ;

[0145] In the formula, It is the local enhancement gain value. This is the base gain, preferably ranging from 1.00 to 1.10. This is the confidence region adjustment coefficient, with a preferred value of 0.20 to 0.45. This is the direction control adjustment coefficient, preferably ranging from 0.10 to 0.30.

[0146] In a preferred embodiment, it is advisable to =1.05, , This indicates that the dominant source of enhancement is still the weight of the trusted region, and the direction control mainly plays a role in correction and constraint, rather than completely dominating the gain allocation. This is more in line with the stability requirements in actual image enhancement scenarios.

[0147] To prevent abnormally high gain in areas with dense texture or residual noise, a global safety gain upper limit parameter G is further set. maxPreferably, the global security gain upper limit parameter is set to 1.45 to 1.80; for fragile cultural relic images with fine surface textures and high requirements for authenticity, it is preferably controlled below 1.50.

[0148] When the local enhancement gain value at a certain spatial location exceeds the global security gain upper limit parameter, it is truncated, that is: if G>G max Then let the actual gain at that position be G. max If G < 1.0, it is preferable to choose 1.0 or the minimum security gain value to avoid unnecessary texture attenuation.

[0149] Furthermore, to accommodate the detail tolerance of different cultural relic materials, in an optional implementation, a material adaptive safety factor can be introduced; for images of cultural relics with relatively soft surface details such as pottery, murals, and textiles, the safety upper limit can be appropriately reduced; for images with relatively stable shallow engraving textures such as stone carvings and bronze inscriptions, the safety upper limit can be appropriately increased, but generally not exceeding 1.80.

[0150] By performing the above operations, high-frequency gain mapping data and safety threshold parameter data are generated, giving the texture enhancement process a clear basis for amplitude adjustment and a safety limiting mechanism.

[0151] Step S45: Enhanced reconstruction generation: Adjust the high-frequency texture component data and the corresponding local enhancement gain value to obtain enhanced high-frequency component data; reconstruct the enhanced high-frequency component data with low-frequency structure data to generate enhanced cultural relic image data; generate enhancement contribution mapping data according to the change ratio of high-frequency components before and after enhancement; output the texture enhancement result data of historical cultural relic image.

[0152] In this embodiment, the high-frequency texture component data I is first processed according to the local enhancement gain value G. H Position-by-position adjustment was performed to obtain enhanced high-frequency component data. Preferably, the adjustment method adopts multiplicative enhancement, that is, multiplying the high-frequency texture component with the corresponding gain value at each spatial location to maintain the relative distribution relationship of the high-frequency texture.

[0153] After obtaining the enhanced high-frequency component data, it is compared with the low-frequency infrastructure image obtained in step S41. Reconstruction is performed to generate enhanced image data of cultural relics. , can be represented as:

[0154] ;

[0155] In the formula, It is enhanced image data of cultural relics. It is a low-frequency basic structure image. It is high-frequency texture component data;

[0156] After reconstruction, to improve the stability and visual continuity of the output, it is preferable to perform a lightweight edge consistency correction process. This process can employ small-scale guided filtering or local contrast callback to mitigate slight transitional abrupt changes that may occur at the enhancement boundaries, without altering the main enhancement result.

[0157] Simultaneously, enhancement contribution mapping data is constructed based on the ratio of changes in high-frequency components before and after enhancement, used to characterize the strength of enhancement sources and the degree of enhancement participation at each spatial location. Preferably, the enhancement contribution mapping data is generated by the ratio or increment of high-frequency amplitudes before and after enhancement, and output in grayscale or pseudo-color image form to help illustrate the correspondence between enhanced regions and original credible evidence.

[0158] The enhanced texture result data of the historical artifact image specifically includes: enhanced artifact image data, high-frequency gain mapping data, enhancement contribution mapping data, and safety threshold parameter data;

[0159] The enhanced artifact image data is used as the final output image;

[0160] The high-frequency gain mapping data is used to characterize the actual enhancement intensity in each region;

[0161] The enhancement contribution mapping data is used to characterize the correspondence between the enhancement results and the original evidence;

[0162] The safety threshold parameter data is used to record the upper limit control value of the gain during this enhancement process.

[0163] By performing the above operations, this method addresses the problem in existing methods for enhancing the texture of historical artifacts that lack a control mechanism for the characteristics of texture direction. This can easily damage the original texture direction structure during enhancement, leading to disordered texture direction or structural distortion after enhancement. For example, in images of ancient brocade artifacts with obvious fabric texture direction, traditional isotropic enhancement methods may produce unreasonable enhancement in non-principal directions, causing the fabric texture direction to become blurred or distorted. This solution adopts direction matching enhancement control. By extracting local structural direction information and matching it with a cross-domain consistent structural direction field, enhancement is only performed on regions with consistent directions, while enhancement is suppressed on regions with deviated directions. This ensures that the enhancement process proceeds along the true main direction of the texture, achieving consistency between the enhanced texture structure and the original structural direction.

[0164] Meanwhile, existing methods for enhancing the texture of historical artifact images suffer from issues such as a lack of security control over enhancement intensity and untraceable enhancement processes. These methods are prone to over-enhancement in local areas or a lack of reliability verification for enhancement results. For example, in images of stone carvings with dense high-frequency textures, traditional enhancement methods may produce excessively high gains in local areas, resulting in unrealistic enhancement effects and making it impossible to determine the reliability of the enhanced region's source. This solution employs a texture enhancement mechanism that combines the construction of credible region weights with security upper limit control. By constructing region credibility weight data, setting global security gain upper limit parameters, and generating enhancement contribution mapping data, the enhancement process achieves technical effects of controlled intensity and traceable enhancement sources, thereby avoiding over-enhancement and improving the security and credibility of the results.

[0165] Example 6, see Figure 1 and Figure 2 Based on the above embodiments, this embodiment provides an intelligent historical artifact image texture enhancement system, including a data acquisition module, a defect modeling module, and a texture enhancement module.

[0166] The data acquisition module is used for multi-source data acquisition, obtaining a multi-source image data set through multi-source data acquisition, and sending the multi-source image data set to the defect modeling module;

[0167] The defect modeling module is used for joint modeling of defect materials. Through joint modeling of defect materials, it obtains joint modeling result data of defect materials and sends the joint modeling result data of defect materials to the texture enhancement module.

[0168] The texture enhancement module is used for cross-domain consistent texture restoration and texture enhancement of historical artifact images. Through cross-domain consistent texture restoration and texture enhancement of historical artifact images, the texture enhancement result data of historical artifact images is obtained.

[0169] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0170] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0171] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. An intelligent method for enhancing the texture of historical artifact images, characterized in that: The method includes the following steps: Step S1: Multi-source data acquisition to obtain a set of multi-source image data with structure alignment and controlled illumination differences; Step S2: Joint modeling of defective materials. Based on the multi-source image data set, a defect saliency distribution model and a material texture statistical model are constructed. Through joint analysis of spatial and frequency domain features, a defect probability distribution map and material prototype matching parameters are generated, and a structural mutual exclusion constraint relationship is formed to obtain the joint modeling result data of defective materials, including defect probability distribution map data, material texture statistical parameter data, and defect texture mutual exclusion constraint parameter data. The defect probability distribution map data is used to characterize the spatial distribution of cracks, stains, reflections, or peeling areas. The material texture statistical parameter data is used to characterize the spectral distribution characteristics, directional characteristics, and contrast distribution characteristics of the current artifact surface material. The defect texture mutual exclusion constraint parameter data is used to limit the upper limit of high-frequency gain for defective areas during the enhancement stage. Step S3: Cross-domain consistent texture restoration. Based on the multi-source image dataset and the joint modeling results of the defective material, a cross-domain consistent texture field is constructed. A texture screening method combining multi-dimensional feature consistency improvement is used to verify the consistency and energy of local structural responses in images from different sources, generating reliable texture seed regions and obtaining the number of cross-domain consistent texture restoration results. This includes the following steps: Step S31: Local structural orientation extraction. Gradient calculation and local structural orientation analysis are performed on each image in the multi-source image dataset to generate a structural orientation field corresponding to each image. Combined with noise estimation parameter data, a preset orientation threshold is adaptively adjusted. The structural orientation of different images at the same spatial location is compared for consistency, and an orientation consistency index is calculated. When the orientation consistency index is greater than the preset orientation threshold, the location is marked as an orientationally stable region, and the orientation consistency judgment result data is output. Step S32: Local frequency domain phase consistency calculation. Step S33: High-frequency energy ratio stability detection. Step S34: Construction of cross-domain consistency comprehensive score. Step S35: Generation of reliable high-frequency candidate regions. Step S4: Historical artifact image texture enhancement. Based on the cross-domain consistent texture restoration result data and the defect material joint modeling result data, a reliable region control optimized texture enhancement method is used to adjust the direction-selective gain of high-frequency components to obtain historical artifact image texture enhancement result data.

2. The intelligent method for enhancing the texture of historical artifact images according to claim 1, characterized in that: In step S1, the multi-source image data set specifically includes: geometrically aligned image data, radiometrically normalized image data, noise estimation parameter data, and uniform resolution image data.

3. The intelligent method for enhancing the texture of historical artifact images according to claim 2, characterized in that: In step S32, the local frequency domain phase consistency calculation involves performing local frequency domain transformation on each image in the multi-source image dataset to extract the corresponding local phase information; performing consistency superposition calculation on the local phases of different images at the same spatial location to generate a phase consistency index; and when the phase consistency index is greater than a preset phase threshold, marking the location as a phase stable region and outputting the phase consistency judgment result data. In step S33, the high-frequency energy ratio stability detection involves performing frequency domain decomposition on each image in the multi-source image dataset to extract high-frequency energy components; calculating the high-frequency energy ratio stability index between different images at the same spatial location; and marking the location as an energy-stable region when the high-frequency energy ratio stability index is greater than a preset energy ratio threshold, and outputting energy ratio stability data. In step S34, the cross-domain consistency comprehensive score construction is based on the direction consistency judgment result data, the phase consistency judgment result data, and the energy ratio stability data to construct a cross-domain consistency comprehensive score model; a comprehensive consistency score value is generated for each spatial location; the defect probability distribution map data in the defect material joint modeling result data is introduced for a one-vote veto filtering: if the defect probability of a certain spatial location is higher than the preset defect safety threshold, the comprehensive consistency score value of that location is directly set to zero or marked as a non-enhanced region to eliminate pseudo-high frequency interference; when the comprehensive consistency score value is greater than the preset comprehensive threshold, the location is determined as a reliable texture seed region.

4. The intelligent method for enhancing the texture of historical artifact images according to claim 3, characterized in that: In step S35, the reliable high-frequency candidate region is generated. Based on the reliable texture seed region, the original high-frequency components are filtered, and only the high-frequency information in the reliable texture seed region is retained. Generate cross-domain consistent texture seed map data, texture consistency score data, cross-domain consistent structure orientation field data, and reliable high-frequency candidate region data to obtain cross-domain consistent texture recovery result data; The cross-domain consistent texture recovery result data specifically includes: cross-domain consistent texture seed map data, texture consistency score data, cross-domain consistent structure orientation field data, and reliable high-frequency candidate region data.

5. The intelligent method for enhancing the texture of historical artifact images according to claim 4, characterized in that: In step S4, the texture enhancement of the historical artifact image includes the following steps: Step S41: Frequency band separation processing, smoothing filtering processing is performed on the original artifact image to extract the low-frequency basic structure image; based on the difference between the original image and the low-frequency basic structure image, high-frequency texture component data is obtained; low-frequency structure data and high-frequency texture component data are output. Step S42: Constructing Trusted Region Weights. Based on the cross-domain consistent texture seed map data and texture consistency score data in the cross-domain consistent texture recovery result data, and the defect probability distribution map data in the defect material joint modeling result data, construct region trust weight data. The region trust weight data is used to characterize the degree of enhancement permission at each spatial location. When a region simultaneously satisfies the trust seed label and the consistency score is higher than a preset threshold, an enhancement weight is assigned. When the defect probability is higher than the preset threshold, the enhancement weight is suppressed. Step S43: Direction matching enhancement control, performing gradient direction analysis on the high-frequency texture component data to extract local structural direction information; matching the local structural direction information with the main direction information determined in the cross-domain consistent texture restoration stage; allowing enhancement when the direction matching degree meets the preset conditions; reducing the enhancement intensity of the corresponding region when the direction deviation exceeds the preset threshold; generating direction control coefficient data; Step S44: Safety upper limit control. Based on the regional confidence weight data and the direction control coefficient data, calculate the local enhancement gain value at each spatial location; set a global safety gain upper limit parameter. When the local enhancement gain value exceeds the safety gain upper limit parameter, it is truncated and limited; generate high-frequency gain mapping data and safety threshold parameter data. Step S45: Enhanced reconstruction generation: Adjust the high-frequency texture component data and the corresponding local enhancement gain value to obtain enhanced high-frequency component data; reconstruct the enhanced high-frequency component data with low-frequency structure data to generate enhanced cultural relic image data; generate enhancement contribution mapping data according to the change ratio of high-frequency components before and after enhancement; output the texture enhancement result data of historical cultural relic image. The enhanced texture result data of the historical artifact image specifically includes: enhanced artifact image data, high-frequency gain mapping data, enhancement contribution mapping data, and safety threshold parameter data.

6. An intelligent historical artifact image texture enhancement system, used to implement the intelligent historical artifact image texture enhancement method as described in any one of claims 1-5, characterized in that: It includes a data acquisition module, a defect modeling module, and a texture enhancement module.

7. The intelligent historical artifact image texture enhancement system according to claim 6, characterized in that: The data acquisition module is used for multi-source data acquisition, obtaining a multi-source image data set through multi-source data acquisition, and sending the multi-source image data set to the defect modeling module; The defect modeling module is used for joint modeling of defect materials. Through joint modeling of defect materials, it obtains joint modeling result data of defect materials and sends the joint modeling result data of defect materials to the texture enhancement module. The texture enhancement module is used for cross-domain consistent texture restoration and texture enhancement of historical artifact images. Through cross-domain consistent texture restoration and texture enhancement of historical artifact images, the texture enhancement result data of historical artifact images is obtained.

Images