A method for fusing two-viewpoint images of a lenticular grating for 3D display

Abstract

The present invention discloses a method for fusing two-viewpoint images of a lenticular grating for 3D display, including: calculating initial coefficients corresponding to the R, G, and B sub-pixels in the first row and the first column from a preset formula to obtain initial coefficient values representing different viewpoint weights; starting from the initial coefficients, performing recursive calculations for the remaining sub-pixel positions through a row recurrence formula and a column recurrence formula to obtain complete coefficient array data; acquiring two-viewpoint image data, and extracting sub-pixel samples from the corresponding viewpoints according to the integer part in the complete coefficient array to obtain a preliminarily collected sub-pixel set; for the preliminarily collected sub-pixel set, using the decimal part in the coefficient array as a basis, performing weight distribution on the sub-pixel samples of the two viewpoints through a linear interpolation method to obtain weighted sub-pixel values; obtaining a fused sub-pixel result through the weighted sub-pixel values, and combining the fused sub-pixel result into a complete image to generate a final 3D display image.

Description

Technical Field

[0001] The present invention relates to the technical field of 3D image generation, and in particular to a method for fusing two-viewpoint images of a lenticular grating for 3D display. Background Art

[0002] As an important branch of modern visual technology, the naked-eye 3D display technology has an irreplaceable position in the fields of augmented reality, virtual reality, and consumer electronics. Its core lies in achieving a stereoscopic visual effect that can be experienced without wearing equipment through optical and image processing means. However, there are still many deficiencies in the current two-viewpoint image fusion method based on lenticular gratings, which limits its wide application.

[0003] When the current two-viewpoint image fusion method based on lenticular gratings fuses two images, it is restricted by lenticular grating parameters (such as tilt angle, grating width) and image resolution, which may cause the fused image to appear blurred or distorted in some details. When traditional methods process images, they often cannot effectively solve the ghosting and distortion problems in the parallax image, affecting the authenticity of the display effect and the user's viewing experience;

[0004] The viewing angle range of naked-eye 3D display devices is usually small, that is, a complete 3D effect can only be observed at a specific angle. This limits the viewing freedom of users and reduces the practicality of the technology;

[0005] Existing image fusion algorithms need to consider various factors, such as lenticular grating parameters, image resolution, number of viewpoints, etc., which results in complex algorithm design and large computational amount. This complexity not only affects the real-time performance of the device but also may lead to a decline in user experience.

[0006] Although existing research has tried to alleviate these problems by adjusting grating parameters or optimizing fusion algorithms, the effects are often not satisfactory, and it is difficult to balance clarity, viewing angle expansion, and real-time performance simultaneously. In this field, the core challenges faced by technological progress mainly focus on three aspects: the generation efficiency of the coefficient array, the sub-pixel acquisition accuracy, and the smoothness of the fusion process. Due to the lack of systematic optimization in the generation of the coefficient array, the calculation is complex and the result consistency is insufficient, which in turn affects the accuracy of subsequent image processing. At the same time, the extraction of viewpoint information in the sub-pixel acquisition process is not fine enough, resulting in blurred or distorted details in the fused image. In addition, the imperfections in weight assignment and weighted calculation in the fusion algorithm often cause unnatural viewpoint transitions, limiting the expansion of the viewing angle range.

[0007] In summary, although the two-viewpoint image fusion technology of lenticular gratings has broad application prospects in the field of autostereoscopic 3D displays, further research and improvement are still needed in aspects such as image resolution, viewing angle range, and image fusion algorithms. How to optimize the generation process of the coefficient array and improve the sub-pixel acquisition and fusion method to enhance image clarity and viewing angle range while reducing the algorithm complexity has become a key issue that this research urgently needs to overcome. The solution to this problem will directly drive the breakthrough of autostereoscopic 3D display technology in real-time performance and user experience, laying a foundation for its popularization in multi-scenario applications. Summary of the Invention

[0008] In order to solve the above-mentioned existing technical problems, the present invention provides a method for fusing two-viewpoint images of a lenticular grating for 3D display.

[0009] The technical solution of the present invention is implemented as follows:

[0010] A method for fusing two-viewpoint images of a lenticular grating for 3D display, comprising

[0011] S1. According to the tilt angle parameter, the sub-pixel width covered by the grating horizontally, the viewing point number control, and the resolution adjustment, calculate the initial coefficients corresponding to the R, G, and B sub-pixels in the first row and the first column from a preset formula to obtain the initial coefficient values representing different viewing point weights;

[0012] S2. Starting from the initial coefficients, perform recursive calculations for the remaining sub-pixel positions through the row recurrence formula and the column recurrence formula to obtain the complete coefficient array data;

[0013] S3. Obtain the two-viewpoint image data, and extract sub-pixel samples from the corresponding viewpoints according to the integer part in the complete coefficient array to obtain a preliminarily collected sub-pixel set;

[0014] S4. For the preliminarily collected sub-pixel set, use the decimal part in the coefficient array as a basis, and perform weight distribution on the sub-pixel samples of the two viewpoints through linear interpolation to obtain the weighted sub-pixel values;

[0015] S5. Through the weighted sub-pixel values, perform fusion calculations on the sub-pixel samples of the two viewpoints to obtain the fused sub-pixel results, and combine the fused sub-pixel results into a complete image to generate the final 3D display image.

[0016] Further, the process of obtaining the initial coefficient values representing different viewing point weights in step S1 includes:

[0017] Obtain the distribution data of the sub-pixel width through the tilt angle parameter and the horizontal grating to obtain a preliminary spatial mapping relationship;

[0018] Calculate the density of the view point distribution according to the sub-pixel width and the number of view points, and determine the preliminary allocation of the view point weights;

[0019] Use a preset formula to calculate the initial coefficients of the R sub-pixels, G sub-pixels, and B sub-pixels, and obtain the initial coefficient values of each sub-pixel;

[0020] Correct the initial coefficient values through resolution adjustment, and determine whether the corrected coefficients meet the view point weight requirements;

[0021] If the corrected coefficients deviate from the view point weights, re-adjust the distribution of the sub-pixel widths horizontally by raster to obtain updated initial coefficient values;

[0022] According to the updated initial coefficient values and the number of view points, use a linear regression algorithm to optimize the allocation of the view point weights and obtain the final weight coefficients.

[0023] Further, the step S1 further includes:

[0024] S101. According to the sub-pixel width and the number of view points, use a linear regression algorithm to optimize the allocation of the view point weights, and verify the display effects of the R sub-pixels, G sub-pixels, and B sub-pixels through the final weight coefficients;

[0025] The process of verifying the display effects in step S101 includes: calculating the preliminary weight coefficients by using a linear regression algorithm according to the sub-pixel width and the number of view points to obtain the optimized allocation data;

[0026] According to the optimized allocation data, adjust the display data of the R sub-pixels to obtain the adjusted first sub-pixel data;

[0027] Match the display data of the G sub-pixels through the adjusted first sub-pixel data to obtain the second sub-pixel data;

[0028] Synchronize the display data of the B sub-pixels according to the second sub-pixel data to obtain the third sub-pixel data;

[0029] Use the mean calculation method to fuse the first sub-pixel data, the second sub-pixel data, and the third sub-pixel data to obtain the comprehensive display data;

[0030] Judge the integrity of the data processing through the matching relationship between the comprehensive display data and the number of view points to obtain the verification result;

[0031] According to the verification result, if the integrity is insufficient, adjust the weight coefficients through the distribution of the sub-pixel widths to obtain the updated comprehensive display data.

[0032] Further, the process of obtaining the complete coefficient array data in the step S2 includes:

[0033] Starting from the initial coefficients, calculate the coefficient increments and coefficient changes for the sub-pixel positions according to the row recurrence formula and the column recurrence formula to obtain the preliminary coefficient distribution;

[0034] Perform a spatial mapping on the preliminary coefficient distribution using the position distribution to obtain the adjusted coefficient distribution data;

[0035] Based on the coefficient increments and coefficient changes, determine whether the adjusted coefficient distribution meets the requirements of the complete array. If not, recalculate and adjust the coefficient increments to obtain the updated coefficient distribution;

[0036] According to the updated coefficient distribution, optimize the coefficient changes using column calculations to obtain the optimized coefficient data;

[0037] Extract the complete array structure from the optimized coefficient data using a data acquisition tool to determine the final coefficient array;

[0038] For the final coefficient array, verify the coefficient distribution using a linear regression algorithm to obtain the verified array data;

[0039] Based on the verified array data, obtain the final distribution of the sub-pixel positions to determine the output result of the complete coefficient array.

[0040] Furthermore, the step S2 further includes:

[0041] S201. Based on the coefficient increments and coefficient changes, determine whether the adjusted coefficient distribution meets the requirements of the complete array. If not, recalculate and adjust the coefficient increments to obtain the updated coefficient distribution;

[0042] Specifically, according to the complete coefficient array data, perform a spatial mapping on the coefficient data using the position distribution to obtain the adjusted coefficient distribution data;

[0043] Based on the coefficient increments and coefficient changes, determine whether the adjusted coefficient distribution meets the requirements of the complete array to obtain a preliminary judgment result;

[0044] If the preliminary judgment result shows non-compliance, recalculate and adjust the coefficient increments to obtain the updated coefficient distribution;

[0045] According to the updated coefficient distribution, optimize the coefficient changes using column calculations to obtain the optimized coefficient data;

[0046] Extract the complete array structure from the optimized coefficient data using a data acquisition tool to determine the final coefficient array;

[0047] For the final coefficient array, verify the coefficient distribution using a linear regression algorithm to obtain the verified array data;

[0048] According to the verified array data, obtain the final distribution of sub-pixel positions and determine the output result of the complete coefficient array.

[0049] Further, the process of obtaining the preliminarily collected sub-pixel set in step S3 includes:

[0050] Obtain two-viewpoint image data, extract sub-pixel samples from the corresponding viewpoints according to the integer part in the complete coefficient array, and obtain the preliminarily collected sub-pixel set;

[0051] Through the preliminarily collected sub-pixel set, determine the sampling positions according to the integer part in the coefficient array to obtain the adjusted sampling position set;

[0052] Adopt a collection method to extract sub-pixel samples from the corresponding viewpoints, update the adjusted sampling position set, and obtain the optimized sub-pixel sample set;

[0053] For the optimized sub-pixel sample set, use a data acquisition tool to determine whether the sub-pixel samples meet the integrity requirements and obtain the integrity judgment result;

[0054] If the integrity judgment result is not met, re-adjust the sampling positions through the extraction process to obtain the re-optimized sub-pixel set;

[0055] According to the re-optimized sub-pixel set, use the support vector machine algorithm to classify the sub-pixel samples and obtain the classified sub-pixel set;

[0056] Through the classified sub-pixel set, obtain the spatial distribution characteristics of the sampling positions to obtain the feature-enhanced sub-pixel set;

[0057] For the feature-enhanced sub-pixel set, use a clustering algorithm to group the sub-pixel samples of the image data to obtain the grouped sub-pixel set;

[0058] Through the grouped sub-pixel set, determine the matching degree between the corresponding viewpoints and the sampling positions to determine the final sub-pixel set.

[0059] Further, the process of obtaining the weighted sub-pixel values in step S4 includes:

[0060] For the preliminarily collected sub-pixel set, use the decimal part in the coefficient array as a basis, and perform weight allocation on the sub-pixel samples of the two viewpoints through linear interpolation to obtain the weighted sub-pixel values;

[0061] Through the weighted sub-pixel values, obtain the corresponding spatial position information from the two-viewpoint samples to determine the sub-pixel data after position adjustment.

[0062] Further, the step S4 further includes:

[0063] For the sub-pixel data after position adjustment, use a clustering algorithm to group the samples to obtain a set of grouped sub-pixels;

[0064] According to the set of grouped sub-pixels, use a data processing tool to judge the distribution consistency among the samples to obtain the sub-pixel data after consistency adjustment;

[0065] The process of obtaining the sub-pixel data after consistency adjustment includes:

[0066] According to the set of grouped sub-pixels, use a data processing tool to analyze the distribution characteristics among the samples to obtain a distribution consistency evaluation result;

[0067] Based on the distribution consistency evaluation result, use a statistical method to calculate the deviation value among the samples to determine the consistency adjustment parameter;

[0068] According to the consistency adjustment parameter, use a linear transformation method to correct the sub-pixel data to obtain the preliminarily adjusted sub-pixel data;

[0069] For the preliminarily adjusted sub-pixel data, use a clustering algorithm to re-group the samples to obtain grouped consistency features;

[0070] Based on the grouped consistency features, judge the matching degree between the sample distribution and the viewpoint samples to obtain a quantization value of the matching degree;

[0071] If the quantization value of the matching degree is lower than the preset threshold, re-adjust the sub-pixel weights through interpolation calculation to obtain the optimized sub-pixel values;

[0072] According to the optimized sub-pixel values, use a support vector machine algorithm to classify the samples to determine the set of classified sub-pixels;

[0073] Based on the set of classified sub-pixels, obtain the spatial distribution features of the samples to obtain the sub-pixel data with enhanced features;

[0074] For the sub-pixel data with enhanced features, use a data processing tool to judge the distribution consistency among the samples to obtain the sub-pixel data after consistency adjustment.

[0075] Further, the process of obtaining the fused sub-pixel result in the step S5 includes:

[0076] Through the weighted sub-pixel values, use the weighted summation formula to perform a fusion calculation on the sub-pixel samples of the two viewpoints to obtain the fused sub-pixel result;

[0077] The step S5 further includes optimizing the fused sub-pixel result, including:

[0078] According to the fusion result, obtain the spatial sample distribution characteristics of the two viewpoints, determine the sub-pixel result after distribution adjustment, group the spatial samples through a clustering algorithm, and obtain the grouped sample set;

[0079] Adopt the grouped sample set to obtain the change trend of the pixel weights and judge the fusion result after the change trend is adjusted;

[0080] If the matching degree between the fusion result after the change trend is adjusted and the viewpoint samples is lower than the preset threshold, re-adjust the pixel weights through interpolation calculation to obtain the optimized sub-pixel result;

[0081] According to the optimized sub-pixel result, classify the sample fusion using a support vector machine algorithm to determine the classified spatial sample set;

[0082] Through the classified spatial sample set, obtain the behavior distribution characteristics of the fusion calculation to get the sub-pixel result with enhanced characteristics;

[0083] Adopt the sub-pixel result with enhanced characteristics and perform a fusion calculation on the viewpoint samples through a weighted summation formula to determine the fusion result after trend optimization;

[0084] According to the fusion result after trend optimization, obtain the final distribution characteristics of the spatial samples to get the finally adjusted sub-pixel set;

[0085] Adopt the finally adjusted sub-pixel set, analyze the spatial distribution trend of the pixel weights through distribution characteristic analysis to obtain the optimized sub-pixel fusion result.

[0086] Furthermore, it also includes:

[0087] S6. Judge whether the fused sub-pixel result exceeds the preset threshold range. If the fused sub-pixel result exceeds the preset threshold range, re-perform weight allocation and fusion calculation by adjusting the weight ratio in the coefficient array to obtain the corrected fused sub-pixel data;

[0088] S7. According to the corrected fused sub-pixel data, combine the resolution adjustment parameters to perform pixel recombination on the image to obtain the final autostereoscopic 3D display image data.

[0089] Compared with the prior art, the present invention has the following beneficial effects:

[0090] 1. Through optimizing the generation of the coefficient array and the acquisition and fusion of sub-pixels, in the initial coefficient calculation stage, by means of the tilt angle parameter, the sub-pixel width covered by the grating horizontally, the control of the number of viewpoints, and the resolution adjustment, combined with the linear regression algorithm to optimize the distribution of the viewpoint weights, the accuracy and reliability of the initial coefficients are ensured. This not only improves the generation efficiency of the coefficient array but also provides accurate input for the subsequent sub-pixel acquisition and fusion, thus reducing the image blurring and distortion phenomena and significantly enhancing the image quality.

[0091] 2. Aiming at the problem of limited viewing angle range, in the process of sub-pixel acquisition and fusion, this invention extracts sub-pixel samples from the corresponding viewpoints according to the integer part in the complete coefficient array and assigns weights through the decimal part, realizing the smooth transition and fusion of the two-viewpoint images. In addition, by using the clustering algorithm to group the spatial samples and dynamically adjusting the sub-pixel acquisition and fusion process according to the number of viewpoints, the 3D display image can adapt to different viewing angle requirements, effectively expanding the viewing angle range and enhancing the user's viewing freedom.

[0092] 3. This invention adopts the row recurrence formula and the column recurrence formula to efficiently calculate the complete coefficient array data starting from the initial coefficients. At the same time, by judging whether the coefficient distribution meets the requirements of the complete array and readjusting the coefficient increment when necessary, the integrity and consistency of the coefficient array are ensured, significantly reducing the computational complexity and improving the real-time performance of the system, making the autostereoscopic 3D display technology more efficient and practical in actual applications.

[0093] 4. Through a series of optimization measures, such as performing threshold judgment and correction on the fusion result, using the support vector machine algorithm to classify the data, and adjusting the sub-pixel display data, etc., the quality and display effect of the fused image are further enhanced. This not only ensures the clarity and consistency of the image at different viewpoints but also enhances the 3D display characteristics, providing users with a more realistic and comfortable 3D visual experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] Figure 1 It is a flowchart of a method for fusing two-viewpoint images of a lenticular grating for 3D display in Embodiment 1.

[0095] Figure 2 It is a flowchart of a method for fusing two-viewpoint images of a lenticular grating for 3D display in Embodiment 2.

[0096] Figure 3 It is another flowchart of a method for fusing two-viewpoint images of a lenticular grating for 3D display in Embodiment 2. DETAILED DESCRIPTION OF THE INVENTION

[0097] To make the objectives, features, and advantages of the present invention more obvious and understandable, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the embodiments described below are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0098] Embodiment 1

[0099] As Figure 1 shown, this embodiment provides a method for fusing two-viewpoint images of a lenticular grating for 3D display, including

[0100] S1. According to the tilt angle parameter, the sub-pixel width covered by the grating horizontally, the viewpoint number control, and the resolution adjustment, calculate the initial coefficients corresponding to the R, G, and B sub-pixels in the first row and first column from a preset formula to obtain the initial coefficient values representing different viewpoint weights;

[0101] S2. Starting from the initial coefficients, perform recursive calculations for the remaining sub-pixel positions through the row recurrence formula and the column recurrence formula to obtain the complete coefficient array data;

[0102] S3. Obtain the two-viewpoint image data, and extract the sub-pixel samples from the corresponding viewpoints according to the integer part in the complete coefficient array to obtain the initially collected sub-pixel set;

[0103] S4. For the initially collected sub-pixel set, use the decimal part in the coefficient array as a basis, and perform weight distribution on the sub-pixel samples of the two viewpoints through linear interpolation to obtain the weighted sub-pixel values;

[0104] S5. Through the weighted sub-pixel values, perform fusion calculations on the sub-pixel samples of the two viewpoints to obtain the fused sub-pixel results, and combine the fused sub-pixel results into a complete image to generate the final 3D display image.

[0105] Further, the process of obtaining the initial coefficient values representing different viewpoint weights in step S1 includes:

[0106] Obtain the distribution data of the sub-pixel width through the tilt angle parameter and the grating horizontally to obtain a preliminary spatial mapping relationship;

[0107] Calculate the density of the viewpoint distribution according to the sub-pixel width and the number of viewpoints to determine the preliminary distribution of the viewpoint weights;

[0108] Perform initial coefficient calculations on the R sub-pixel, G sub-pixel, and B sub-pixel using a preset formula to obtain the initial coefficient values of each sub-pixel;

[0109] Calibrate the initial coefficient value through resolution adjustment, and determine whether the calibrated coefficient meets the view point weight requirement;

[0110] If the calibrated coefficient deviates from the view point weight, re-adjust the distribution of the sub-pixel width horizontally through the grating to obtain the updated initial coefficient value;

[0111] According to the updated initial coefficient value and the number of view points, use the linear regression algorithm to optimize the allocation of view point weights to obtain the final weight coefficient;

[0112] In one embodiment, the process of obtaining the initial coefficient values representing different view point weights in step S1 can be as follows:

[0113] According to the tilt angle parameter of 15 degrees and the sub-pixel width covered horizontally by the grating of 0.2 microns, obtain the distribution data of the sub-pixel width through the spatial mapping algorithm to obtain the preliminary spatial mapping relationship;

[0114] Through the preliminary spatial mapping relationship and the number of view points of 8, use the density distribution formula to calculate the density of the view point distribution, and determine the preliminary allocation of view point weights as [0.12, 0.15, 0.18, 0.20, 0.18, 0.15, 0.12, 0.10];

[0115] Use the preset formula to calculate the initial coefficient of the first R sub-pixel in the first row, input the tilt angle parameter and the sub-pixel width, and obtain the initial coefficient value of the R sub-pixel as 0.25;

[0116] Use the preset formula to calculate the initial coefficient of the first G sub-pixel in the first row, input the same parameters, and obtain the initial coefficient value of the G sub-pixel as 0.30;

[0117] Use the preset formula to calculate the initial coefficient of the first B sub-pixel in the first row, input the same parameters, and obtain the initial coefficient value of the B sub-pixel as 0.35;

[0118] Through the resolution adjustment to 1920x1080, calibrate the initial coefficient values of the R sub-pixel, G sub-pixel, and B sub-pixel, and obtain the calibrated coefficients as 0.24, 0.29, and 0.34 respectively. If the calibrated coefficients deviate from the view point weight requirement, re-adjust the distribution of the sub-pixel width horizontally through the grating to 0.18 microns, and obtain the updated initial coefficient values as 0.23, 0.28, and 0.33.

[0119] Furthermore, step S1 further includes:

[0120] S101. According to the sub-pixel width and the number of view points, use the linear regression algorithm to optimize the allocation of view point weights, and verify the display effects of the R sub-pixel, G sub-pixel, and B sub-pixel through the final weight coefficient;

[0121] The process of verifying the display effect in step S101 includes:

[0122] Using the sub-pixel width and the number of viewpoints, the preliminary weight coefficients are calculated by the linear regression algorithm to obtain the optimized allocation data;

[0123] According to the optimized allocation data, the display data of the R sub-pixel is adjusted to obtain the adjusted first sub-pixel data;

[0124] Using the adjusted first sub-pixel data, the display data of the G sub-pixel is matched to obtain the second sub-pixel data;

[0125] According to the second sub-pixel data, the display data of the B sub-pixel is synchronized to obtain the third sub-pixel data;

[0126] Using the mean calculation method, the first sub-pixel data, the second sub-pixel data, and the third sub-pixel data are fused to obtain the comprehensive display data;

[0127] Based on the matching relationship between the comprehensive display data and the number of viewpoints, the integrity of the data processing is judged to obtain the verification result;

[0128] According to the verification result, if the integrity is insufficient, the weight coefficients are adjusted by the distribution of the sub-pixel width to obtain the updated comprehensive display data;

[0129] In one embodiment, the process of verifying the display effect in step S101 may be as follows:

[0130] When calculating the preliminary weight coefficients by the linear regression algorithm using the sub-pixel width and the number of viewpoints, it can be regarded as a data-based distribution optimization process;

[0131] The core of linear regression is to find a reasonable weight allocation relationship through the known sub-pixel width and the number of viewpoints. For example, assuming the sub-pixel width is 0.2 mm and the number of viewpoints is 8, the weight coefficients of each viewpoint can be preliminarily determined through regression analysis. For example, the weight of the viewpoint close to the center is set to 0.15, while the weight of the edge viewpoint is 0.10;

[0132] Specifically, when adjusting the display data of the R sub-pixel according to the optimized allocation data, it can be processed from the perspective of brightness distribution;

[0133] Exemplarily, if the brightness value of the original R sub-pixel is 200, after being adjusted by the weight coefficient of 0.15, the new brightness value may become 230;

[0134] In one embodiment, when performing matching processing on the display data of G sub-pixels, the adjustment result of R sub-pixels can be referred to. For example, if the brightness of R sub-pixels is adjusted to 230, the brightness of G sub-pixels can be matched and adjusted to 225 according to their spatial relationship with R sub-pixels. For example, when synchronizing the display data of B sub-pixels, it can be achieved through data alignment on the time axis;

[0135] If the brightness of the second sub-pixel data is 225 at a certain moment, then the B sub-pixel can be synchronously adjusted to 220 to ensure the time consistency in display among the three;

[0136] It can be understood that when fusing the first, second, and third sub-pixel data using the mean calculation method, the aim is to obtain a balanced comprehensive display data;

[0137] For example, if the brightness values of the three are 230, 225, and 220 respectively, a comprehensive value of 225 may be obtained after mean calculation.

[0138] This fusion method can smooth the differences between sub-pixels, make the display data more integral, and effectively improve the display quality under multiple viewpoints;

[0139] In a possible implementation, when judging the integrity through the matching relationship between the comprehensive display data and the number of viewpoints, a threshold standard can be set. For example, if the number of viewpoints is 8, the brightness distribution of the comprehensive display data should fluctuate between 220 and 230. If it exceeds this range, it is considered that the integrity is insufficient;

[0140] Specifically, if the verification result shows insufficient integrity, the weight coefficient is adjusted through the distribution of the sub-pixel width. For example, the sub-pixel width is fine-tuned from 0.2 mm to 0.22 mm. After recalculating the weight coefficient, the brightness value may be updated from 225 to 228;

[0141] It should be noted that the above examples are all centered around the processing of sub-pixel data, and through the progressive logic from preliminary calculation to final adjustment, the rigor of the scheme is ensured;

[0142] Exemplarily, this method can improve the efficiency of data processing and the stability of display in multi-viewpoint display technology, bringing a higher-quality visual experience to users.

[0143] Furthermore, the process of obtaining the complete coefficient array data in step S2 includes:

[0144] Starting from the initial coefficient, through the row recurrence formula and the column recurrence formula, calculating the coefficient increment and coefficient change for the sub-pixel positions to obtain the preliminary coefficient distribution;

[0145] Perform spatial mapping on the preliminary coefficient distribution using the position distribution to obtain the adjusted coefficient distribution data;

[0146] Based on the coefficient increment and coefficient change, determine whether the adjusted coefficient distribution meets the requirements of the complete array. If not, recalculate the coefficient increment through calculation to obtain the updated coefficient distribution;

[0147] According to the updated coefficient distribution, optimize the coefficient change using column calculation to obtain the optimized coefficient data;

[0148] Extract the complete array structure from the optimized coefficient data through a data acquisition tool to determine the final coefficient array;

[0149] For the final coefficient array, verify the coefficient distribution using a linear regression algorithm to obtain the verified array data;

[0150] Based on the verified array data, obtain the final distribution of the sub-pixel positions to determine the output result of the complete coefficient array;

[0151] In one embodiment, the process of obtaining the complete coefficient array data in step S2 can be as follows:

[0152] Starting from the initial coefficient, perform recursive calculation for the remaining sub-pixel positions using the row recurrence formula and column recurrence formula. For example, use the recurrence formula: Calculate to obtain the complete coefficient array data. According to the complete coefficient array data, perform spatial mapping on the coefficient data using the position distribution. For example, map the coefficient data to a two-dimensional spatial grid to obtain the adjusted coefficient distribution data. Based on the coefficient increment and coefficient change, for example, calculate and , determine whether the adjusted coefficient distribution meets the requirements of the complete array to obtain a preliminary judgment result. If the preliminary judgment result shows non-compliance, recalculate the coefficient increment through calculation. For example, adjust to 0.6 to obtain the updated coefficient distribution;

[0153] According to the updated coefficient distribution, optimize the coefficient change using column calculation. For example, optimize to 0.4 to obtain the optimized coefficient data;

[0154] Extract the complete array structure from the optimized coefficient data through a data acquisition tool. For example, use a matrix extraction algorithm to determine the final coefficient array;

[0155] For the final coefficient array, verify the coefficient distribution using a linear regression algorithm. For example, use the least squares method to fit the coefficient distribution to obtain the verified array data;

[0156] Based on the verified array data, obtain the final distribution of sub-pixel positions, for example, extract the sub-pixel coordinates (x, y) to determine the output result of the complete coefficient array.

[0157] Furthermore, the step S2 further includes:

[0158] S201. According to the coefficient increment and coefficient change, determine whether the adjusted coefficient distribution meets the requirements of the complete array. If not, recalculate and adjust the coefficient increment to obtain the updated coefficient distribution;

[0159] Specifically, according to the complete coefficient array data, perform spatial mapping on the coefficient data using the position distribution to obtain the adjusted coefficient distribution data;

[0160] Determine whether the adjusted coefficient distribution meets the requirements of the complete array through the coefficient increment and coefficient change to obtain a preliminary judgment result;

[0161] If the preliminary judgment result shows non-compliance, recalculate and adjust the coefficient increment to obtain the updated coefficient distribution;

[0162] According to the updated coefficient distribution, optimize the coefficient change using column calculation to obtain the optimized coefficient data;

[0163] Extract the complete array structure from the optimized coefficient data through a data acquisition tool to determine the final coefficient array;

[0164] For the final coefficient array, verify the coefficient distribution using a linear regression algorithm to obtain the verified array data;

[0165] Based on the verified array data, obtain the final distribution of sub-pixel positions to determine the output result of the complete coefficient array;

[0166] In one embodiment, according to the coefficient increment Δk (such as Δk = 0.05) and coefficient change ΔC (such as ΔC = 0.1), calculate the adjusted coefficient distribution data D_adj using matrix operations;

[0167] Judge through a preset complete array condition (such as the row-column uniformity threshold ε = 0.01) whether it meets the condition. If the uniformity error exceeds ε, then use row calculation (such as Gaussian elimination) to readjust to obtain the updated (such as = 0.03);

[0168] Based on , calculate the updated coefficient distribution through a row recurrence formula (such as ) ;

[0169] Use a column recurrence formula (such as ) to perform optimization, and obtain the optimized (such as = 0.08);

[0170] Through judgment to determine whether it meets the requirements of the complete array. If it still does not meet the requirements, use spatial mapping (such as bilinear interpolation) to adjust the position distribution to obtain the adjusted ;

[0171] Based on , use a linear regression algorithm (such as least squares fitting) to verify the coefficient distribution, and obtain the verified array data . Through extract the final coefficient array (such as a 5×5 matrix) to determine the output result.

[0172] Furthermore, the process of obtaining the preliminarily collected sub-pixel set in step S3 includes:

[0173] Obtain two viewpoint image data, and extract sub-pixel samples from the corresponding viewpoints according to the integer part in the complete coefficient array to obtain the preliminarily collected sub-pixel set;

[0174] Through the preliminarily collected sub-pixel set, determine the sampling positions according to the integer part in the coefficient array to obtain the adjusted sampling position set;

[0175] Adopt a collection method to extract sub-pixel samples from the corresponding viewpoints and update the adjusted sampling position set to obtain the optimized sub-pixel sample set;

[0176] For the optimized sub-pixel sample set, use a data acquisition tool to determine whether the sub-pixel samples meet the integrity requirements to obtain an integrity judgment result;

[0177] If the integrity judgment result does not meet the requirements, re-adjust the sampling positions through the extraction process to obtain the re-optimized sub-pixel set;

[0178] According to the re-optimized sub-pixel set, use a support vector machine algorithm to classify the sub-pixel samples to obtain the classified sub-pixel set;

[0179] Through the classified sub-pixel set, obtain the spatial distribution characteristics of the sampling positions to obtain the feature-enhanced sub-pixel set;

[0180] For the sub-pixel set with enhanced features, a clustering algorithm is used to group the sub-pixel samples of the image data to obtain the grouped sub-pixel set;

[0181] Based on the grouped sub-pixel set, the matching degree between the corresponding viewpoints and the sampling positions is judged to determine the final sub-pixel set;

[0182] In one embodiment, the process of obtaining the preliminarily collected sub-pixel set in step S3 can be as follows:

[0183] Two viewpoint image data are obtained. According to the integer part in the complete coefficient array, sub-pixel samples are extracted from the corresponding viewpoints. For example, sub-pixel samples with coordinates (10, 20) and (15, 25) are respectively extracted from viewpoint A and viewpoint B to obtain the preliminarily collected sub-pixel set;

[0184] Based on the preliminarily collected sub-pixel set, according to the integer part in the coefficient array, the sampling positions are determined. For example, the sampling positions are adjusted to (12, 22) and (17, 27) to obtain the adjusted sampling position set. Using the acquisition method, sub-pixel samples are extracted from the corresponding viewpoints. For example, sub-pixel samples with coordinates (12, 22) and (17, 27) are respectively extracted from viewpoint A and viewpoint B to update the adjusted sampling position set to obtain the optimized sub-pixel sample set;

[0185] For the optimized sub-pixel sample set, through a data acquisition tool, it is judged whether the sub-pixel samples meet the integrity requirements. For example, it is checked whether the RGB values of each sub-pixel sample are complete to obtain the integrity judgment result. If the integrity judgment result is not satisfied, the sampling positions are readjusted through the extraction process. For example, the sampling positions are adjusted to (13, 23) and (18, 28) to obtain the re-optimized sub-pixel set;

[0186] Based on the re-optimized sub-pixel set, a support vector machine algorithm is used to classify the sub-pixel samples. For example, a support vector machine with a kernel function of RBF is used to classify the sub-pixel samples into two categories to obtain the classified sub-pixel set;

[0187] Based on the classified sub-pixel set, the spatial distribution characteristics of the sampling positions are obtained. For example, the density distribution of the sub-pixel samples in the image is calculated to obtain the sub-pixel set with enhanced features;

[0188] For the sub-pixel set with enhanced features, a clustering algorithm is used to group the sub-pixel samples of the image data. For example, the K-means algorithm is used to group the sub-pixel samples into three groups to obtain the grouped sub-pixel set;

[0189] Based on the grouped sub-pixel sets, determine the matching degree between the corresponding viewpoints and the sampling positions. For example, calculate the similarity of the sampling positions in viewpoints A and B to determine the final sub-pixel set.

[0190] Further, the process of obtaining the weighted sub-pixel values in step S4 includes:

[0191] For the initially collected sub-pixel sets, use the fractional parts in the coefficient array as a basis, and through linear interpolation, allocate weights to the sub-pixel samples of the two viewpoints to obtain the weighted sub-pixel values;

[0192] Based on the weighted sub-pixel values, obtain the corresponding spatial position information from the two viewpoint samples to determine the sub-pixel data after position adjustment;

[0193] In one embodiment, the process of obtaining the weighted sub-pixel values in step S4 can be as follows:

[0194] For the initially collected sub-pixel sets, use the fractional parts in the coefficient array (such as 0.25 and 0.75) as the weight basis, and through bilinear interpolation, perform weighted calculations on the sub-pixel samples (such as the red channel values R1 = 120 and R2 = 160) of the left and right viewpoints to obtain the weighted sub-pixel values (R = 0.25×120 + 0.75×160 = 150);

[0195] Based on the weighted sub-pixel values, in combination with the disparity mapping relationship (such as the disparity offset Δd = 2.3 pixels), extract the spatial coordinates (x1 = 105.2, y1 = 58.7; x2 = 107.5, y2 = 58.7) from the two viewpoint samples to determine the sub-pixel data after position adjustment;

[0196] For the sub-pixel data after position adjustment, use the K-means clustering algorithm (K = 3, Euclidean distance threshold ε = 5.0) to group the samples to obtain a set of sub-pixels containing 3 categories;

[0197] According to the grouped sub-pixel sets, calculate the sample distribution variance (such as the variance ratio 0.92) through principal component analysis (PCA). If the variance ratio is lower than the threshold 0.95, then use cubic spline interpolation to recalculate the weights (such as the adjusted R = 148) to obtain the optimized sub-pixel values;

[0198] For the optimized sub-pixel values, use an SVM classifier (with the kernel function as RBF, γ = 0.1) to classify the samples and divide them into a set of sub-pixels with clear boundaries (classification accuracy 98.2%);

[0199] From the classified sub-pixel set, local binary pattern (LBP) features (such as 8-neighborhood coding 10110011) are extracted to obtain sub-pixel data with enhanced features. For the sub-pixel data with enhanced features, a connected component detection algorithm (minimum area threshold A = 10 pixels) is used to judge the integrity of the sample. If the missing area is greater than A, the viewpoint image is resampled (such as resampling coordinates x = 110, y = 60) to obtain an adjusted sub-pixel set;

[0200] According to the adjusted sub-pixel set, the viewpoint matching degree is judged by normalized cross-correlation (NCC > 0.9) to determine the final sub-pixel set.

[0201] Furthermore, the step S4 further includes:

[0202] For the sub-pixel data after position adjustment, a clustering algorithm is used to group the samples to obtain a grouped sub-pixel set;

[0203] According to the grouped sub-pixel set, a data processing tool is used to judge the distribution consistency among the samples to obtain sub-pixel data after consistency adjustment;

[0204] The process of obtaining the sub-pixel data after consistency adjustment includes:

[0205] According to the grouped sub-pixel set, a data processing tool is used to analyze the distribution characteristics among the samples to obtain a distribution consistency evaluation result;

[0206] Based on the distribution consistency evaluation result, a statistical method is used to calculate the deviation value among the samples to determine the consistency adjustment parameter;

[0207] According to the consistency adjustment parameter, the sub-pixel data is corrected by a linear transformation method to obtain preliminarily adjusted sub-pixel data;

[0208] For the preliminarily adjusted sub-pixel data, a clustering algorithm is used to regroup the samples to obtain grouped consistency features;

[0209] Based on the grouped consistency features, the matching degree between the sample distribution and the viewpoint sample is judged to obtain a quantization value of the matching degree;

[0210] If the quantization value of the matching degree is lower than the preset threshold, the sub-pixel weights are readjusted by interpolation calculation to obtain optimized sub-pixel values;

[0211] According to the optimized sub-pixel values, a support vector machine algorithm is used to classify the samples to determine the classified sub-pixel set;

[0212] From the classified sub-pixel set, the spatial distribution characteristics of the samples are obtained to obtain sub-pixel data with enhanced features;

[0213] For the sub-pixel data with enhanced features, the data processing tool is used to judge the distribution consistency among samples, and the sub-pixel data after consistency adjustment is obtained.

[0214] Further, the process of obtaining the fused sub-pixel result in step S5 includes:

[0215] Using the weighted sub-pixel values, the sub-pixel samples of two viewpoints are fused and calculated by the weighted summation formula to obtain the fused sub-pixel result;

[0216] Step S5 further includes optimizing the fused sub-pixel result, including:

[0217] According to the fusion result, the spatial sample distribution characteristics of two viewpoints are obtained, the sub-pixel result after distribution adjustment is determined, and the spatial samples are grouped by the clustering algorithm to obtain the grouped sample set;

[0218] Using the grouped sample set, the change trend of the pixel weights is obtained, and the fused result after change trend adjustment is judged;

[0219] If the matching degree between the fused result after change trend adjustment and the viewpoint samples is lower than the preset threshold, the pixel weights are re-adjusted by interpolation calculation to obtain the optimized sub-pixel result;

[0220] According to the optimized sub-pixel result, the support vector machine algorithm is used to classify the sample fusion to determine the classified spatial sample set;

[0221] Through the classified spatial sample set, the behavior distribution characteristics of the fusion calculation are obtained to get the sub-pixel result with enhanced characteristics;

[0222] Using the sub-pixel result with enhanced characteristics, the viewpoint samples are fused and calculated by the weighted summation formula to determine the fused result after trend optimization;

[0223] According to the fused result after trend optimization, the final distribution characteristics of the spatial samples are obtained to get the finally adjusted sub-pixel set;

[0224] Using the finally adjusted sub-pixel set, through the distribution characteristic analysis, the spatial distribution trend of the pixel weights is judged to obtain the optimized sub-pixel fusion result;

[0225] In one embodiment, according to the fusion result, the K-means clustering algorithm is used to analyze the spatial sample distribution characteristics of two viewpoints, the number of clustering centers is set to 5, and the sub-pixel result after distribution adjustment is obtained;

[0226] The spatial samples are grouped by the DBSCAN clustering algorithm, the neighborhood radius is set to 0.5, and the minimum number of samples is 10 to obtain the grouped sample set;

[0227] Using the grouped sample set, calculate the changing trend of pixel weights, fit the trend curve using a linear regression model, and judge the fused result after adjusting the changing trend. If the matching degree between the fused result after adjusting the changing trend and the viewpoint samples is lower than the preset threshold of 0.8, recalculate and adjust the pixel weights through bilinear interpolation to obtain an optimized sub-pixel result;

[0228] According to the optimized sub-pixel result, use the support vector machine algorithm to classify the sample fusion, set the kernel function as the radial basis function, and the classification accuracy as 95%, and determine the classified spatial sample set;

[0229] Through the classified spatial sample set, calculate the behavioral distribution characteristics of the fusion calculation, use principal component analysis to extract the feature vectors, and obtain a sub-pixel result with enhanced characteristics;

[0230] Using the sub-pixel result with enhanced characteristics, perform a fusion calculation on the viewpoint samples through the weighted summation formula, with weight coefficients of 0.6 and 0.4, and determine the fused result after trend optimization;

[0231] According to the fused result after trend optimization, use the Gaussian mixture model to obtain the final distribution characteristics of the spatial samples, and obtain the finally adjusted sub-pixel set;

[0232] Using the finally adjusted sub-pixel set, through the analysis of the distribution characteristics, use the entropy method to judge the spatial distribution trend of the pixel weights, and obtain an optimized fused result.

[0233] Furthermore, it also includes:

[0234] S6. Judge whether the fused sub-pixel result exceeds the preset threshold range. If the fused sub-pixel result exceeds the preset threshold range, re-perform weight allocation and fusion calculation by adjusting the weight ratio in the coefficient array to obtain corrected fused sub-pixel data;

[0235] Exemplarily, if the fused sub-pixel result exceeds the preset threshold range (such as the brightness value exceeds the range of 0 - 255), re-perform weight allocation by adjusting the weight ratio in the coefficient array (such as adjusting the weight from [0.3, 0.7] to [0.4, 0.6]), and calculate using the weighted fusion formula (such as the weighted average method) to obtain corrected fused sub-pixel data;

[0236] S7. According to the corrected fused sub-pixel data, combined with the resolution adjustment parameters, perform pixel recombination on the image to obtain the final naked-eye 3D display image data;

[0237] Specifically, based on the corrected fused sub-pixel data, combined with the resolution adjustment parameters, the pixel recombination method is adjusted by interpolation calculation to obtain the recombined image data. Through the recombined image data, spatial distribution analysis is performed on the sub-pixel values to obtain the data after distribution adjustment;

[0238] Based on the data after distribution adjustment, combined with the adjustment parameters, the clustering algorithm is used to group the pixel recombination results to obtain the grouped data set. If the grouped data set exceeds the preset threshold range, the spatial distribution characteristics are adjusted through the calculation process to obtain the optimized data;

[0239] Through the optimized data, enhancement processing is performed on the naked-eye 3D characteristics to obtain the image data with enhanced characteristics. Based on the image data with enhanced characteristics, combined with the display image requirements, the support vector machine algorithm is used to classify the data to obtain the final naked-eye 3D display data;

[0240] Exemplarily, based on the corrected fused sub-pixel data, combined with the resolution adjustment parameters, the bilinear interpolation algorithm is used to adjust the pixel recombination method. For example, the resolution is adjusted from 1920×1080 to 3840×2160 to obtain the recombined image data;

[0241] Through the recombined image data, spatial distribution analysis is performed on the sub-pixel values, and the Gaussian filtering algorithm is used to smooth the sub-pixel values to obtain the data after distribution adjustment;

[0242] Based on the data after distribution adjustment, combined with the adjustment parameters, the K-means clustering algorithm is used to group the pixel recombination results, and the number of clustering centers is set to 8 to obtain the grouped data set. If the grouped data set exceeds the preset threshold range, for example, the pixel value difference exceeds 0.5, the spatial distribution characteristics are adjusted through the calculation process, and the Laplace operator is used to recalculate the pixel distribution to obtain the optimized data. Through the optimized data, enhancement processing is performed on the naked-eye 3D characteristics, and the edge enhancement algorithm is used to sharpen the image to obtain the image data with enhanced characteristics;

[0243] Based on the image data with enhanced characteristics, combined with the display image requirements, the support vector machine algorithm is used to classify the data, and the kernel function is set to the radial basis function to obtain the final naked-eye 3D display data;

[0244] Through the sub-pixel width and the number of viewpoints, the linear regression algorithm is used to calculate the preliminary weight coefficients. For example, the sub-pixel width is 0.25mm and the number of viewpoints is 9 to obtain the optimized allocation data;

[0245] Adjust the display data of the R sub-pixels according to the optimized allocation data, and use the weighted average algorithm to adjust the R sub-pixel value to 0.8 to obtain the adjusted first sub-pixel data;

[0246] Match the display data of the G sub-pixels through the adjusted first sub-pixel data, and use the least squares method to match the G sub-pixel value with the R sub-pixel value to obtain the second sub-pixel data.

[0247] Embodiment 2

[0248] As Figure 2 shown, this embodiment provides a method for fusing two-viewpoint images of a lenticular grating for 3D display, including

[0249] S1. Calculate the initial coefficients corresponding to the R, G, and B sub-pixels in the first row and first column from a preset formula according to the tilt angle parameter, the sub-pixel width covered by the grating horizontally, the viewpoint number control, and the resolution adjustment, to obtain the initial coefficient values representing different viewpoint weights;

[0250] S2. Starting from the initial coefficients, perform recursive calculations for the remaining sub-pixel positions through the row recurrence formula and the column recurrence formula to obtain the complete coefficient array data;

[0251] S3. Obtain the two-viewpoint image data, and extract the sub-pixel samples from the corresponding viewpoints according to the integer part in the complete coefficient array to obtain the initially collected sub-pixel set;

[0252] S4. For the initially collected sub-pixel set, use the decimal part in the coefficient array as a basis, and perform weight allocation on the sub-pixel samples of the two viewpoints through linear interpolation to obtain the weighted sub-pixel values;

[0253] S5. Perform fusion calculations on the sub-pixel samples of the two viewpoints through the weighted sub-pixel values to obtain the fused sub-pixel results, and combine the fused sub-pixel results into a complete image to generate the final 3D display image.

[0254] Further, the process of obtaining the initial coefficient values representing different viewpoint weights in step S1 includes:

[0255] Obtain the distribution data of the sub-pixel width through the tilt angle parameter and the grating horizontally to obtain a preliminary spatial mapping relationship;

[0256] Calculate the density of the viewpoint distribution according to the sub-pixel width and the number of viewpoints, and determine the preliminary allocation of the viewpoint weights;

[0257] Use a preset formula to calculate the initial coefficients of the R sub-pixels, G sub-pixels, and B sub-pixels to obtain the initial coefficient values of each sub-pixel;

[0258] Calibrate the initial coefficient value through resolution adjustment, and determine whether the calibrated coefficient meets the view point weight requirement;

[0259] If the calibrated coefficient deviates from the view point weight, redistribute the sub-pixel width horizontally through the grating to obtain the updated initial coefficient value;

[0260] According to the updated initial coefficient value and the number of view points, optimize the distribution of view point weights using the linear regression algorithm to obtain the final weight coefficient.

[0261] Furthermore, step S1 further includes:

[0262] S101. According to the sub-pixel width and the number of view points, optimize the distribution of view point weights using the linear regression algorithm, and verify the display effects of the R sub-pixels, G sub-pixels, and B sub-pixels through the final weight coefficient;

[0263] The process of verifying the display effect in step S101 includes: calculating the preliminary weight coefficient using the linear regression algorithm based on the sub-pixel width and the number of view points to obtain the optimized distribution data;

[0264] Adjust the display data of the R sub-pixels according to the optimized distribution data to obtain the adjusted first sub-pixel data;

[0265] Match the display data of the G sub-pixels with the adjusted first sub-pixel data to obtain the second sub-pixel data;

[0266] Synchronize the display data of the B sub-pixels according to the second sub-pixel data to obtain the third sub-pixel data;

[0267] Use the mean calculation method to fuse the first sub-pixel data, the second sub-pixel data, and the third sub-pixel data to obtain the comprehensive display data;

[0268] Judge the integrity of the data processing through the matching relationship between the comprehensive display data and the number of view points to obtain the verification result;

[0269] According to the verification result, if the integrity is insufficient, adjust the weight coefficient through the distribution of the sub-pixel width to obtain the updated comprehensive display data.

[0270] Furthermore, the process of obtaining the complete coefficient array data in step S2 includes:

[0271] Starting from the initial coefficient, calculate the coefficient increment and coefficient change for the sub-pixel position through the row recurrence formula and the column recurrence formula to obtain the preliminary coefficient distribution;

[0272] Perform spatial mapping on the preliminary coefficient distribution using the position distribution to obtain the adjusted coefficient distribution data;

[0273] Judge whether the adjusted coefficient distribution meets the requirements of the complete array through the coefficient increment and coefficient change. If not, recalculate and adjust the coefficient increment to obtain the updated coefficient distribution.

[0274] According to the updated coefficient distribution, optimize the coefficient change by column calculation to obtain the optimized coefficient data.

[0275] Extract the complete array structure from the optimized coefficient data through a data acquisition tool to determine the final coefficient array.

[0276] For the final coefficient array, verify the coefficient distribution using a linear regression algorithm to obtain the verified array data.

[0277] Through the verified array data, obtain the final distribution of the sub-pixel positions to determine the output result of the complete coefficient array.

[0278] Furthermore, the step S2 further includes:

[0279] S201. Judge whether the adjusted coefficient distribution meets the requirements of the complete array through the coefficient increment and coefficient change. If not, recalculate and adjust the coefficient increment to obtain the updated coefficient distribution.

[0280] Specifically, according to the complete coefficient array data, perform spatial mapping on the coefficient data using the position distribution to obtain the adjusted coefficient distribution data.

[0281] Judge whether the adjusted coefficient distribution meets the requirements of the complete array through the coefficient increment and coefficient change to obtain a preliminary judgment result.

[0282] If the preliminary judgment result shows non-compliance, recalculate and adjust the coefficient increment to obtain the updated coefficient distribution.

[0283] According to the updated coefficient distribution, optimize the coefficient change by column calculation to obtain the optimized coefficient data.

[0284] Extract the complete array structure from the optimized coefficient data through a data acquisition tool to determine the final coefficient array.

[0285] For the final coefficient array, verify the coefficient distribution using a linear regression algorithm to obtain the verified array data.

[0286] According to the verified array data, obtain the final distribution of the sub-pixel positions to determine the output result of the complete coefficient array.

[0287] Furthermore, the process of obtaining the initially collected sub-pixel set in the step S3 includes:

[0288] Obtain two viewpoint image data, extract sub-pixel samples from the corresponding viewpoints according to the integer part in the complete coefficient array, and obtain the initially collected sub-pixel set;

[0289] Based on the initially collected sub-pixel set, determine the sampling positions according to the integer part in the coefficient array, and obtain the adjusted sampling position set;

[0290] Adopt a collection method to extract sub-pixel samples from the corresponding viewpoints, update the adjusted sampling position set, and obtain the optimized sub-pixel sample set;

[0291] For the optimized sub-pixel sample set, use a data acquisition tool to determine whether the sub-pixel samples meet the integrity requirements, and obtain the integrity judgment result;

[0292] If the integrity judgment result is not satisfied, re-adjust the sampling positions through the extraction process to obtain the re-optimized sub-pixel set;

[0293] According to the re-optimized sub-pixel set, use the support vector machine algorithm to classify the sub-pixel samples, and obtain the classified sub-pixel set;

[0294] Through the classified sub-pixel set, obtain the spatial distribution characteristics of the sampling positions, and obtain the feature-enhanced sub-pixel set;

[0295] For the feature-enhanced sub-pixel set, use a clustering algorithm to group the sub-pixel samples of the image data, and obtain the grouped sub-pixel set;

[0296] Through the grouped sub-pixel set, judge the matching degree between the corresponding viewpoints and the sampling positions, and determine the final sub-pixel set.

[0297] Furthermore, the process of obtaining the weighted sub-pixel values in step S4 includes:

[0298] For the initially collected sub-pixel set, use the decimal part in the coefficient array as a basis, and perform weight assignment on the sub-pixel samples of the two viewpoints through linear interpolation to obtain the weighted sub-pixel values;

[0299] Through the weighted sub-pixel values, obtain the corresponding spatial position information from the two viewpoint samples, and determine the sub-pixel data after position adjustment.

[0300] Furthermore, step S4 also includes:

[0301] For the sub-pixel data after position adjustment, use a clustering algorithm to group the samples, and obtain the grouped sub-pixel set;

[0302] Based on the grouped sub-pixel sets, use a data processing tool to judge the distribution consistency among samples, and obtain the sub-pixel data after consistency adjustment;

[0303] The process of obtaining the sub-pixel data after consistency adjustment includes:

[0304] Based on the grouped sub-pixel sets, use a data processing tool to analyze the distribution characteristics among samples, and obtain the distribution consistency evaluation result;

[0305] Based on the distribution consistency evaluation result, use a statistical method to calculate the deviation value among samples, and determine the consistency adjustment parameter;

[0306] Based on the consistency adjustment parameter, use a linear transformation method to correct the sub-pixel data, and obtain the sub-pixel data after preliminary adjustment;

[0307] For the sub-pixel data after preliminary adjustment, use a clustering algorithm to re-group the samples, and obtain the grouping consistency feature;

[0308] Based on the grouping consistency feature, judge the matching degree between the sample distribution and the viewpoint samples, and obtain the matching degree quantization value;

[0309] If the matching degree quantization value is lower than the preset threshold, re-adjust the sub-pixel weights through interpolation calculation to obtain the optimized sub-pixel values;

[0310] Based on the optimized sub-pixel values, use a support vector machine algorithm to classify the samples, and determine the grouped sub-pixel sets after classification;

[0311] Based on the grouped sub-pixel sets after classification, obtain the spatial distribution feature of the samples, and obtain the sub-pixel data with enhanced features;

[0312] For the sub-pixel data with enhanced features, use a data processing tool to judge the distribution consistency among samples, and obtain the sub-pixel data after consistency adjustment.

[0313] Furthermore, the process of obtaining the fused sub-pixel result in step S5 includes:

[0314] Based on the weighted sub-pixel values, use the weighted summation formula to perform fusion calculation on the sub-pixel samples of the two viewpoints, and obtain the fused sub-pixel result;

[0315] Step S5 further includes optimizing the fused sub-pixel result, including:

[0316] Based on the fusion result, obtain the spatial sample distribution features of the two viewpoints, determine the sub-pixel result after distribution adjustment, and group the spatial samples through a clustering algorithm to obtain the grouped sample sets;

[0317] Using the grouped sample set, obtain the changing trend of pixel weights, and judge the fused result after adjusting the changing trend;

[0318] If the matching degree between the fused result after adjusting the changing trend and the viewpoint sample is lower than the preset threshold, re-adjust the pixel weights through interpolation calculation to obtain the optimized sub-pixel result;

[0319] According to the optimized sub-pixel result, use the support vector machine algorithm to classify the sample fusion and determine the classified spatial sample set;

[0320] Through the classified spatial sample set, obtain the behavior distribution characteristics of the fusion calculation to get the sub-pixel result with enhanced characteristics;

[0321] Using the sub-pixel result with enhanced characteristics, perform fusion calculation on the viewpoint samples through the weighted summation formula to determine the fused result after trend optimization;

[0322] According to the fused result after trend optimization, obtain the final distribution characteristics of the spatial samples to get the finally adjusted sub-pixel set;

[0323] Using the finally adjusted sub-pixel set, judge the spatial distribution trend of the pixel weights through distribution characteristic analysis to obtain the optimized sub-pixel fusion result.

[0324] Furthermore, it also includes:

[0325] S6. Judge whether the fused sub-pixel result exceeds the preset threshold range. If the fused sub-pixel result exceeds the preset threshold range, re-perform weight allocation and fusion calculation by adjusting the weight ratio in the coefficient array to obtain the corrected fused sub-pixel data;

[0326] S7. According to the corrected fused sub-pixel data, combine the resolution adjustment parameters to perform pixel recombination on the image to obtain the final naked-eye 3D display image data.

[0327] As Figure 3 shown, furthermore, the method for fusing two-viewpoint images of the lenticular grating for 3D display further includes:

[0328] When the viewpoint number control parameter changes, obtain new viewpoint image data, and repeat the coefficient array optimization and sub-pixel acquisition and fusion process to obtain 3D display image data adapted to the new viewpoint;

[0329] Specifically, when the viewpoint number and the control parameter are associated and changed, obtain the image data corresponding to the new viewpoint, collect the original data through the sensor and store it as the first image data;

[0330] Extract sub-pixel information from the first image data, and use the acquisition process to locate and segment the sub-pixels to obtain the second image data;

[0331] For the association between the second image data and the coefficient array, perform an optimization process, and adjust the coefficient array through a convolutional neural network to obtain the third image data;

[0332] If the third image data is related to the fusion process, integrate the sub-pixel information through the fusion process, and use the weighted average algorithm to generate the fourth image data;

[0333] According to the relationship between the fourth image data and the display data, adjust the 3D display parameters to obtain the fifth image data;

[0334] By matching the fifth image data with the new viewpoint, determine whether the display requirements are met. If not, repeat the optimization process to adjust the coefficient array to obtain the final display data;

[0335] For the association between the final display data and the 3D display, output the 3D display image data adapted to the new viewpoint.

[0336] By recording the time-consuming data of each coefficient array optimization and fusion calculation, use the real-time performance optimization algorithm to dynamically adjust the calculation process to obtain the optimized processing sequence;

[0337] Specifically, by recording the time-consuming data of the fusion calculation, obtain the time-consuming information of each operation to obtain the first processing data;

[0338] According to the association between the first processing data and the real-time performance, use the support vector machine algorithm to analyze the time-consuming distribution to obtain the second processing data;

[0339] If the second processing data exceeds the preset threshold, perform dynamic adjustment on the calculation process to obtain the third processing data;

[0340] Through the correspondence between the third processing data and the coefficient array, adjust the array parameters to obtain the fourth processing data;

[0341] Use the relationship between the fourth processing data and the processing sequence to generate the optimized sequence structure to obtain the fifth processing data;

[0342] According to the matching between the fifth processing data and the performance optimization, determine whether the real-time requirements are met to obtain the sixth processing data;

[0343] Through the association between the sixth processing data and the sequence generation, output the final processing sequence data.

[0344] The specific embodiments of the invention have been described in detail above, but they are only examples. The present invention is not limited to the specific embodiments described above. Those skilled in the art should understand that the above embodiments and the descriptions in the specification only illustrate the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for fusion of two-viewpoint images of cylindrical lens grating for 3D display, characterized in that: include: S1, according to the tilt angle parameter, the sub-pixel width of the grating horizontal coverage, the number of viewpoints control and the resolution adjustment, the initial coefficients corresponding to the R, G, and B sub-pixels in the first row and the first column are calculated from the preset formula to obtain the initial values ​​of the coefficients representing the weights of different viewpoints; S2. Starting from the initial coefficients, recursive calculations are performed for the remaining sub-pixel positions using a row recursive formula and a column recursive formula to obtain complete coefficient array data; S3, acquiring two viewpoint image data, extracting sub-pixel samples from corresponding viewpoints according to the integer part in the complete coefficient array, and obtaining a preliminarily collected sub-pixel set; S4, for the initially collected sub-pixel set, using the decimal part in the coefficient array as a basis, weighting the sub-pixel samples of the two viewpoints by a linear interpolation method to obtain a weighted sub-pixel value; S5, performing fusion calculation on the sub-pixel samples of the two viewpoints by using the weighted sub-pixel values ​​to obtain fused sub-pixel results, and combining the fused sub-pixel results into a complete image to generate a final 3D display image; Also includes: S6, determining whether the fused sub-pixel result exceeds a preset threshold range. If the fused sub-pixel result exceeds the preset threshold range, re-performing weight distribution and fusion calculation by adjusting the weight ratio in the coefficient array to obtain corrected fused sub-pixel data; S7, performing pixel reorganization on the image according to the corrected fused sub-pixel data and in combination with the resolution adjustment parameter to obtain the final naked-eye 3D display image data; When the viewpoint number control parameter changes, new viewpoint image data is acquired, and the coefficient array optimization and sub-pixel acquisition and fusion processes are repeated to obtain 3D display image data adapted to the new viewpoint; By recording the time-consuming data of each coefficient array optimization and fusion calculation, the real-time performance optimization algorithm is used to dynamically adjust the calculation process to obtain an optimized processing sequence.

2. The method for fusion of two-viewpoint images of cylindrical lens grating for 3D display according to claim 1, characterized in that: The process of obtaining the initial values ​​of coefficients representing weights of different viewpoints in step S1 includes: The distribution data of sub-pixel width is obtained through the tilt angle parameter and the grating level to obtain the preliminary spatial mapping relationship; Calculate the density of viewpoint distribution according to the sub-pixel width and the number of viewpoints, and determine the preliminary distribution of viewpoint weights; The initial coefficients of the R sub-pixel, the G sub-pixel, and the B sub-pixel are calculated using a preset formula to obtain the initial value of the coefficient of each sub-pixel; Correct the initial value of the coefficient by adjusting the resolution, and determine whether the corrected coefficient meets the viewpoint weight requirement; If the corrected coefficient deviates from the viewpoint weight, the distribution of the sub-pixel width is readjusted at the raster level to obtain the updated initial value of the coefficient; According to the updated initial value of the coefficient and the number of viewpoints, the linear regression algorithm is used to optimize the distribution of viewpoint weights to obtain the final weight coefficient.

3. The method for fusion of two-viewpoint images of cylindrical lens grating for 3D display according to claim 2, characterized in that: The step S1 further comprises: S101, according to the sub-pixel width and the number of viewpoints, a linear regression algorithm is used to optimize the distribution of viewpoint weights, and the display effects of the R sub-pixel, the G sub-pixel, and the B sub-pixel are verified by the final weight coefficients; The process of verifying the display effect in step S101 includes: calculating preliminary weight coefficients by using a linear regression algorithm through sub-pixel width and the number of viewpoints to obtain optimized allocation data; According to the optimized allocation data, the display data of the R sub-pixel is adjusted to obtain the adjusted first sub-pixel data; Matching the display data of the G sub-pixel with the adjusted first sub-pixel data to obtain the second sub-pixel data; According to the second sub-pixel data, synchronously process the display data of the B sub-pixel to obtain the third sub-pixel data; Using a mean value calculation method, the first sub-pixel data, the second sub-pixel data and the third sub-pixel data are merged to obtain comprehensive display data; By comprehensively displaying the matching relationship between data and the number of viewpoints, the integrity of data processing can be judged and verification results can be obtained; According to the verification result, if the integrity is insufficient, the weight coefficient is adjusted through the distribution of sub-pixel width to obtain updated comprehensive display data.

4. The method for fusion of two-viewpoint images of cylindrical lens grating for 3D display according to claim 1, characterized in that: The process of obtaining the complete coefficient array data in step S2 includes: By using the row recursion formula and the column recursion formula, starting from the initial coefficient, the coefficient increment and coefficient change are calculated for the sub-pixel position to obtain the preliminary coefficient distribution; Using the position distribution to spatially map the preliminary coefficient distribution, and obtaining the adjusted coefficient distribution data; Through the coefficient increment and coefficient change, it is determined whether the adjusted coefficient distribution meets the requirements of the complete array. If not, the coefficient increment is readjusted through calculation to obtain the updated coefficient distribution; According to the updated coefficient distribution, the column calculation is used to optimize the coefficient change and obtain the optimized coefficient data; Extract the complete array structure from the optimized coefficient data through the data acquisition tool to determine the final coefficient array; For the final coefficient array, a linear regression algorithm is used to verify the coefficient distribution to obtain the verified array data; The final distribution of sub-pixel positions is obtained through the verified array data, and the output result of the complete coefficient array is determined.

5. The method for fusion of two-viewpoint images of cylindrical lens grating for 3D display according to claim 4, characterized in that: The step S2 further comprises: S201, judging whether the adjusted coefficient distribution meets the complete array requirement according to the coefficient increment and the coefficient change, and if not, re-adjusting the coefficient increment by calculation to obtain an updated coefficient distribution; Specifically, according to the complete coefficient array data, the coefficient data is spatially mapped using position distribution to obtain adjusted coefficient distribution data; Through the coefficient increment and coefficient change, it is judged whether the adjusted coefficient distribution meets the requirements of the complete array, and a preliminary judgment result is obtained; If the preliminary judgment result shows that it is not satisfied, the coefficient increment is readjusted by calculation to obtain the updated coefficient distribution; According to the updated coefficient distribution, the column calculation is used to optimize the coefficient change and obtain the optimized coefficient data; Extract the complete array structure from the optimized coefficient data through the data acquisition tool to determine the final coefficient array; For the final coefficient array, a linear regression algorithm is used to verify the coefficient distribution to obtain the verified array data; Based on the verified array data, the final distribution of sub-pixel positions is obtained to determine the output result of the complete coefficient array.

6. The method for fusion of two-viewpoint images of cylindrical lens grating for 3D display according to claim 1, characterized in that: The process of obtaining the initially collected sub-pixel set in step S3 includes: Obtaining two viewpoint image data, extracting sub-pixel samples from corresponding viewpoints according to the integer part in the complete coefficient array, and obtaining a preliminarily collected sub-pixel set; Determine the sampling position according to the integer part in the coefficient array through the initially collected sub-pixel set, and obtain an adjusted sampling position set; Adopting the acquisition method, extracting sub-pixel samples from the corresponding viewpoints, updating the adjusted sampling position set, and obtaining an optimized sub-pixel sample set; For the optimized sub-pixel sample set, using a data acquisition tool, determine whether the sub-pixel sample meets the integrity requirement, and obtain an integrity determination result; If the integrity judgment result is not satisfied, the sampling position is readjusted through the extraction process to obtain a re-optimized sub-pixel set; According to the re-optimized sub-pixel set, a support vector machine algorithm is used to classify the sub-pixel samples to obtain a classified sub-pixel set; The spatial distribution characteristics of the sampling position are obtained through the classified sub-pixel set, and the feature-enhanced sub-pixel set is obtained; For the feature-enhanced sub-pixel set, a clustering algorithm is used to group the sub-pixel samples of the image data to obtain a grouped sub-pixel set; The matching degree between the corresponding viewpoint and the sampling position is determined by the grouped sub-pixel set to determine the final sub-pixel set.

7. The method for fusion of two-viewpoint images of cylindrical lens grating for 3D display according to claim 1, characterized in that: The process of obtaining the weighted sub-pixel value in step S4 includes: For the initially collected sub-pixel set, the decimal part in the coefficient array is used as a basis, and the sub-pixel samples of the two viewpoints are weighted by linear interpolation method to obtain the weighted sub-pixel value; The corresponding spatial position information is obtained from the two viewpoint samples through the weighted sub-pixel values ​​to determine the sub-pixel data after position adjustment.

8. The method for fusion of two-viewpoint images of cylindrical lens grating for 3D display according to claim 7, characterized in that: The step S4 further comprises: For the sub-pixel data after position adjustment, a clustering algorithm is used to group the samples to obtain a grouped sub-pixel set; According to the grouped sub-pixel set, the distribution consistency between samples is determined by a data processing tool to obtain the sub-pixel data after consistency adjustment; The process of obtaining the sub-pixel data after consistency adjustment includes: According to the grouped sub-pixel set, data processing tools are used to analyze the distribution characteristics between samples to obtain the distribution consistency evaluation results; Based on the distribution consistency evaluation results, the deviation values ​​between samples are calculated using statistical methods to determine the consistency adjustment parameters; According to the consistency adjustment parameter, the sub-pixel data is corrected by a linear transformation method to obtain the preliminarily adjusted sub-pixel data; For the sub-pixel data after preliminary adjustment, a clustering algorithm is used to regroup the samples to obtain group consistency features; By grouping consistency features, the matching degree between sample distribution and viewpoint samples is judged, and the matching quantization value is obtained; If the matching quantization value is lower than the preset threshold, the sub-pixel weight is readjusted through interpolation calculation to obtain the optimized sub-pixel value; According to the optimized sub-pixel values, the samples are classified using a support vector machine algorithm to determine a classified sub-pixel set; The spatial distribution characteristics of the samples are obtained through the classified sub-pixel set, and the feature-enhanced sub-pixel data is obtained; For the feature-enhanced sub-pixel data, the distribution consistency between samples is determined by a data processing tool to obtain the sub-pixel data after consistency adjustment.

9. The method for fusion of two-viewpoint images of cylindrical lens grating for 3D display according to claim 1, characterized in that: The process of obtaining the fused sub-pixel result in step S5 includes: The weighted sub-pixel values ​​are used to fuse the sub-pixel samples of the two viewpoints using a weighted summation formula to obtain a fused sub-pixel result; The step S5 further includes optimizing the fusion sub-pixel result, including: According to the fusion results, the spatial sample distribution characteristics of the two viewpoints are obtained, the sub-pixel results after the distribution adjustment are determined, and the spatial samples are grouped by a clustering algorithm to obtain a grouped sample set; The grouped sample set is used to obtain the changing trend of pixel weights and determine the fusion result after the changing trend is adjusted; If the matching degree between the fusion result after the change trend adjustment and the viewpoint sample is lower than the preset threshold, the pixel weight is readjusted through interpolation calculation to obtain the optimized sub-pixel result; According to the optimized sub-pixel results, the support vector machine algorithm is used to classify the sample fusion and determine the classified spatial sample set; Through the classified spatial sample set, the behavior distribution characteristics of the fusion calculation are obtained, and the sub-pixel result of the characteristic enhancement is obtained; Using the sub-pixel results of feature enhancement, the viewpoint samples are fused and calculated through the weighted summation formula to determine the fusion result after trend optimization; According to the trend-optimized fusion results, the final distribution characteristics of the spatial samples are obtained to obtain the final adjusted sub-pixel set; The final adjusted sub-pixel set is used to determine the spatial distribution trend of pixel weights through distribution characteristic analysis to obtain the optimized sub-pixel fusion result.

Images