A Data Processing Method for Inkjet Printing of Calligraphy and Painting Reproduction on Xuan Paper

Abstract

This application discloses a data processing method for inkjet printing of calligraphy and painting reproductions on Xuan paper. The method includes: reading paper attribute parameters according to Xuan paper type information; preprocessing and structural region identification of digital images of calligraphy and painting to obtain a skeleton line set, boundary distance map, boundary sensitive area mask, pore mask, and slow-release spread area mask; then generating a basic coverage map, a neighborhood expected coverage map, and a local orientation map by combining the channel configuration of the printing device; further generating a diffusion risk map; and performing channel reallocation and partition grid rearrangement on the basic coverage map accordingly to obtain a target inkjet dot map and target grid data, which are then output to the inkjet printing device.

Description

Technical Field

[0001] This application relates to the field of printing data processing technology for Xuan paper, specifically to a data processing method for inkjet printing of calligraphy and painting reproductions on Xuan paper. Background Technology

[0002] Current inkjet printing technologies for reproducing calligraphy and paintings typically focus on comprehensive color reproduction, channel ink matching, and conventional rasterization. The substrates are mostly relatively stable materials such as coated paper, photographic paper, or ordinary art paper. In these solutions, print data is generally generated based on the pixel color values ​​of the input image to determine the coverage of each channel, and then combined with conventional halftone or dot control rules to complete the output. Other inkjet printing control solutions focus on switching between different print data based on print area matching, or on improving jetting timing and droplet positioning accuracy based on motion position and time compensation. These solutions have application value in general inkjet printing scenarios, but their focus is mainly on area matching printing, print timing control, or imaging accuracy control on ordinary substrates.

[0003] However, Xuan paper differs significantly from ordinary printing materials in fiber structure, ink absorption characteristics, and diffusion patterns. Xuan paper, after being inked, tends to diffuse unevenly along the fiber direction, and the degree of diffusion in different areas is influenced by the local brushstroke shape, boundary transition, and the amount of ink in neighboring areas. In calligraphy and traditional Chinese painting reproduction scenarios, flying white strokes, dry brushstrokes, fine lines, the fading of ink at the boundary, and the gaps left in the paper are all key details that significantly impact the visual effect. If the uniform coverage generation method and uniform grid layout method used for ordinary paper materials are still applied, problems such as fine lines becoming thicker, edges becoming fuzzy, flying white pores being filled, and ink color transitions becoming distorted can easily occur. For inkjet printing of calligraphy and painting reproductions on Xuan paper, the influence of the skeleton lines, boundary distance, pore areas, local direction, and neighboring coverage on the image must be comprehensively considered, but existing conventional inkjet data processing workflows lack specific handling for these factors.

[0004] Therefore, there is an urgent need for an inkjet printing data processing method suitable for the reproduction of calligraphy and paintings on Xuan paper, which can distinguish the structural regions in the input calligraphy and painting images and perform targeted processing on the coverage distribution and inkjet dot arrangement in combination with the properties of Xuan paper. Summary of the Invention

[0005] This application provides a data processing method for inkjet printing of calligraphy and painting reproductions on Xuan paper, in order to at least solve some of the technical problems existing in the related technologies described above.

[0006] According to a first aspect of the embodiments of this application, a data processing method for inkjet printing of calligraphy and painting reproductions on Xuan paper is provided, including:

[0007] The paper attribute parameters are read according to the Xuan paper type information. The paper attribute parameters include at least the main fiber orientation angle, boundary influence distance threshold, neighborhood statistical window size, skeleton continuity constraint coefficient, boundary expansion penalty coefficient, pore protection buffer width, dark channel reduction upper limit, and permeation risk weight reorganization.

[0008] The digital image is preprocessed to generate an analysis block set. Structural region recognition is performed on the analysis block set to obtain a skeleton line set, a boundary distance map, a boundary sensitive region mask, a pore mask, and a slow-release layup region mask.

[0009] A basic coverage map is generated based on the digital images of calligraphy and painting and the channel configuration of the printing device, and a neighborhood expected coverage map and a local orientation map are generated based on the basic coverage map;

[0010] A bleeding risk map is generated based on the skeleton line set, boundary distance map, pore mask, slow-release spread area mask, neighborhood expected coverage map, local orientation map, and the paper property parameters.

[0011] Based on the erosion risk map, the basic coverage map is reassigned to obtain a corrected coverage map;

[0012] Based on the corrected coverage map, the skeleton line set, the aperture mask, and the boundary distance map, a partitioned raster rearrangement is performed to generate a target nozzle bitmap; target raster data is generated based on the target nozzle bitmap and output to the inkjet printing device.

[0013] As an optional approach, performing structural region identification on the analysis block set specifically includes:

[0014] Each analysis block is input into the structural segmentation model, which outputs a class probability map of the main handwriting skeleton candidate region, boundary-sensitive candidate region, pore-preserving candidate region, and slow-release spread candidate region. An initial class map is generated based on the class probability map. The main handwriting skeleton candidate region is refined to obtain the skeleton line set. A distance transformation is performed based on the initial class map to obtain the boundary distance map. A boundary-sensitive region mask is determined based on the boundary-sensitive candidate region. A connected component filtering is performed on the pore-preserving candidate region to obtain the pore mask. The remaining slow-release regions are determined as the slow-release spread region mask.

[0015] As an optional approach, the step of generating a basic coverage map based on the digital image of the calligraphy and painting and the channel configuration of the printing device, and generating a neighborhood predicted coverage map and a local orientation map based on the basic coverage map, specifically includes:

[0016] Based on the pixel color values ​​of the digital image of the calligraphy and painting and the channel configuration of the printing device, the basic coverage rate of each pixel on each printing channel is determined to form the basic coverage rate map; within a statistical window centered on each pixel and with a side length determined by the size of the neighborhood statistical window, the basic coverage rate of each pixel on each channel is statistically analyzed to obtain the neighborhood expected coverage rate map; based on the grayscale gradient of the digital image of the calligraphy and painting, the local principal direction of each pixel is calculated to form the local direction map.

[0017] As an optional approach, the generation of the permeation risk map specifically includes:

[0018] For each pixel, a region item, a boundary item, a direction item, and a coverage item are determined. The region item is determined based on the neighboring skeleton line set, pore mask, slow-release spread area mask, and boundary-sensitive region mask to which each pixel belongs. The boundary item is determined based on the boundary distance of each pixel in the boundary distance map and the boundary influence distance threshold. The direction item is determined based on the angle between the local direction of each pixel and the main fiber direction angle. The coverage item is determined based on the neighborhood predicted coverage map. The region item, boundary item, direction item, and coverage item are weighted according to the percolation risk weighting to obtain the percolation risk value for each pixel.

[0019] As an optional approach, when generating the diffusion risk map, the method further includes: adjusting the diffusion risk value of pixels located inside the pore mask to be no lower than a preset pore protection threshold; and incrementally correcting the diffusion risk value of pixels located outside the pore mask and at a distance from the pore mask no greater than the pore protection buffer width.

[0020] As an optional approach, the process of reallocating channels in the basic coverage map based on the erosion risk map to obtain a corrected coverage map specifically includes:

[0021] Based on the comprehensive chromaticity information of each pixel, each pixel is divided into neutral color pixels and non-neutral color pixels; based on the diffusion risk value of each pixel and the upper limit of the dark channel reduction, the dark channel reduction ratio of each pixel is determined; for the neutral color pixels, the dark channel coverage is reduced according to the dark channel reduction ratio; for the non-neutral color pixels, the dark channel coverage is reduced according to the dark channel reduction ratio, and the reduced portion is distributed to the remaining channels according to the original proportion of the remaining channels; for pixels located inside the aperture mask, the coverage of each channel is set to zero to form the corrected coverage map.

[0022] As an optional approach, the step of performing partitioned raster rearrangement based on the corrected coverage map, the skeleton line set, the aperture mask, and the boundary distance map specifically includes: dividing the corrected coverage map into multiple micro-blocks; for each printing channel in each micro-block, determining the number of target landing points based on the corrected coverage of each pixel in the corresponding channel; establishing a candidate position set in each micro-block, and setting no-entry markers for candidate positions located within the aperture mask and candidate positions that form cross-aperture connections with the aperture mask; determining candidate position scores based on the corrected coverage of the candidate positions, their adjacency with the skeleton line set, and their adjacency with the high-risk direction of the boundary; selecting target landing point positions according to the candidate position scores, and updating the adjacency state after each point selection until the target number of landing points is reached.

[0023] As an optional approach, determining the candidate location score based on the corrected coverage of the candidate location, its adjacency with the skeleton line set, and its adjacency with the high-risk direction of the boundary specifically includes:

[0024] The basic coverage contribution is determined based on the corrected coverage rate of the candidate location on the corresponding channel; the skeleton continuity reward is determined based on the continuity relationship between the candidate location and the landing point of the selected skeleton direction, combined with the skeleton continuity constraint coefficient; the boundary expansion penalty is determined based on the selected continuity length of the candidate location in the high-risk direction of the boundary, combined with the boundary expansion penalty coefficient; and the candidate location score is determined based on the basic coverage contribution, skeleton continuity reward, and boundary expansion penalty.

[0025] As an optional solution, the step of generating target raster data based on the target dot bitmap and outputting it to the inkjet printing device specifically includes:

[0026] The segmented target inkjet point bitmaps corresponding to each micro-block are stitched together according to their position index in the whole map to obtain the full-frame target inkjet point bitmap; when there are conflicting landing points in the overlapping area, the target landing points corresponding to the boundary sensitive area or the skeleton line set neighborhood are retained; the target raster data is generated according to the stitched full-frame target inkjet point bitmap and the channel configuration of the printing device, and the target raster data is sent to the inkjet printing device.

[0027] According to a second aspect of the embodiments of this application, a data processing system for inkjet printing of calligraphy and painting reproduction on Xuan paper is also provided, comprising:

[0028] The parameter reading module is configured to read the corresponding paper attribute parameters according to the Xuan paper type information. The paper attribute parameters include at least the main fiber direction angle, boundary influence distance threshold, neighborhood statistical window size, skeleton continuity constraint coefficient, boundary expansion penalty coefficient, pore protection buffer width, dark channel reduction upper limit, and diffusion risk weight reorganization.

[0029] The preprocessing and recognition module is configured to preprocess the digital image of calligraphy and painting and generate a set of analysis blocks, perform structural region recognition on the set of analysis blocks, and obtain a skeleton line set, a boundary distance map, a boundary sensitive region mask, a pore mask, and a slow-release paving area mask.

[0030] The coverage generation module is configured to generate a basic coverage map based on the digital image of the calligraphy and painting and the channel configuration of the printing device, and to generate a neighborhood expected coverage map and a local orientation map based on the basic coverage map;

[0031] The risk map generation module is configured to generate a bleeding risk map based on the skeleton line set, boundary distance map, pore mask, slow-release spread area mask, neighborhood expected coverage map, local orientation map, and the paper property parameters.

[0032] The channel reallocation module is configured to reallocate channels on the base coverage map based on the permeation risk map to obtain a modified coverage map.

[0033] The grid rearrangement and data output module is configured to perform partitioned grid rearrangement based on the corrected coverage map, the skeleton line set, the aperture mask, and the boundary distance map, generate a target nozzle bitmap, generate target grid data based on the target nozzle bitmap, and output it to the inkjet printing device.

[0034] This application identifies structural regions in digital images of calligraphy and paintings to be printed, extracts skeleton line sets, boundary distance maps, pore masks, and slow-release spreading area masks, and generates a diffusion risk map by combining the main fiber orientation angle corresponding to the type of Xuan paper, the boundary influence distance threshold, the neighborhood statistical window size, and the diffusion risk weight reorganization. On this basis, channel reallocation is performed on the basic coverage map, and further, the partitioned raster is rearranged according to the skeleton continuity constraint, boundary expansion penalty, and pore entry prohibition rule to generate a target spray point map and target raster data adapted to the Xuan paper material.

[0035] Therefore, without changing the existing mechanical structure and jet control link of inkjet printing equipment, more targeted printing data processing can be carried out on fine lines, dry brush, ink blot boundaries and areas where ink accumulation turns light, based on the directional ink absorption and local diffusion differences of Xuan paper fibers. This can reduce phenomena such as widening of fine lines, edge fuzzing, dry brush blockage and local merging, while maintaining the stability of the main structure and overall color layer of the calligraphy and painting.

[0036] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Furthermore, no embodiment in this disclosure is required to achieve all the effects described above. Attached Figure Description

[0037] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0038] Figure 1 This is a schematic diagram of a data processing method for inkjet printing of calligraphy and paintings on Xuan paper, provided in an embodiment of this disclosure.

[0039] Figure 2 This is a schematic diagram of the structural region identification process provided in an embodiment of the present disclosure.

[0040] Figure 3 This is a schematic diagram of the target spray point bitmap generation process provided in the embodiments of this disclosure.

[0041] Figure 4 This is a schematic diagram of the structure of a data processing system for inkjet printing of calligraphy and painting reproduction on Xuan paper, provided in an embodiment of this disclosure. Detailed Implementation

[0042] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0043] The implementation process of the method described in this application will be described in detail below with reference to specific embodiments. It should be noted that this embodiment is only used to explain this application and is not intended to limit the scope of protection of this application. Without departing from the concept of this application, conventional adjustments or substitutions of each step by those skilled in the art should be included in the scope of protection of this application.

[0044] This implementation method is applicable to inkjet printing scenarios for replicating calligraphy, traditional Chinese painting, ink wash painting, flower-and-bird painting, and paintings containing seal areas on Xuan paper. The executing entity can be a front-end data processing device, which can be integrated into the front-end controller of the inkjet printer or deployed in an image processing server communicatively connected to the inkjet printer. This device must at least possess image reading, parameter retrieval, model inference, coverage calculation, raster reordering, and print data output capabilities. Input data includes digital images of the paintings to be printed, Xuan paper type information, and printer channel configuration. Before operation, the device pre-stores Xuan paper parameter files, color mapping tables, and a trained structural segmentation model.

[0045] Figure 1 This is a flowchart of a data processing method for inkjet printing of calligraphy and painting reproduction on Xuan paper according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes steps S1-S6:

[0046] In step S1, the corresponding paper attribute parameters are read according to the Xuan paper type information. The paper attribute parameters include at least the main fiber orientation angle, boundary influence distance threshold, neighborhood statistical window size, skeleton continuity constraint coefficient, boundary expansion penalty coefficient, pore protection buffer width, dark channel reduction upper limit, and diffusion risk weight reorganization.

[0047] In some embodiments, the digital image of the calligraphy or painting to be printed can be a high-resolution image obtained through scanning, or a photographically corrected replica image. The Xuan paper type information is stored in the form of paper category identifiers, at least distinguishing between raw Xuan paper, semi-raw and sized Xuan paper, and can be further subdivided into different supply batches. The printing device channel configuration at least indicates the number of available ink channels, channel identifiers, and channel order of the current printing device.

[0048] Before processing begins, the print front-end data processing device first establishes the data context for this task; the data context includes the processing image identifier, paper type identifier, channel sequence, processing resolution, temporary cache index, and task output identifier.

[0049] For Xuan paper parameters, a parameter recording method organized by paper type is adopted. Each parameter record includes at least the main fiber orientation angle, boundary influence distance threshold, neighborhood statistical window size, skeleton continuity constraint coefficient, boundary expansion penalty coefficient, pore protection buffer width, dark channel reduction upper limit, permeation risk weight reorganization, boundary redistribution threshold, pore protection threshold, and continuous length upper limit.

[0050] Among them, the main fiber orientation angle represents the main fiber orientation of this type of Xuan paper in the plane, which can be determined by observing the ink spread direction under standard printing conditions on an offline sample; the boundary influence distance threshold represents the range near the boundary that still needs to be treated as a high-risk pixel; the neighborhood statistical window size represents the side length of the statistical window for the local expected coverage; the skeleton continuity constraint coefficient represents the reward intensity for continuous arrangement of points in the skeleton direction; and the boundary outward expansion penalty coefficient represents the suppression intensity for continuous arrangement of points in the high-risk direction of the boundary.

[0051] The pore protection buffer width indicates the distance at which drop points are still prohibited or restricted outside the pore; the dark channel reduction upper limit indicates the maximum proportion of dark channels that can be reduced at high-risk pixels; the diffusion risk weighting reconfiguration indicates the weight of region, boundary, direction, and coverage items in the comprehensive risk calculation; the boundary redistribution threshold indicates the risk threshold for initiating additional coverage decay outside the boundary; the pore protection threshold indicates the minimum risk value inside the pore; and the continuous length upper limit indicates the maximum length of continuous drop points allowed in high-risk directions.

[0052] The above parameters are obtained through statistical analysis of standard samples during the offline maintenance phase. For example, multiple rounds of proofing can be performed on the same type of Xuan paper using a fixed pattern library to statistically analyze the boundary widening, fly white blocking ratio, fine line breakage ratio, and overall color deviation. Then, the corresponding parameter groups are selected through a comprehensive evaluation function. During offline maintenance, if the same paper type shows stable differences in subsequent batches, a new parameter record can be created for that batch. Once a single printing task begins, no further online changes are made to this parameter record.

[0053] The color mapping table uses a channel coverage lookup table generated after equipment calibration. The lookup table takes the color coordinates of the comprehensive color space as input and the theoretical coverage ratio of each printing channel as output. This table is established based on the standard color block printing results during the equipment maintenance phase and is called in online tasks.

[0054] In step S2, the digital image of the calligraphy and painting is preprocessed to generate an analysis block set. Structural region recognition is performed on the analysis block set to obtain a skeleton line set, a boundary distance map, a boundary sensitive region mask, a pore mask, and a slow-release paving area mask.

[0055] Please see Figure 2 , Figure 2 A schematic diagram of the structural region identification process provided in an embodiment of this disclosure is shown. Figure 2 As shown, in step S201, the digital images of calligraphy and painting are preprocessed and an analysis block set is generated.

[0056] After the digital image of the calligraphy and painting enters the processing flow, basic preprocessing resolution normalization is first performed. Specifically, the device resamples the original image into a processed image based on the output resolution given by the printing task. If the original image resolution is higher than the output resolution, an edge-preserving downsampling method is used; if the original image resolution is lower than the output resolution, a smooth constraint interpolation method is used. Based on the pixel coordinates of this processed image, subsequent skeleton line sets, boundary distance maps, pore masks, slow-release spread area masks, basic coverage maps, neighborhood expected coverage maps, local orientation maps, diffusion risk maps, corrected coverage maps, and target spray point maps are obtained.

[0057] Preprocessing also includes generating a set of analysis blocks; since structural segmentation models typically use fixed-size inputs, the image to be processed is divided into multiple analysis blocks to balance computational efficiency and local continuity. Overlapping regions are preserved between adjacent analysis blocks, and the width of the overlap region can be configured to a fixed value to ensure that the block edge structure does not break during subsequent stitching.

[0058] Each analysis block records its position index in the entire image. The position index includes at least the coordinates of the top left corner, the block width, and the block height. In addition to the original color channels, auxiliary feature channels are constructed for each analysis block. Specifically, the analysis block is first converted into a grayscale image, and then the horizontal and vertical gradients are calculated on the grayscale image to obtain the gradient magnitude map. The input tensor constructed in this way contains both color information and edge intensity information, which is beneficial for subsequent identification of the handwriting subject, boundary transitions, and pore areas.

[0059] In step S202, the analysis blocks are converted into model input vectors according to the segmentation model data format. The data structure of each analysis block consists of a multi-channel tensor, which includes at least a red channel, a green channel, a blue channel, and a gradient magnitude channel. In one example, if the device already has a comprehensive color space conversion module, the color channels can be replaced with the luminance and chromaticity channels in the comprehensive color space. After the analysis block tensor is generated, it is entered into the inference queue of the structural segmentation model according to its position index.

[0060] In step S203, each analysis block is input into the structural segmentation model, and initial class probability maps of the main handwriting skeleton candidate region, boundary-sensitive candidate region, porosity-preserving candidate region, and slow-release spread candidate region are output. In this embodiment, the structural segmentation model adopts a U-Net, which consists of an encoding path, a bottleneck layer, and a decoding path in sequence.

[0061] The encoding path contains multiple levels of downsampling units. Each downsampling unit consists of two convolutional layers, a batch normalization layer, a linear rectified activation layer, and a pooling layer. The first convolutional layer receives the output feature map from the previous layer and extracts local texture. The second convolutional layer further enhances the structural representation at the same scale. Then, the batch normalization layer stabilizes the feature distribution, the linear rectified activation layer introduces a nonlinear response, and the pooling layer reduces the spatial resolution and expands the receptive field. Adjacent downsampling units are cascaded and connected, and the output of the lower layer is transmitted to the higher layer.

[0062] The bottleneck layer is located between the deepest layer of the encoding path and the decoding path. Its input is the output feature map of the last downsampling unit. Internally, it employs two convolutional layers, a batch normalization layer, and a linear rectified activation layer to aggregate global structural semantics. The bottleneck layer output is simultaneously provided to the first upsampling unit of the decoding path.

[0063] The decoding path contains multiple upsampling units corresponding to the scale of the encoding path. Each upsampling unit first upsamples the feature map of the previous layer, and then concatenates it with the feature map of the corresponding scale of the encoding path. After concatenation, the features are fused with local edges and high-level semantics through convolutional layers, batch normalization layers, and linear rectified activation layers. Through this cross-layer connection, boundary position details and region semantic judgments can be preserved in the same branch. The end of the decoding path connects to the output layer, which uses a single convolutional mapping to create a four-channel probability map. The four channels correspond to the main handwriting skeleton candidate region, the boundary-sensitive candidate region, the aperture preservation candidate region, and the gradual spread candidate region, respectively. A normalization classification layer can be connected after the output layer to make the sum of the probabilities of each pixel in the four classes equal to 1.

[0064] In one embodiment, model training is conducted offline. The training samples consist of composite images of calligraphy, landscapes, flowers and birds, figures, and seals, with each image labeled by annotators to identify four types of regions. The candidate regions for the main stroke skeleton include the main strokes of calligraphy, the brushstroke lines, the branch lines, and the central area of ​​the main ink mark; the candidate regions for boundary sensitivity include the edges of ink blots, the boundaries of ink accumulation and fading, the outer edges of flying white strokes, the edges of seals, and the transition areas on both sides of fine lines; the candidate regions for porosity preservation include flying white stroke holes, gaps in dry brush strokes, white lines on paper, and negative white spaces that are meaningful for preservation; and the candidate regions for slow-release gradation include large areas of gently stained areas, background transition color blocks, and non-critical low-frequency areas.

[0065] During model training, the input is an analysis block tensor, and the output is a four-class label image with the same size as the input. The loss function can be composed of cross-entropy loss and Dice loss. Cross-entropy loss constrains pixel class discrimination, while Dice loss constrains sparse region segmentation contours. The optimization phase uses an iterative update method until the segmentation accuracy on the validation set stabilizes. After training, the model weights are solidified and deployed to the printing front-end data processing device.

[0066] In the online task, each analysis block is sequentially input into the structural segmentation model, which outputs a four-channel probability map. The device selects the category with the highest probability for each pixel to generate an initial category map. In some embodiments, if there are overlapping regions between analysis blocks, the overlapping parts of the four-category probability maps are weighted by center distance and merged before generating the initial category map. This can reduce the probability of category jumps at block boundaries.

[0067] In step S204, each initial category candidate region of the initial category probability map is processed to obtain the final region result. For the region marked as the main handwriting skeleton candidate region, a thinning process is performed. The thinning process can use a morphological thinning algorithm, which maintains the connectivity by gradually peeling off edge pixels, and finally obtains a pixel-level center line. This center line constitutes a skeleton line set throughout the entire image.

[0068] A distance transformation is performed on the regions related to the ink blot in the initial category map to obtain a boundary distance map. The input of the distance transformation is the boundary of the ink blot, and the output is the distance value from each pixel to the nearest boundary. This distance value is used for subsequent boundary item calculations. Therefore, the boundary distance map is kept at the same size as the processed image after generation and is not resampled.

[0069] The boundary-sensitive candidate regions are filtered to obtain a boundary-sensitive region mask. The boundary-sensitive region mask is used to characterize the pixel distribution of the boundary-sensitive regions in the processed image. The region corresponding to the marked pixels in the boundary-sensitive region mask is the boundary-sensitive region.

[0070] Specifically, an initial boundary-sensitive region can be formed based on the pixels identified as boundary-sensitive candidate regions in the initial category image. By combining connectivity, region area, or adjacency relationship with the main boundary of the ink blot, obviously isolated false detection regions can be removed, thereby obtaining a boundary-sensitive region mask. The boundary-sensitive region mask is consistent with the processed image in terms of spatial size and is used for subsequent region item calculation and boundary outer coverage attenuation processing.

[0071] Connected component filtering is performed on the pore retention candidate regions. During the filtering, the area of ​​all connected components is calculated first, and then isolated regions smaller than the preset area lower limit are deleted. The area lower limit can be determined statistically based on the minimum effective size of the retained pores in the training samples. After filtering, the pore mask is obtained. Regions that are not covered by the above three types and are identified by the model as slow-release spreading candidate regions are directly formed into slow-release spreading region masks.

[0072] Therefore, the structural region identification stage outputs a skeleton line set, a boundary distance map, a boundary sensitive region mask, a pore mask, and a slow-release spread area mask.

[0073] In step S3, a basic coverage map is generated based on the digital image of the calligraphy and painting and the channel configuration of the printing device, and a neighborhood expected coverage map and a local orientation map are generated based on the basic coverage map.

[0074] After obtaining the processed image and region result set, the device generates a basic coverage map according to the printing device channel configuration. The printing device channel configuration includes the number of channels, the channel number, and the ink type corresponding to each channel. The channel sequence is fixed at the beginning of the task, and the channel order of all subsequent coverage vectors and target ink dot bitmaps is arranged according to this sequence.

[0075] For each pixel in the processed image, its color value is first read. If the color mapping table is based on the comprehensive color space, the pixel color is first converted from the input color space to the comprehensive color space, and then the theoretical coverage of each channel is obtained by looking up the table. The basic coverage map is a three-dimensional array, with the first two dimensions being spatial position and the third dimension being the channel dimension. The coverage vector at any pixel position represents the theoretical ink coverage ratio of that position in each channel. This ratio is the initial basis for subsequent channel redistribution.

[0076] When generating the neighborhood predicted coverage map, a square statistical window with a side length determined by the size of the neighborhood statistical window is extracted centered on each pixel. The basic coverage of all channels for all pixels within the window is accumulated, and then averaged based on the total number of pixels and the total number of channels within the window. The calculation process can be expressed as follows:

[0077]

[0078] in, Indicates position The expected coverage of the neighborhood, This indicates a statistical window centered on this location. Indicates the number of pixels within the window. Indicates the total number of channels. Indicates position In the The higher the base coverage rate on each channel, the higher the expected cumulative liquid load around that location, and the more cautious you need to be when continuing to check for liquids.

[0079] Local orientation maps are generated based on grayscale gradients. Specifically, the processed image is converted to a grayscale image, and the horizontal and vertical gradients are calculated in the neighborhood of each pixel, and a structure tensor is constructed. The local principal direction is determined based on the direction of the principal feature vector of the structure tensor. If the pixel is located near the skeleton line set, the direction can be regarded as the stroke direction; if it is located in a boundary-sensitive area, it can be regarded as the local boundary tangential direction; if it is located in a slow-release layer, the direction is only used as a weak reference, and its influence can be reduced in subsequent risk calculations. In order to ensure that the local orientation map is consistent with the coordinates of other layers, the local orientation map is not scaled after generation.

[0080] In step S4, a bleeding risk map is generated based on the skeleton line set, boundary distance map, pore mask, slow-release spread area mask, neighborhood expected coverage map, local orientation map, and the paper property parameters.

[0081] For each pixel in the processed image, calculate the region term, boundary term, orientation term, and coverage term. The region term is determined based on the region to which the pixel belongs. If the pixel is located in the region corresponding to the boundary sensitive region mask, it is assigned a higher region term; if it is located inside the aperture mask, it is assigned a high region term; if it is located in the neighborhood of the skeleton line set, it is assigned a medium region term; if it is located in the slow-release spread region, it is assigned a lower region term. The value of the region term can be preset to different levels according to the offline calibration results.

[0082] The boundary term is calculated based on the boundary distance in the boundary distance map and the boundary influence distance threshold. When the distance from a pixel to the nearest boundary is less than the boundary influence distance threshold, the boundary term decreases as the distance increases; when the distance is not less than the threshold, the boundary term is recorded as zero. The calculation relationship can be expressed as follows:

[0083]

[0084] in, For position Boundary terms, This represents the distance value at this location on the boundary distance map. The distance threshold is affected by the boundary.

[0085] The direction term is determined based on the angle between the local direction and the main fiber direction. Since Xuan paper is more prone to directional expansion along the main fiber direction, when the local brushstroke direction or boundary tangent is close to the main fiber direction, the risk of bleed in that pixel should be increased. The direction term can be calculated using the following formula:

[0086]

[0087] in, For direction terms, For local orientation patterns at location Direction value, The direction term is the main fiber direction angle. The closer the local direction is to the main fiber direction, the larger the direction term is. When the two are close to orthogonal, the direction term decreases. If there is no effective local direction value at a certain position, the direction term can be recorded as zero.

[0088] The coverage value is taken directly from the corresponding position in the neighborhood's expected coverage map. This value reflects the expected coverage intensity around the current pixel. The higher the cumulative coverage of the neighborhood, the greater the liquid load on the local area of ​​the paper. If densely packed dots in the same direction are added later, it will be easier for the liquid to spread outward.

[0089] After obtaining the above components, the device performs weighted fusion according to the diffusion risk weighting to obtain the diffusion risk value of each pixel. The calculation relationship can be expressed as follows:

[0090]

[0091] in, Indicates position The risk value of leaching, For regional items, For boundary terms, For direction terms, For coverage items, , , , These are the corresponding weighting coefficients.

[0092] The weighting coefficients are given in the Xuan paper parameter archive and are derived from the offline calibration results. During calibration, the boundary morphology, white streaking retention, and fine line stability of multiple sets of parameters on actual samples are compared. The parameter sets with lower overall costs of color deviation and structural distortion are selected and written into the archive. To ensure calculation stability, the weighting coefficient set can be set to a fixed value under a paper category, or it can be set to a normalized weight that satisfies the sum of 1.

[0093] In some embodiments, the pixels in the aperture mask are more sensitive to subsequent imaging. To prevent the aperture region from being covered by neighboring spray points during channel redistribution and grid rearrangement, the device performs a risk value lower limit correction on the pixels inside the aperture mask.

[0094] Specifically, if a pixel is located inside the pore mask, its bleed risk value is adjusted to be no less than the pore protection threshold determined according to the paper property parameters. If a pixel is located outside the pore mask and the distance from the pore mask is no greater than the pore protection buffer width, its bleed risk value is incrementally corrected. This increment can decrease with the distance from the pore boundary to form a buffer zone. The width of the buffer zone is derived from the Xuan paper parameter file.

[0095] After the diffusion risk map is generated, it maintains the same spatial size as the processed image and is accompanied by a task-level index identifier. Subsequently, coverage correction, boundary attenuation, and partition raster rearrangement all call this map, and the risk value is no longer recalculated. This process ensures the consistency of risk assessment on the one hand, and avoids local inconsistencies caused by repeated calculations on the other hand.

[0096] In step S5, the basic coverage map is reassigned according to the permeation risk map to obtain a modified coverage map.

[0097] Specifically, for each pixel, the comprehensive color information is calculated. If the comprehensive color space is described by luminance and two chromaticity components, the pixel is determined to be a neutral color pixel or a non-neutral color pixel based on the sum of the absolute values ​​of the chromaticity components. Neutral color pixels usually correspond to black, gray, light ink and some low-saturation areas. Non-neutral color pixels correspond to colored areas with obvious comprehensive color tendencies or stamp comprehensive color areas. The comprehensive color chromaticity classification threshold is determined during the device calibration stage and remains unchanged for the same device.

[0098] For neutral color pixels, the device determines the dark channel reduction ratio of the pixel based on the bleed risk value and the dark channel reduction limit. In one example, the bleed risk value is normalized and mapped to the range of 0 to 1. The reduction ratio can be linearly mapped to the range of zero to the reduction limit according to the risk value. If the risk value at a certain position is low, the reduction ratio is small. If the risk value is close to a high value, the reduction ratio is close to the reduction limit. After reduction, the coverage of the dark channel decreases proportionally. If the pixel is also located in the neighborhood of the skeleton line set, the minimum coverage requirement of the corresponding channel of the skeleton is maintained under the constraint of the dark channel reduction limit to avoid the main handwriting skeleton from breaking due to excessive reduction.

[0099] For non-neutral color pixels, first determine the dark channel reduction ratio in the same way, and then distribute the reduced coverage to the other channels according to the original coverage ratio of the other channels. This can keep the overall color direction basically unchanged, while reducing the situation where high-risk areas are concentrated on the coverage of the dark channel. During the redistribution process, the original channel coverage vector of the pixel is used as the allocation benchmark to avoid introducing new overall color shifts.

[0100] For pixels located inside the aperture mask, all channel coverage is set to zero. For pixels located within the aperture buffer zone, an upper limit for coverage can be set based on their distance from the aperture boundary. The closer the upper limit is to the aperture boundary, the lower it is, thus avoiding the formation of a cross-aperture surrounding the aperture. After the zeroing of the aperture interior and the buffer zone limiting processing are completed, aperture-related coverage constraints are formed.

[0101] For the mask corresponding to the boundary sensitive area, the device further determines whether the pixel is located inside or outside the boundary. The inside or outside of the boundary can be determined by combining the skeleton position and the boundary normal relationship. If a pixel is located outside the boundary and its diffusion risk value is higher than the boundary redistribution threshold determined according to the paper attribute parameters, then an attenuation coefficient is applied to the coverage of each channel of the pixel and written into the corrected coverage map. The attenuation coefficient outside the boundary can be determined by combining the risk value and the boundary distance. The closer to the boundary and the higher the risk value, the more obvious the attenuation. This kind of attenuation only acts on the outside of the boundary, thereby forming an asymmetric structure inside and outside the boundary in the coverage distribution.

[0102] After completing the above operations, the base coverage map is converted into a modified coverage map. The modified coverage map is still a three-dimensional array, and its spatial location and channel index are completely consistent with the base coverage map.

[0103] In step S6, partitioned grid rearrangement is performed based on the corrected coverage map, the skeleton line set, the aperture mask, and the boundary distance map to generate a target nozzle bitmap; target grid data is generated based on the target nozzle bitmap and output to the inkjet printing device.

[0104] Please see Figure 3 , Figure 3 A schematic diagram illustrating the target spray point bitmap generation process provided in an embodiment of this disclosure is shown. For example... Figure 3 As shown, in step S301, the micro-blocks are divided and the number of target landing points is calculated.

[0105] The device divides the corrected coverage map into multiple micro-blocks. The size of the micro-blocks is set according to the printing resolution and the minimum control particle size of the printhead. The spatial index of the micro-blocks is consistent with the processed image. For each micro-block and each printing channel, the corrected coverage of all pixels in the micro-block on the channel is summed and then converted into the number of target landing points for that micro-block and that channel. If the conversion result is a decimal, rounding rules can be used.

[0106] In step S302, a candidate location set is established based on the divided micro-blocks and their pixel positions. Within each micro-block, all pixel positions can be used as candidate locations. The device first performs an entry restriction determination on the candidate locations; if a candidate location falls within the aperture mask, it is directly marked as an entry restriction location and does not participate in subsequent scoring.

[0107] If a candidate position is not within the aperture mask, but its selection may form a cross-aperture connection with the other side of the aperture, it is also marked as a forbidden position. A cross-aperture connection means that the position and the position on the opposite side of the aperture form a continuous path in the local neighborhood, so that the aperture is surrounded by spray points on both sides. Optionally, the device determines whether a cross-aperture connection is formed by checking whether the shortest path from the candidate position to the boundary point on the opposite side of the aperture passes through the central region of the aperture.

[0108] After the prohibition determination is completed, skeleton adjacency markers and boundary high-risk direction adjacency markers are established for non-prohibited locations. Skeleton adjacency reflects whether the location is adjacent to the skeleton line set and whether, if selected, it helps to form a continuous spray chain along the skeleton direction. Boundary high-risk direction adjacency reflects whether, if the location is selected, it will extend the continuous spray length in the high-permeability direction. The high-permeability direction can be determined by the local direction map and the main fiber direction angle. When the local direction is close to the main fiber direction, the direction is considered a high-risk direction.

[0109] In step S303, for each non-forbidden candidate location, its score is calculated, and target landing points are selected in descending order of score. The score consists of at least a basic coverage contribution, a skeleton continuous reward, and a boundary expansion penalty; the basic coverage contribution comes directly from the corrected coverage rate of the location on the corresponding channel, and the higher the coverage rate, the higher the basic contribution.

[0110] The skeleton continuity reward is determined based on the continuity relationship between the candidate position and the selected skeleton direction landing point. If the position is located in the neighborhood of the skeleton line set and is adjacent to the selected landing point along the skeleton direction, a reward is given, and the reward strength is controlled by the skeleton continuity constraint coefficient. The boundary expansion penalty is determined based on the selected continuity length of the position in the high-risk direction of the boundary. If selecting it would extend the continuity chain in the high-risk direction, a penalty is applied, and the penalty strength is controlled by the boundary expansion penalty coefficient. Their calculation relationship can be expressed as follows:

[0111]

[0112] in, Score the candidate positions. For position In the Corrected coverage on each channel The coefficient represents the skeleton continuity constraint. For the skeleton continuous reward item, The boundary expansion penalty coefficient, For the boundary expansion penalty term, when a position is not in the neighborhood of the skeleton line set, its skeleton continuous reward term is zero; when a position is not in the boundary sensitive area or is not on the high-risk direction chain, its boundary expansion penalty term is zero.

[0113] After obtaining the candidate position scores, the device selects the target landing point in descending order of scores. For each selected position, the selected position set and adjacency status are updated in the micro-block status table. The adjacency status includes at least the continuous length in each direction, the continuous segment position in the skeleton direction, and the continuous segment position in the high-risk direction of the boundary. If the continuous length in a certain direction reaches the upper limit of the continuous length, the score of the candidate position that continues to extend in that direction is reduced, or it is directly marked as unselectable. The upper limit of the continuous length comes from the Xuan paper parameter file and remains unchanged in the same task.

[0114] For micro-blocks traversed by the skeleton line set, the lowest point density along the skeleton direction is prioritized during iterative point selection. Specifically, several positions with the highest scores can be selected first in the skeleton neighborhood to ensure the continuity of the skeleton trunk. Then, points are selected for the remaining positions according to the regular scoring. This ensures that the central orientation of the main strokes of calligraphy, mountain and rock texture lines, and branch lines remains stable. For micro-blocks located in boundary-sensitive areas, if a continuous high-risk point chain has already formed in a certain direction, the priority of candidate positions for extending the chain is reduced to control the outward expansion of the edge. For micro-blocks located in the slow-release paving area, the point selection is mainly determined by the basic coverage contribution, with weaker boundary penalties to maintain a smooth transition at low frequencies.

[0115] Optionally, in certain microblocks with dense pores or high-risk boundaries, the prohibition rules and penalty rules may cause the number of legal candidate locations to be less than the number of target landing points. In this case, the device performs microblock insufficiency compensation. The compensation order is as follows: within the same microblock, the boundary penalty item of the low-risk area is appropriately relaxed, and the scores of some candidate locations are recalculated. If it is still insufficient, the number of missing target landing points is transferred proportionally to adjacent low-risk microblocks. If the adjacent microblocks have also reached their tolerable upper limit, the local coverage gap of this microblock is retained and recorded in the comprehensive color error record table. This compensation only changes the local target landing point allocation and does not write back to modify the coverage map and permeation risk map.

[0116] In step S304, the target spray point bitmaps of each block are stitched together to form a full-frame target spray point bitmap. After completing the sorting of all micro-blocks and all channels, a block-based target spray point bitmap is obtained. The block target spray point bitmap corresponding to each micro-block is backfilled into the full-frame coordinates according to its position index in the full-frame map. For conflict points that occur at the stitching boundary, if the conflict location is located in the boundary sensitive area or the neighborhood of the skeleton line set, the result from the block center that is closer to that location is retained, or the result with the higher score is retained. If it is located in the slow-release spread area, an averaging strategy can be used to retain one of them. After stitching, a full-frame target spray point bitmap is formed. This bitmap is still stored according to the channel dimension and corresponds one-to-one with the channel configuration of the printing device.

[0117] After the target dot bitmap is generated, the device encapsulates it into target raster data according to the channel configuration of the printing equipment and writes it into the print job package; the print job package includes at least the channel bitmap, channel order, image size and job identifier; after receiving it, the inkjet printing equipment can execute printing according to the existing mechanical motion and jet control process.

[0118] As an example, if the object to be printed is a cursive calligraphy work, there are many flying white strokes, abrupt turns, and dry brush cracks in the image; after region recognition, the center of the main stroke of the cursive script is included in the skeleton line set, the turning edges and the outer edge of the flying white strokes are included in the boundary sensitive area, and the paper white gaps inside the flying white strokes form a pore mask; after the basic coverage map is generated, the turning parts with more ink accumulation have higher values ​​in the neighborhood expected coverage map.

[0119] In areas where the direction of the main brushstroke is close to the direction of the main fibers of Xuan paper, the directional term is also higher. After the diffusion risk map is calculated, the risk value of these areas increases. During the channel redistribution stage, dark channels are reduced in these positions and partially allocated to other channels. The coverage of each channel in the pore area is set to zero. After entering the micro-block dot-mapping stage, the candidate positions inside the pores and across the pores are prohibited from selection. Continuous dot-mapping along the high-risk direction outside the boundary is penalized, while continuous dot-mapping in the direction of the skeleton center is rewarded. The final output bitmap is closer to the original image in overall color and is more suitable for actual imaging under the conditions of Xuan paper fibers.

[0120] Therefore, by identifying structural regions in calligraphy and painting images and establishing a diffusion risk map based on the properties of Xuan paper, and then performing channel redistribution and partitioned grid rearrangement on this basis, the target ink dot map can simultaneously reflect the continuous constraint of the brushstroke skeleton, the suppression of high-risk directions at the boundary, and the constraint of porosity retention. This can alleviate phenomena such as the widening of fine lines, edge fuzzing, white clogging, and the merging of ink patches in areas where ink has faded on Xuan paper, while maintaining the stability of the main structure and overall color gradation of the calligraphy and painting.

[0121] Please see Figure 4 , Figure 4 This is a schematic diagram of a data processing system for inkjet printing of calligraphy and paintings on Xuan paper, provided in an embodiment of this application. As shown in the figure, the system includes:

[0122] The parameter reading module 401 is configured to read the corresponding paper attribute parameters according to the Xuan paper type information. The paper attribute parameters include at least the main fiber direction angle, boundary influence distance threshold, neighborhood statistical window size, skeleton continuity constraint coefficient, boundary expansion penalty coefficient, pore protection buffer width, dark channel reduction upper limit, and diffusion risk weight reorganization.

[0123] The preprocessing and recognition module 402 is configured to preprocess the digital image of calligraphy and painting and generate a set of analysis blocks, perform structural region recognition on the set of analysis blocks, and obtain a skeleton line set, a boundary distance map, a boundary sensitive region mask, a pore mask, and a slow-release paving area mask.

[0124] The coverage generation module 403 is configured to generate a basic coverage map based on the digital image of the calligraphy and painting and the channel configuration of the printing device, and to generate a neighborhood expected coverage map and a local orientation map based on the basic coverage map.

[0125] The risk map generation module 404 is configured to generate a bleeding risk map based on the skeleton line set, boundary distance map, pore mask, slow-release spread area mask, neighborhood expected coverage map, local orientation map and the paper property parameters.

[0126] The channel reallocation module 405 is configured to reallocate channels on the basic coverage map based on the permeation risk map to obtain a modified coverage map.

[0127] The grid rearrangement and data output module 406 is configured to perform partitioned grid rearrangement based on the corrected coverage map, the skeleton line set, the aperture mask and the boundary distance map, generate a target nozzle bitmap, generate target grid data based on the target nozzle bitmap and output it to the inkjet printing device.

[0128] Each processing unit and / or module in the embodiments of this application can be implemented by an analog circuit that implements the functions described in the embodiments of this application, or by software that executes the functions described in the embodiments of this application.

[0129] In the above embodiments, the descriptions of each embodiment have different focuses. Parts not described in detail in a certain embodiment can be referred to in the relevant descriptions of other embodiments. The above descriptions are merely preferred embodiments of this application and explanations of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to the technical solutions formed by specific combinations of the above technical features, but should also cover other technical solutions formed by arbitrary combinations of the above technical features or their equivalent features without departing from the inventive concept.

Claims

1. A data processing method for inkjet printing of calligraphy and painting reproductions on Xuan paper, characterized in that, include: The paper attribute parameters are read according to the Xuan paper type information. The paper attribute parameters include at least the main fiber orientation angle, boundary influence distance threshold, neighborhood statistical window size, skeleton continuity constraint coefficient, boundary expansion penalty coefficient, pore protection buffer width, dark channel reduction upper limit, and permeation risk weight reorganization. The digital image is preprocessed to generate an analysis block set. Structural region recognition is performed on the analysis block set to obtain a skeleton line set, a boundary distance map, a boundary sensitive region mask, a pore mask, and a slow-release layup region mask. A basic coverage map is generated based on the digital images of calligraphy and painting and the channel configuration of the printing device, and a neighborhood expected coverage map and a local orientation map are generated based on the basic coverage map; A bleeding risk map is generated based on the skeleton line set, boundary distance map, pore mask, slow-release spread area mask, neighborhood expected coverage map, local orientation map, and the paper property parameters. Based on the erosion risk map, the basic coverage map is reassigned to obtain a corrected coverage map; Based on the corrected coverage map and the skeleton line set, pore mask and boundary distance map, perform partitioned grid rearrangement to generate a target spray point map; Target raster data is generated based on the target dot bitmap and output to the inkjet printing device.

2. The method according to claim 1, characterized in that, The structural region identification performed on the analysis block set specifically includes: Each analysis block is input into the structural segmentation model, which outputs a class probability map of the main handwriting skeleton candidate region, boundary-sensitive candidate region, pore-preserving candidate region, and slow-release spread candidate region. An initial class map is generated based on the class probability map. The main handwriting skeleton candidate region is refined to obtain the skeleton line set. A distance transformation is performed based on the initial class map to obtain the boundary distance map. A boundary-sensitive region mask is determined based on the boundary-sensitive candidate region. A connected component filtering is performed on the pore-preserving candidate region to obtain the pore mask. The remaining slow-release regions are determined as the slow-release spread region mask.

3. The method according to claim 1, characterized in that, The step of generating a basic coverage map based on the digital image of the calligraphy and painting and the channel configuration of the printing device, and generating a neighborhood predicted coverage map and a local orientation map based on the basic coverage map, specifically includes: Based on the pixel color values ​​of the digital image of the calligraphy and painting and the channel configuration of the printing device, the basic coverage rate of each pixel on each printing channel is determined to form the basic coverage rate map; within a statistical window centered on each pixel and with a side length determined by the size of the neighborhood statistical window, the basic coverage rate of each pixel on each channel is statistically analyzed to obtain the neighborhood expected coverage rate map; based on the grayscale gradient of the digital image of the calligraphy and painting, the local principal direction of each pixel is calculated to form the local direction map.

4. The method according to claim 3, characterized in that, The generation of the permeation risk map specifically includes: For each pixel, a region item, a boundary item, a direction item, and a coverage item are determined. The region item is determined based on the neighboring skeleton line set, pore mask, slow-release spread area mask, and boundary-sensitive region mask to which each pixel belongs. The boundary item is determined based on the boundary distance of each pixel in the boundary distance map and the boundary influence distance threshold. The direction item is determined based on the angle between the local direction of each pixel and the main fiber direction angle. The coverage item is determined based on the neighborhood predicted coverage map. The region item, boundary item, direction item, and coverage item are weighted according to the percolation risk weighting to obtain the percolation risk value for each pixel.

5. The method according to claim 4, characterized in that, When generating the diffusion risk map, the method further includes: adjusting the diffusion risk value of pixels located inside the pore mask to be no less than a preset pore protection threshold; and incrementally correcting the diffusion risk value of pixels located outside the pore mask and whose distance from the pore mask is no greater than the pore protection buffer width.

6. The method according to claim 1, characterized in that, The process of reallocating channels in the base coverage map based on the erosion risk map to obtain a corrected coverage map specifically includes: Based on the comprehensive chromaticity information of each pixel, each pixel is divided into neutral color pixels and non-neutral color pixels; based on the diffusion risk value of each pixel and the upper limit of the dark channel reduction, the dark channel reduction ratio of each pixel is determined; for the neutral color pixels, the dark channel coverage is reduced according to the dark channel reduction ratio; for the non-neutral color pixels, the dark channel coverage is reduced according to the dark channel reduction ratio, and the reduced portion is distributed to the remaining channels according to the original proportion of the remaining channels; for pixels located inside the aperture mask, the coverage of each channel is set to zero to form the corrected coverage map.

7. The method according to claim 1, characterized in that, The step of performing partitioned raster rearrangement based on the corrected coverage map, the skeleton line set, the aperture mask, and the boundary distance map specifically includes: dividing the corrected coverage map into multiple micro-blocks; determining the number of target landing points for each printing channel in each micro-block based on the corrected coverage of each pixel in the corresponding channel; establishing a candidate position set in each micro-block, and setting no-entry markers for candidate positions located within the aperture mask and candidate positions that form a cross-aperture connection with the aperture mask; determining candidate position scores based on the corrected coverage of the candidate positions, their adjacency with the skeleton line set, and their adjacency with the high-risk direction of the boundary; selecting target landing point positions according to the candidate position scores, and updating the adjacency state after each point selection until the target number of landing points is reached.

8. The method according to claim 7, characterized in that, The process of determining candidate location scores based on the corrected coverage of the candidate location, its adjacency with the skeleton line set, and its adjacency with high-risk boundary directions specifically includes: The basic coverage contribution is determined based on the corrected coverage rate of the candidate location on the corresponding channel; the skeleton continuity reward is determined based on the continuity relationship between the candidate location and the landing point of the selected skeleton direction, combined with the skeleton continuity constraint coefficient; the boundary expansion penalty is determined based on the selected continuity length of the candidate location in the high-risk direction of the boundary, combined with the boundary expansion penalty coefficient; and the candidate location score is determined based on the basic coverage contribution, skeleton continuity reward, and boundary expansion penalty.

9. The method according to claim 1, characterized in that, The step of generating target raster data based on the target dot map and outputting it to the inkjet printing device specifically includes: The segmented target inkjet point bitmaps corresponding to each micro-block are stitched together according to their position index in the whole map to obtain the full-frame target inkjet point bitmap; when there are conflicting landing points in the overlapping area, the target landing points corresponding to the boundary sensitive area or the skeleton line set neighborhood are retained; the target raster data is generated according to the stitched full-frame target inkjet point bitmap and the channel configuration of the printing device, and the target raster data is sent to the inkjet printing device.

10. A data processing system for inkjet printing of calligraphy and paintings on Xuan paper, characterized in that, include: The parameter reading module is configured to read the corresponding paper attribute parameters according to the Xuan paper type information. The paper attribute parameters include at least the main fiber direction angle, boundary influence distance threshold, neighborhood statistical window size, skeleton continuity constraint coefficient, boundary expansion penalty coefficient, pore protection buffer width, dark channel reduction upper limit, and diffusion risk weight reorganization. The preprocessing and recognition module is configured to preprocess the digital image of calligraphy and painting and generate a set of analysis blocks, perform structural region recognition on the set of analysis blocks, and obtain a skeleton line set, a boundary distance map, a boundary sensitive region mask, a pore mask, and a slow-release paving area mask. The coverage generation module is configured to generate a basic coverage map based on the digital image of the calligraphy and painting and the channel configuration of the printing device, and to generate a neighborhood expected coverage map and a local orientation map based on the basic coverage map; The risk map generation module is configured to generate a bleeding risk map based on the skeleton line set, boundary distance map, pore mask, slow-release spread area mask, neighborhood expected coverage map, local orientation map, and the paper property parameters. The channel reallocation module is configured to reallocate channels on the base coverage map based on the permeation risk map to obtain a modified coverage map. The grid rearrangement and data output module is configured to perform partitioned grid rearrangement based on the corrected coverage map, the skeleton line set, the aperture mask, and the boundary distance map, generate a target nozzle bitmap, generate target grid data based on the target nozzle bitmap, and output it to the inkjet printing device.

Images