Image signal processing system and processing method

Abstract

The present invention discloses an image signal processing system and method, which relates to the field of image signal processing technology. The system includes extracting structural features from the collected original image signal and generating a region classification mask; predicting the processing path of each region based on the region classification mask; performing corresponding compression operations on each region according to the processing path, and embedding corresponding structural identification information in the compressed data; and performing enhancement processing on key structural regions based on the structural identification information during decoding, thereby improving the image restoration quality. Continuous adjustment of the number of channel retention and the quantization step size is achieved, enhancing the fidelity control capability of complex structural regions. Improving the image restoration quality and detail retention effect, and establishing a closed-loop optimization process between compression and enhancement.

Description

Technical Field

[0001] The present invention relates to the technical field of image signal processing, and in particular to an image signal processing system and a processing method. Background Art

[0002] In the field of image signal processing, compression and enhancement are two key processes for improving image storage efficiency and visual quality. Existing image compression methods primarily employ a uniform compression ratio or employ compression configuration based on global statistical features, seeking a balance between coding efficiency and visual quality. However, these methods often fail to tailor compression strategies to regions with significant structural differences or complex details. This can lead to problems such as loss of detail in key areas or insufficient compression in low-structure regions, compromising image reconstruction quality and subsequent processing accuracy.

[0003] At the same time, image enhancement technology often operates as a standalone process, typically performed after decoding through methods such as edge filtering, super-resolution interpolation, or residual compensation. Due to the lack of structural semantics during the compression phase, the enhancement module cannot obtain structural information from the image compression process and can only perform secondary estimation based on the decoded data. This can lead to localized jumps in the enhancement effect, excessive edges, or increased artifacts, making it difficult to ensure the stability and consistency of the restoration quality.

[0004] Furthermore, existing methods often rely on fixed rules to divide image regions and employ static strategies for channel control and quantization configuration, lacking pixel-level structural awareness and adaptive adjustment capabilities. Traditional compression methods are unable to effectively allocate resources to preserve detail in structurally complex or prominent areas, and they also struggle to dynamically improve compression rates and reduce redundant data in background areas. Summary of the Invention

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] Therefore, the technical problem solved by the present invention is: the existing image compression and enhancement methods have the problems of weak structural perception ability, disconnection between compression and enhancement, easy distortion of key areas, and low compression efficiency in low-structure areas, as well as how to achieve pixel-level path prediction guided by structural features, dynamic regulation of compression parameters and cross-stage collaborative transmission of structural information.

[0007] To solve the above technical problems, the present invention provides the following technical solutions: an image signal processing system, comprising:

[0008] The image acquisition module includes an image acquisition unit and a preprocessing unit, which is used to acquire the original image and perform noise suppression and normalization;

[0009] A structural modeling module includes a feature extraction unit and a region division unit, wherein the feature extraction unit is used to construct a structural feature vector and the region division unit is used to generate a region classification mask;

[0010] The path prediction module includes a partition analysis unit and a path allocation unit, which is used to extract the structural priority distribution characteristics, neighborhood variability and boundary correlation to generate a path mapping matrix;

[0011] A compression execution module includes a structure density calculation unit and a compression control unit, wherein the structure density calculation unit is used to weightedly fuse and normalize the structure priority score, structure complexity index, and boundary correlation index of each pixel to construct a structure density function and output a structure density response value; the compression control unit is used to determine compression parameters based on the structure density response value and execute a compression strategy corresponding to the path, wherein the compression parameters include an integer value corresponding to the number of channels retained and an integer value corresponding to the quantization step size, wherein the number of channels retained is obtained by multiplying the structure density function value by a preset maximum number of channels and rounding down, and the quantization step size is obtained by multiplying the complement of the structure density function value by a preset maximum quantization precision and rounding down, wherein the complement value is 1 minus the structure density function value;

[0012] a structure embedding module, comprising an embedding construction unit and a data alignment unit, wherein the embedding construction unit is used to generate a structure embedding vector, and the data alignment unit is used to integrate the structure information with the compressed data and the path information at the pixel level;

[0013] The enhanced processing module includes an identification parsing unit and an enhanced execution unit. The identification parsing unit is used to extract structure identification information and path type, and the enhanced execution unit is used to perform enhanced processing of key structural areas.

[0014] As a preferred solution of the image signal processing method of the present invention, wherein: structural features are extracted from the collected original image signal to generate a region classification mask;

[0015] predicting a processing path for each region based on the region classification mask;

[0016] According to the processing path, a corresponding compression operation is performed on each region, and corresponding structure identification information is embedded in the compressed data;

[0017] During decoding, enhancement processing is performed on key structural areas based on the structural identification information, thereby improving image restoration quality.

[0018] As a preferred embodiment of the image signal processing method of the present invention, the extracting structural features includes performing noise suppression and normalization on the collected original image signal to obtain a pre-processed image;

[0019] Extracting multi-scale edge response, multi-directional texture intensity and frequency domain residual saliency heat map from the preprocessed image respectively, and combining the edge response, texture intensity and saliency heat map as a structural feature vector of the pixel;

[0020] Generating the region classification mask includes calculating structural field similarity between pixels; the structural field similarity is calculated by using a cosine similarity algorithm on the structural feature vectors of any two pixels; selecting pixels whose feature response values ​​are in the top X percent of all pixels as initial cores, the feature response values ​​being determined by a weighted average calculation of the structural feature vectors to represent the strength or significance of the structural response of the pixels;

[0021] For each structural consistency region, the regional significance mean and structural feature variance are calculated, and a structural priority score is assigned to all pixels in the region;

[0022] According to the structure priority score, pixels are graded by adaptive thresholding to generate region classification masks that classify pixels into structure-critical regions, general regions, or background regions.

[0023] As a preferred embodiment of the image signal processing method of the present invention, the predicting of the processing path of each region includes executing a structure-guided path allocation rule based on the distribution characteristics of the structural priority, neighborhood variability, and boundary correlation;

[0024] The distribution characteristic of the structural priority is to extract the structural partition to which each pixel belongs, and to count the discrete degree between the structural priorities of all pixels in the structural partition and the mean value of the structural priority in the partition;

[0025] The neighborhood variability represents the amount of change in the structural priority of pixels in the neighborhood centered on the current pixel relative to the structural priority of the current pixel.

[0026] The boundary correlation is to calculate, for a boundary pixel in the structural key area, a structural priority difference between the boundary pixel and pixels in the neighborhood whose labels do not belong to the structural key area.

[0027] As a preferred embodiment of the image signal processing method of the present invention, the path assignment rule includes, when a pixel label in the region classification mask is a structural key region, extracting the structural partition to which the pixel belongs, statistically analyzing the distribution dispersion of the structural priority, calculating the variability of the structural priority within the pixel neighborhood, and detecting its structural correlation with adjacent regions;

[0028] When the distribution dispersion reaches a structural complexity threshold, or the structural correlation reaches a boundary correlation threshold, allocating an enhanced compression path;

[0029] When the distribution dispersion is lower than a structural complexity threshold and the structural correlation is lower than a boundary correlation threshold, assigning a denoising and enhancement post-compression path;

[0030] When the pixel label is a general area, a post-enhancement compression path or a compression path is assigned according to whether the neighborhood structure priority variability reaches a detail enhancement threshold;

[0031] When the pixel label is a background area, a post-denoising compression path is assigned.

[0032] As a preferred embodiment of the image signal processing method of the present invention, the compression operation includes constructing a structure density function based on the output of the path allocation rule, combining the structure priority score, structure complexity index and boundary correlation index of each pixel;

[0033] Determining the compression parameter by using a structural density response value output by the structural density function;

[0034] The compression parameters include an integer value corresponding to the number of channels to be retained and an integer value corresponding to the quantization step size.

[0035] The number of channels retained is obtained by multiplying the structure density function value by the preset maximum number of channels and discarding the decimal part to take the integer, and the quantization step is obtained by multiplying the complement of the structure density function value by the preset maximum quantization accuracy and discarding the decimal part to take the integer, wherein the complement is 1 minus the structure density function value.

[0036] As a preferred solution of the image signal processing method of the present invention, wherein: structure identification information is synchronously constructed during the compression process, and the structure identification information includes a structure density response value, a compression parameter corresponding to the selected path type, and compression residual information;

[0037] The structure identification information is integrated with the compressed data and the path mapping matrix in the form of an embedded vector and aligned according to pixel indices to form a compressed output set.

[0038] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement steps of a method for processing an image signal.

[0039] A computer-readable storage medium stores a computer program, which implements the steps of a method for processing image signals when executed by a processor.

[0040] Beneficial effects of the present invention: The image signal processing method provided by the present invention can achieve fine division of structural key areas, general areas and background areas in the image by introducing structural feature modeling and regional classification masks, thereby improving processing granularity and perception accuracy. Combining structural priority distribution, neighborhood variability and boundary correlation, a path prediction mechanism is constructed to realize dynamic scheduling of processing paths. A structural density function is constructed as the compression parameter control core to achieve continuous adjustment of the number of channel retention and quantization step size, thereby enhancing the fidelity control capability of complex structural areas. During the compression process, a structural embedding vector is generated synchronously, and the structural response, path type and residual information are integrated into the compressed data to achieve end-to-end transmission of structural information. The decoding stage can perform targeted enhancement processing based on the structural identification information, perform residual compensation or structural sensitive recovery in the structural key areas, thereby improving the image restoration quality and detail retention effect, and building a closed-loop optimization process between compression and enhancement. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0042] Figure 1 This is an overall flow chart of an image signal processing method provided by the first embodiment of the present invention. DETAILED DESCRIPTION

[0043] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0044] Example 1, reference Figure 1 , as an embodiment of the present invention, provides a method for processing an image signal, comprising:

[0045] S1: Extract structural features from the collected original image signal and generate a region classification mask.

[0046] The original image signal acquired As input, the image is first subjected to noise suppression and normalization to obtain a preprocessed image. .exist Calculate the multi-scale edge response , multi-directional texture strength And the frequency domain residual significance heat map For each pixel, 、 、 Combination as feature vector :

[0047] ;

[0048] Where x represents the horizontal coordinate of the pixel; y represents the vertical coordinate of the pixel; Indicates the two-dimensional spatial position of the current pixel in the image; Indicates the horizontal and vertical coordinates of the first pixel involved in the structural similarity calculation; Indicates the horizontal and vertical coordinates of the second pixel participating in the structural similarity calculation.

[0049] For all pixel feature vectors, calculate the structural field similarity one by one:

[0050] ;

[0051] in, Indicates location and location The similarity of structural features, Indicates location The structural eigenvector at Indicates location The structural eigenvector at express The transpose of .

[0052] The pixels whose feature response values ​​are in the top X percent (such as the top 5%) of all pixels are selected as the initial core, based on the similarity threshold Perform density adaptive clustering and expand the structural consistency areas in sequence , each pixel belongs to only one cluster area. For each cluster area , calculate the significance mean of the region and structural characteristic variance , and assign structural priority scores to all pixels in the area. The formula is expressed as:

[0053] ;

[0054] in, Represents pixels Structural priority score, Indicates area The significant mean of Indicates area The variance of the internal structural features, represents the maximum value of the structural feature variance in all regions, represents pixel coordinates, Indicates the A structural partition.

[0055] by Based on this, by setting the threshold 、 Each pixel is divided into structural key area, general area or background area to obtain the area classification mask :

[0056] ;

[0057] in, Represents pixel points The regional classification label is one of 2, 1, and 0; Indicates the adaptive segmentation threshold between the key structural area and the general area; Represents the adaptive segmentation threshold between general area and background area; 2 represents the key structural area, 1 represents the general area, and 0 represents the background area.

[0058] Finally, the region classification mask is obtained The size of the mask matrix is ​​consistent with the input image, which provides a partition basis for subsequent adaptive path prediction and compression processing based on the structure field.

[0059] During structural feature extraction and region classification mask generation, noise suppression and normalization of the original image signal effectively eliminate environmental interference and signal deviations during acquisition, providing a unified foundation for subsequent structural feature segmentation. Multi-scale edge response, texture intensity, and frequency-domain saliency features are jointly extracted within the same image space, significantly improving the resolution of local contours, texture patterns, and salient regions. A vectorized description based on structural field features enables comprehensive quantification of complex structural distributions, laying the foundation for subsequent spatial partitioning and structural recognition.

[0060] Furthermore, during the structural similarity calculation and density-adaptive clustering stages, the cosine similarity of structural feature vectors is used to measure the spatial structural relationships between pixels, replacing the traditional methods of linear weighted fusion and simple threshold segmentation. This mechanism supports the dynamic coupling expression of high-dimensional structural features. Through iterative clustering, it automatically merges sets of pixels with highly consistent structures into adaptive regions, achieving precise segmentation and deformable partitioning of the image's local structural continuity. This process overcomes the problems of conventional partitioning methods based on single features or fixed templates, which suffer from insufficient generalization and poor robustness when dealing with multi-scale and multi-type structures.

[0061] Furthermore, during the hierarchical mask generation phase, the importance and complexity of regions can be adaptively ranked through a comprehensive score of the structural significance mean and cluster variance. The final output regional classification mask not only accurately distinguishes between key structural areas, general areas, and background areas, but also dynamically adapts to image inputs of different types and structural complexities. The overall solution breaks through the barriers of traditional image partitioning, which can only process regular areas and has difficulty adapting to structural diversity. It achieves a full-process innovation from pixel-level feature expression, spatial structure clustering, to multi-level differentiation, effectively improving the accuracy and robustness of subsequent path prediction and structural adaptive compression processing.

[0062] S2: Predicting a processing path for each region based on the region classification mask.

[0063] First, for each pixel , according to the mask label Distinguish the region type to which it belongs. For the key structural region, further analyze the degree of local structural change in the key structural region. The neighborhood of the center Compute the local variability of structural priorities within:

[0064] ;

[0065] in, Represents pixels The local structural priority variability of N Indicated in pixels The set of neighborhood pixels centered on ; Represents pixels in the neighborhood Structural priority score of Indicates the current pixel Structural priority score.

[0066] In the structural key area, for each pixel, extract the cluster partition to which it belongs , and based on the discrete degree of the structural priority distribution within the same partition, the structural complexity index is calculated:

[0067] ;

[0068] in, Indicates location Structural complexity score of the pixel at ; Representation and Pixels The structural clustering regions of the same category are indexed as ; Represents the structural area The total number of pixels in the set (i.e., the cardinality of the set); Represents cluster areas Any pixel position index in ; Represents pixels Structural priority score at Represents the structural area The average of all pixel structure priorities within .

[0069] For pixels at the boundary of the structural key area , that is, there are pixels with labels other than 2 in its neighborhood. We further introduce the structural correlation index. This index measures the difference in structural features between boundary pixels and adjacent areas. It is specifically defined as follows:

[0070] ;

[0071] ;

[0072] in, Indicated in pixels The set of all pixels whose labels are not equal to 2 in the center and neighborhood; Represents the horizontal and vertical coordinates of any pixel point in the neighborhood; Indicated in pixels The neighborhood set centered on Represents pixel points Regional classification labels; Indicates that the pixel does not belong to the key structural area; Represents pixels Structural relevance index of Indicates taking the maximum value of all items in the collection; Indicates selecting pixels from the set of pixels in the neighborhood non-structural key area.

[0073] Pixels with high boundary correlation are more likely to have cross-regional structural breaks or detail mismatches, and should be given higher enhancement and collaborative processing priority when allocating paths.

[0074] Define the structural complexity threshold during the path assignment process , determines the enhancement requirements of complex structural areas; boundary correlation threshold , distinguish high-variability boundaries from ordinary boundaries; general regional priority variability threshold , used for detail compensation.

[0075] These thresholds can be set through partition statistics, image scene experience, or adaptive optimization.

[0076] For each pixel, , combining the above structural indicators and regional labels, the signal processing path is allocated according to the following rules:

[0077] If satisfied, , and at the same time satisfy, ,or, , it indicates that the pixel is located in an area with complex structure or more sudden boundary changes. Its details are enhanced first and then compressed. The allocation path is "enhancement followed by compression".

[0078] If satisfied, , and at the same time satisfy, ,and, , indicating that the structure distribution is stable. To prevent the weak structure texture from being distorted during compression, denoising is performed first, followed by enhancement and compression. The allocation path is "denoising, enhancement, and then compression."

[0079] If satisfied, ,and, , indicating that the details in the general area are complex and the fidelity before compression should be improved through enhancement, and the assignment path is "enhancement followed by compression".

[0080] If satisfied, ,and, ,The structural changes are limited, and the compression operation is directly performed, and the path is assigned as “compression”.

[0081] If satisfied, , that is, the pixel is located in the background area and usually does not contain significant structural information. To improve compression efficiency, denoising is performed first to reduce redundancy, and then compression is performed. The allocation path is "denoising followed by compression".

[0082] The processing path of each pixel is written into the path assignment matrix in the task identification sequence. , the matrix size is consistent with the original image, which serves as the scheduling basis for subsequent compression and enhancement processes, and realizes fine control and dynamic allocation of paths based on structural features.

[0083] By constructing two structural behavior metrics—structural complexity and boundary correlation—the team can further refine the image structure within the same region label, avoiding the problem of traditional region classification methods lacking recognition and response to internal structural features after labeling. A joint approach that combines local variability, cluster distribution dispersion, and cross-region priority differences improves the path prediction process's sensitivity to complex textures, edge attenuation, and multi-scale structures, providing a higher-resolution basis for assigning compression and enhancement strategies.

[0084] Furthermore, during path allocation, a structure-guided five-level task flow is introduced, enabling flexible switching of signal processing paths based on the functional roles of different regions within the image structure field. Compared to traditional strategies that only support static processing sequences of "compression" or "compression + enhancement," this mechanism supports multiple combinations such as "denoising → enhancement → compression" and "enhancement → compression," and can dynamically adjust path levels within regions. This path matrix construction, driven by structural judgment results, enables pixel-level processing flow scheduling and binding of execution units, providing the logical foundation for subsequent adaptive compression.

[0085] Furthermore, when addressing the intersection of boundary regions and general areas, correlation analysis based on maximum priority differences effectively addresses reconstruction artifacts caused by inter-regional path breakage and structural discontinuities in existing methods. A collaborative enhancement strategy driven by cross-regional structural differences enables detailed consistency compensation between complex and adjacent regions before compression. This overcomes the limitations of traditional path allocation strategies, which suffer from fixed strategies, delayed responses, and rigid processing at region boundaries, and enables the ability to compress and enhance collaborative scheduling in structurally continuous regions.

[0086] S3: According to the processing path, perform corresponding compression operations on each region, and embed corresponding structure identification information in the compressed data.

[0087] Combined structural priority scoring , structural complexity index Correlation index with boundary , construct a structure-driven compression mapping function and implement differentiated compression strategies according to different path types. First, in order to dynamically respond to the compression fidelity requirements of the structural region, a structural density function is introduced. , used to control channel retention rate and quantization accuracy:

[0088] ;

[0089] in, Represents pixels The structural response density at ; Represents the Sigmoid function; 、 、 Represents structural priority , structural complexity Correlation with boundaries Weight coefficient for structural density; Represents the structural priority score of the pixel; Indicates the structural complexity index within the partition to which it belongs; Indicates the structural boundary correlation between the point and the adjacent area.

[0090] In the actual compression process, σ calculates the number of compression channels and quantization step size based on μ(x, y):

[0091] ;

[0092] in, Represents pixels The corresponding number of compression channels retained; Indicates the corresponding quantization step size; Indicates the maximum number of compression channels that can be allocated; Indicates the maximum configurable quantization accuracy; Indicates a rounding-down operation. This mechanism allows regions with high structural density to retain more feature channels and reduce quantization error, while regions with low structural density automatically shrink the coding capacity to improve compression efficiency.

[0093] The compression module uses a split execution strategy for different path types. If the path is "enhancement followed by compression" or "denoising, enhancement followed by compression," it prioritizes preserving more structural feature dimensions before compression and uses a lower quantization step size. If the path is "compression" or "denoising followed by compression," the compression ratio is increased, while the number of retained channels is reduced to improve the overall compression rate.

[0094] The structure embedding vector is constructed synchronously during the compression process:

[0095] ;

[0096] in, Represents the output value of the structure density function, represents the channel compression residual, represents the quantization error response.

[0097] The embedding vector is written in full form in the key area of ​​the structure and in compressed form in the general area. The embedding field can be omitted in the background area to reduce data redundancy.

[0098] The final output consists of compressed coded data, structural embedded information and the path mapping matrix In the compression process, each data is aligned and integrated according to the pixel position. In order to unify the output content, the structure compression output set is defined as follows:

[0099] ;

[0100] in, Indicates the compressed encoding result, s is the structure embedding vector, Indicates the path information.

[0101] Introducing a structural density function during the compression process effectively addresses the problem of traditional image compression methods being unable to adjust compression parameters based on structural differences in image regions. Traditional methods often employ a uniform compression rate or compression based on static region division rules, resulting in over-compression in key structural regions and under-compression in low-structure regions, impacting the overall image restoration quality. By constructing a structural density function with structural priority scores, structural complexity, and boundary relevance as inputs, the structural response value of each pixel position can be continuously calculated. Based on this, the number of channels retained and the quantization step size can be dynamically adjusted, enabling the compression process to adapt to image structural features in a fine-grained manner and achieving a coordinated adjustment between compression strength and image structural complexity.

[0102] Based on compression parameter control, a structure embedding vector mechanism is introduced to avoid the problem of structural information being completely discarded during the compression process. Traditional compression methods typically only retain pixel encoding information, and structural semantics must be re-estimated at the decoding end, resulting in a lack of path and structural prior support during the enhancement phase. The structure embedding vector consists of three parts: the structure density function value, the channel compression residual, and the quantization error. It can record the response relationship between compression behavior and structural features, and is integrated with the compressed data for encoding. It directly provides a structural enhancement reference on the decoding side, improving the targetedness and consistency of structural recovery.

[0103] The compressed output data uses a pixel-aligned combination of a path mapping matrix, compression results, and structural embedding vectors, effectively breaking through the problem of the traditional separate compression-enhancement processing chain. In the original scheme, the compression and enhancement processes were independent of each other, lacking a cross-stage structural semantic transfer mechanism, making it difficult to build a structural field-driven enhancement logic. By unifying the compressed output structure, the combined output of path, structural density, and compression behavior is achieved, providing continuous structural semantic input for subsequent enhancement modules, giving the entire image processing chain structural adaptive processing capabilities and making the enhanced restoration effect more refined and stable.

[0104] S4: During decoding, enhancement processing is performed on key structural areas according to the structural identification information, thereby improving image restoration quality.

[0105] Extract the structural embedding vector for each pixel position , which contains the structural density response value, channel compression residual and quantization error information recorded during the compression process. , the structural fidelity of the pixel in the compression stage is estimated as the initial judgment basis for whether the current pixel enters the enhancement processing flow.

[0106] For satisfaction The structural key area pixels are further combined with the structural density response value and the path label , determine the strength level of the enhancement strategy. If the path type is "enhancement followed by compression" or "denoising, enhancement followed by compression", it means that the original structure has been enhanced before compression. In this case, the decoding enhancement should prioritize residual recovery to compress the residual information. and quantization error information As the reference value, correct the main decoding channel output; if the path type is "compression" or "denoising and compression", it means that the area has not been strongly enhanced, then enable the structure-sensitive filter to reconstruct local details and use It is a guiding factor that dynamically adjusts the response range of the enhancement core.

[0107] The enhancement process uses a pixel-level adaptive filtering module. For each pixel in the key structural area, the enhancement method is selected according to the following logic:

[0108] When the structural density Higher than the set enhancement start threshold , a significant enhancement of structural response guidance is performed.

[0109] When the structure density value is in the middle range, the channel decoding error is compensated by combining the compressed residual information.

[0110] When the structural density value is lower than the set threshold, the enhancement process is skipped and only conventional interpolation restoration is performed.

[0111] The enhanced pixel values ​​are weightedly fused with the original decoded pixel values. The fusion weights are controlled by the output of the structure density function, achieving a joint update of compressed residual restoration and structure-guided enhancement. The enhanced result maintains the same size as the original image, and the enhanced image frame is output pixel by pixel.

[0112] The final enhanced output is dominated by structural key areas, reconstructing texture, contour and edge features in significant areas, effectively offsetting the loss of high-frequency information during the compression process, and improving the structural fidelity and detail quality of the overall restored image.

[0113] Example 2 is an embodiment of the present invention, which provides a method for processing image signals. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0114] Sixty high-resolution images (1024×768) from the COST2100 visual scene database, including images of people, buildings, and natural environments, were selected as the evaluation sample set. The experimental platform utilized an NVIDIA RTX 3080 GPU and an Intel i9-12900KF processor, with Python and CUDA for algorithm acceleration. The original images were first processed by the image acquisition module, where they were subjected to YUV channel separation and brightness normalization. After image resizing, they were fed into the structure modeling module. The structural feature extraction unit computed the multi-scale gradient orientation distribution, texture coherence, and frequency domain saliency response of the image to generate a structural feature vector. The region partitioning unit used a density clustering method (DBSCAN) combined with a structural field similarity metric to generate a structural classification mask, separating key structural regions, general regions, and background regions.

[0115] The path prediction module constructs a three-dimensional feature tensor using the variance of the structural priority score, the neighborhood deviation coefficient, and the edge transition amplitude. Combined with a heuristic partitioning scoring function, it calculates a per-pixel path map. The compression execution module uses a structural density function to adjust channel allocation and quantization bit width, controlling the per-pixel compression strength and forming a dynamic compression map. During the compression process, a structural embedding vector is generated and integrated into the output bitstream. After decoding, the enhancement module performs residual inversion and saliency-aware enhancement based on the path type and structural density response. The entire process batch processes 60 images, with an average processing time of 1.26 seconds per image.

[0116] Average original image size: 746.21KB.

[0117] Traditional mean block compression method (fixed path): PSNR: 27.84dB, SSIM: 0.8013, average PSNR of structural key areas: 25.66dB, file size: 132.15KB.

[0118] WebP image compression algorithm (Google): PSNR: 29.92dB, SSIM: 0.8489, average PSNR of structural key areas: 27.11dB, file size: 124.50KB.

[0119] The structure-guided system: PSNR: 31.34dB, SSIM: 0.8796, average PSNR of the structure key area: 29.76dB, file size: 120.03KB.

[0120] The average improvement after image enhancement module processing (structural key areas): PSNR increased by 2.65dB, and edge restoration accuracy increased by 14.32%.

[0121] The compression rate is improved by 3.59% compared with the WebP algorithm, and the detail recognition score is improved by 18.45% after enhancement.

[0122] Experimental results demonstrate that the structure-field-guided compression enhancement system outperforms existing mainstream approaches in terms of compression ratio, overall image quality, and detail recovery in structural areas. Traditional methods using a fixed compression path perform poorly in recovering details in key areas, suffer from severe edge information loss, and blur textures in areas of structural transitions. While the WebP algorithm utilizes lossy compression optimization, it does not model image structural features, which can lead to fragmented details in high-frequency structural areas.

[0123] The system uses a structural modeling module to identify regions of image structure and, in conjunction with a path prediction mechanism, assigns optimal processing chains to different regions, avoiding the structural confusion caused by "uniform compression." Adjusting compression parameters under the control of a structure density function increases the number of channels and converges the quantization step size in regions with dense structures, significantly improving structural fidelity. The structure embedding mechanism simultaneously encapsulates structural semantics and compression behavior during the compression phase. The enhancement module accurately performs targeted repairs, avoiding the miscompensation caused by distorted structural position estimates by traditional enhancement algorithms.

[0124] In particular, in areas with fine structures like building edges and human hair, the enhancement module guides the filtering range and intensity based on the structural density response, achieving layered detail restoration and effectively resolving edge distortion and image artifacts caused by compression. While maintaining an improved compression rate, the overall system achieves a 2.65dB improvement in PSNR for structural areas compared to WebP, and a nearly 18% increase in subjective detail scores, demonstrating the effectiveness and technological foresight of the structure-driven mechanism in the coordinated optimization of image compression and enhancement.

[0125] Embodiment 3, an embodiment of the present invention, provides an image signal processing system, including:

[0126] The image acquisition module includes an image acquisition unit and a preprocessing unit, which is used to acquire the original image and complete noise suppression and normalization.

[0127] The structural modeling module includes a feature extraction unit and a region division unit. The feature extraction unit is used to construct a structural feature vector, and the region division unit is used to generate a region classification mask.

[0128] The path prediction module, including the partition analysis unit and the path allocation unit, is used to extract the structural priority distribution characteristics, neighborhood variability and boundary correlation, and generate the path mapping matrix.

[0129] The compression execution module includes a structural density calculation unit and a compression control unit. The structural density calculation unit is used to construct a structural density function and output a structural density response value. The compression control unit is used to calculate compression parameters and execute a path corresponding compression strategy.

[0130] The structure embedding module includes an embedding construction unit and a data alignment unit. The embedding construction unit is used to generate a structure embedding vector, and the data alignment unit is used to integrate the structure information with the compressed data and the path information at the pixel level.

[0131] The enhanced processing module includes an identification parsing unit and an enhanced execution unit. The identification parsing unit is used to extract structure identification information and path type, and the enhanced execution unit is used to perform enhanced processing of key structural areas.

[0132] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the method of the present invention. The aforementioned storage medium includes various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0133] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0134] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting, or processing it in another suitable manner as necessary, and then storing it in a computer memory.

[0135] It should be understood that various aspects of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the aforementioned embodiments, multiple steps or methods may be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, implementation may be achieved using any combination of the following technologies known in the art: discrete logic circuits having logic gates for implementing logic functions on data signals, application-specific integrated circuits having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), and the like. It should be noted that the above embodiments are intended only to illustrate the technical solutions of the present invention and are not intended to be limiting. While the present invention has been described in detail with reference to preferred embodiments, those skilled in the art will appreciate that modifications or equivalent substitutions may be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and such modifications are intended to be encompassed by the claims of the present invention.

[0136] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A system for processing an image signal, characterized in that: include: The image acquisition module includes an image acquisition unit and a preprocessing unit, which is used to acquire the original image and perform noise suppression and normalization; A structural modeling module includes a feature extraction unit and a region division unit, wherein the feature extraction unit is used to construct a structural feature vector and the region division unit is used to generate a region classification mask; The path prediction module includes a partition analysis unit and a path allocation unit, which is used to extract the structural priority distribution characteristics, neighborhood variability and boundary correlation to generate a path mapping matrix; A compression execution module includes a structure density calculation unit and a compression control unit, wherein the structure density calculation unit is used to weightedly fuse and normalize the structure priority score, structure complexity index, and boundary correlation index of each pixel to construct a structure density function and output a structure density response value; the compression control unit is used to determine compression parameters based on the structure density response value and execute a compression strategy corresponding to the path, wherein the compression parameters include an integer value corresponding to the number of channels retained and an integer value corresponding to the quantization step size, wherein the number of channels retained is obtained by multiplying the structure density function value by a preset maximum number of channels and rounding down, and the quantization step size is obtained by multiplying the complement of the structure density function value by a preset maximum quantization precision and rounding down, wherein the complement value is 1 minus the structure density function value; a structure embedding module, comprising an embedding construction unit and a data alignment unit, wherein the embedding construction unit is used to generate a structure embedding vector, and the data alignment unit is used to integrate the structure information with the compressed data and the path information at the pixel level; The enhanced processing module includes an identification parsing unit and an enhanced execution unit. The identification parsing unit is used to extract structure identification information and path type, and the enhanced execution unit is used to perform enhanced processing of key structural areas.

2. A method for processing an image signal, characterized in that: The image signal processing system according to claim 1 comprises the following steps: Extract structural features from the collected original image signal and generate a region classification mask; predicting a processing path for each region based on the region classification mask; According to the processing path, a corresponding compression operation is performed on each region, and corresponding structure identification information is embedded in the compressed data; During decoding, enhancement processing is performed on key structural areas based on the structural identification information, thereby improving image restoration quality.

3. The image signal processing method according to claim 2, wherein: The extracting of structural features includes performing noise suppression and normalization on the collected original image signal to obtain a preprocessed image; Extracting multi-scale edge response, multi-directional texture intensity and frequency domain residual saliency heat map from the preprocessed image respectively, and combining the edge response, texture intensity and saliency heat map as a structural feature vector of the pixel; Generating the region classification mask includes calculating structural field similarity between pixels based on the structural feature vector; the structural field similarity is calculated by using a cosine similarity algorithm on the structural feature vectors of any two pixels; selecting pixels whose feature response values ​​are in the top X percent of all pixels as initial cores, the feature response values ​​being determined by weighted average calculation of the structural feature vectors to represent the strength or significance of the structural response of the pixels; For each structural consistency region, the regional significance mean and structural feature variance are calculated, and a structural priority score is assigned to all pixels in the region; According to the structure priority score, pixels are graded by adaptive thresholding to generate region classification masks that classify pixels into structure-critical regions, general regions, or background regions.

4. The method for processing an image signal according to claim 3, wherein: The predicting of processing paths for each region includes executing a structure-guided path assignment rule based on the distribution characteristics of structural priorities, neighborhood variability, and boundary correlation; The distribution characteristic of the structural priority is to extract the structural partition to which each pixel belongs, and to count the discrete degree between the structural priorities of all pixels in the structural partition and the mean value of the structural priority in the partition; The neighborhood variability represents the amount of change in the structural priority of pixels in the neighborhood centered on the current pixel relative to the structural priority of the current pixel. The boundary correlation is to calculate, for a boundary pixel in the structural key area, a structural priority difference between the boundary pixel and pixels in the neighborhood whose labels do not belong to the structural key area.

5. The method for processing an image signal according to claim 4, wherein: The path assignment rule includes, when a pixel label in the region classification mask is a structural key region, extracting the structural partition to which the pixel belongs, statistically analyzing the distribution dispersion of the structural priority, calculating the variability of the structural priority within the pixel neighborhood, and detecting its structural correlation with adjacent regions; When the distribution dispersion reaches a structural complexity threshold, or the structural correlation reaches a boundary correlation threshold, allocating an enhanced compression path; When the distribution dispersion is lower than a structural complexity threshold and the structural correlation is lower than a boundary correlation threshold, assigning a denoising and enhancement post-compression path; When the pixel label is a general area, a post-enhancement compression path or a compression path is assigned according to whether the neighborhood structure priority variability reaches a detail enhancement threshold; When the pixel label is a background area, a post-denoising compression path is assigned.

6. The method for processing an image signal according to claim 5, wherein: The compression operation includes constructing a structural density function based on the output of the path allocation rule and combining the structural priority score, structural complexity index and boundary correlation index of each pixel; Determining the compression parameter by using a structural density response value output by the structural density function; The compression parameter includes an integer value corresponding to the number of channels to be retained and an integer value corresponding to the quantization step size; The number of channels retained is obtained by multiplying the structure density function value by the preset maximum number of channels and discarding the decimal part to take the integer, and the quantization step is obtained by multiplying the complement of the structure density function value by the preset maximum quantization accuracy and discarding the decimal part to take the integer, wherein the complement is 1 minus the structure density function value.

7. The method for processing an image signal according to claim 6, wherein: Synchronously constructing the structure identification information during the compression process, the structure identification information including the structure density response value, the compression parameter corresponding to the selected path type, and the compression residual information; The structure identification information is integrated with the compressed data and the path mapping matrix in the form of an embedded vector and aligned according to pixel indices to form a compressed output set.

8. The method for processing an image signal according to claim 7, wherein: The enhancement processing includes extracting structure identification information and path mapping matrix during decoding, and determining whether to perform enhancement processing in combination with the region label corresponding to the pixel; When the path type is compression followed by enhancement or denoising and compression followed by enhancement, error compensation enhancement processing is performed based on the compression residual information and quantization error information in the structure identification information; When the path type is compression or compression after denoising, performing structure-sensitivity enhancement processing based on the structure density information in the structure identification information, and adjusting the application range of the enhancement processing according to the structure density response value; The enhancement result is fused with the decoded image at the pixel level, the fusion weight is determined by the structural density response value, and the enhanced image frame is output.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the image signal processing method according to any one of claims 2 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the image signal processing method according to any one of claims 2 to 8 are implemented.

Images