Digital printing image registration method fusing feature registration and deformation optimization

Abstract

The invention provides a digital printing image registration method fusing feature registration and deformation optimization, and relates to the field of digital printing, and the method comprises the steps: employing a feature extraction algorithm to achieve global registration; local deformation optimization is carried out in combination with a deformation estimation algorithm; a reduction factor is adaptively introduced in the global registration stage to improve the registration efficiency; in the local deformation optimization stage, superpixel segmentation is performed on the pre-printing original image, and a nonlinear brightness compensation mechanism is introduced according to difference information of corresponding superpixel areas of the printing image and the pre-printing original image, so that the constraint of brightness on a deformation estimation algorithm is overcome. According to the invention, various rigid deformations, elastic deformations and brightness differences between the printing image and the original image before printing which may occur in the printing process and the process of collecting the printing image can be dealt with, so that the registration precision of the digital printing image is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of image registration, and relates to a high-precision registration method for digital printing images, and in particular to a digital printing image registration method that integrates feature registration and deformation optimization. Background Art

[0002] Image registration is the process of matching two or more images acquired at different times, with different sensors or under different conditions. It has been widely used in remote sensing data analysis, computer vision, image processing and other fields, and is also a hot research field for correcting image transformation. According to the deformation mode of the image, there are two types of image registration methods: rigid registration and non-rigid registration. Image registration technology is mainly divided into two categories: registration based on spatial features and registration based on grayscale information. Image registration methods based on spatial features are widely used. However, when the number of feature points extracted from the image is large, although this helps to improve the accuracy of registration, it will also lead to a significant decrease in the registration speed; on the contrary, if the number of feature points extracted is reduced in order to increase the speed, the accuracy of registration will inevitably be sacrificed. Image registration methods based on image grayscale information require the analysis of the grayscale information of the entire image, which is computationally intensive and the registration algorithm is inefficient. In recent years, registration algorithms based on feature extraction have been a research hotspot in the field of image registration and are also commonly used image registration algorithms. The displacement field solved in the registration is similar to the velocity field solved in the optical flow. In recent years, the optical flow computing technology has developed rapidly. The image registration method based on the optical flow model has been able to obtain high-quality registration results. The optical flow field belongs to the registration method based on grayscale information, which analyzes the grayscale of the image.

[0003] In the field of digital printing, the registration accuracy of printed images and reference images is crucial for the detection of surface defects of printed products. At present, in the field of printed image registration, the mainstream registration methods are registration methods based on feature extraction such as SIFT and SURF. However, due to factors such as equipment vibration and paper instability during the printing process, the collected printed images have slight elastic deformation and other rigid deformations, and feature extraction registration algorithms such as SIFT are difficult to solve elastic deformation. Although the Active Demons deformation estimation algorithm based on optical flow field can solve elastic deformation, the algorithm is sensitive to brightness. When detecting some special defects, such as the interruption of the printing process caused by the problem of the printing nozzle, which causes the detection of subtle printing stripe defects, the pre-press original image is usually used as a reference image. The collected printed image undergoes two layers of complex mapping of the printing equipment and the image acquisition device, so its brightness must be different from the pre-press original image. Therefore, the use of deformation estimation algorithms to register the printed image with the pre-press original image is bound to affect the registration accuracy.

[0004] In order to improve the registration accuracy of digital printed images as much as possible and correct the elastic deformation of printed images caused by equipment vibration, paper instability and other reasons, the present invention proposes a digital printed image registration method that integrates feature registration and deformation optimization. Summary of the invention

[0005] In view of the problems existing in the prior art, the present invention provides a digital printing image registration method integrating feature registration and deformation optimization.

[0006] The technical solution of the present invention is as follows:

[0007] A digital printing image registration method integrating feature registration and deformation optimization comprises: firstly, using a feature extraction algorithm to realize global registration between a pre-press original image and a printed image, then combining a deformation estimation algorithm to perform local deformation optimization, thereby improving registration accuracy and solving elastic deformation; and adaptively introducing a reduction factor in the global registration stage to improve registration efficiency; and in the local deformation optimization stage, introducing a nonlinear brightness compensation mechanism to overcome the constraint of brightness difference on the deformation estimation algorithm.

[0008] The specific steps include:

[0009] The first step is to adaptively introduce a reduction factor of 1 / n, and obtain the corresponding reduced affine matrix according to the reduction factor. and the corresponding magnified affine matrix

[0010] In the second step, after resizing the prepress original image O and the printed image P by reducing the affine matrix S, the global registration of the printed image P and the prepress original image O is completed by the feature extraction algorithm;

[0011] In the third step, the global registration result described in the second step is locally deformed and optimized through a deformation estimation algorithm, while a nonlinear brightness compensation mechanism is introduced.

[0012] Furthermore, the specific steps of adaptively introducing the reduction factor in the first step are as follows:

[0013] 1.1) Select the smaller value of the length and width of the pre-press original image, divide 240 by this value, and the result is the initial reduction factor;

[0014] 1.2) introducing the reduction factor, resizing the printed image and the prepress original image, and globally registering the resized printed image with the prepress original image;

[0015] 1.3) Calculate the correlation coefficient CC value between the global registration result and the prepress original image, and set a threshold (usually 0.9). If the CC value exceeds the set threshold, output the global registration result; if it does not exceed the set threshold, double the reduction factor, and then repeat steps 1.2) and 1.3) until the CC value exceeds the set threshold, and output a global registration result that meets the requirements.

[0016] Specifically, the correlation coefficient The value of CC ranges from -1 to 1, where 1 represents a perfect match, 0 represents no correlation, and -1 represents a complete negative correlation; where N is the total number of pixels, O i is the pixel value of the i-th pixel in the original image before printing, is the average value of all pixel intensity values ​​of the original image before printing, G i is the intensity value corresponding to the i-th pixel in the registration image, is the average intensity value of all pixels in the registered image.

[0017] Furthermore, the specific steps of the second step are as follows:

[0018] 2.1) Resize the prepress original image O and the printed image P by using the reduced affine matrix S, and at the same time, update the printed image spatial position matrix to A ′ =SA, A is the spatial position matrix of the printed image before resizing;

[0019] 2.2) Extract feature points of the resized pre-press original image and the printed image through a feature extraction algorithm, locate the feature points, determine the direction of the feature points, describe the feature points, and perform feature matching;

[0020] 2.3) Obtain the affine transformation coefficients of the resized pre-press original image and the printed image through random sample consensus (RANSAC), and obtain the application transformation matrix H;

[0021] 2.4) Through matrix calculation, the spatial position matrix T = UHA′ of the printed image after global registration is obtained. According to A in 2.1) ′ , we can get T = UHSA, so we can get the final application transformation matrix Finally, the printed image P is affine transformed through M to obtain the global registration result G;

[0022] Furthermore, the deformation estimation algorithm in the third step is the Active Demons algorithm, which specifically refers to:

[0023] The global registration result G in the second step is regarded as a frame of the prepress original image O after slight movement and deformation. The brightness value of the point (x, y) on the image O at time t is O(x, y, t). According to the motion process: O(x, y, t) = O(x + Δx, y + Δy, t + Δt), the Taylor expansion on the right is:

[0024] G is considered as a frame of O after slight movement and deformation, dx, dy and dt are very small, and ε after Taylor expansion can be ignored: remember, but,

[0025] Specifically, the deformation estimation algorithm can be based on the gradient information of the pre-press original image O. The difference information O between the prepress original image O and the global registration result G t To calculate the offset between the original image before printing and the printed image The O t The difference between O and G: O t =diff=GO, in the present invention, a nonlinear brightness compensation mechanism is introduced to eliminate the error in deformation estimation caused by the nonlinear brightness difference between the pre-press original image O and the global registration result G.

[0026] Furthermore, after the nonlinear brightness compensation mechanism is introduced, the third step is specifically as follows:

[0027] 3.1) The pixels in the prepress original image O are combined into larger areas with similar properties through the superpixel segmentation SLIC algorithm, and the brightness of the areas with similar properties is compensated. Specifically, according to the difference information diff between the prepress original image O and the global registration result G, the brightness of the prepress original image is compensated, O_c(x,y)=O(x,y)+δ*diff avg (x,y), where the diff avg (x, y) is the average pixel value of the superpixel region where the point (x, y) is located on diff, δ*diff avg (x, y) is the brightness compensation value of point (x, y), δ is the compensation coefficient, and the value range is 0.8~1.2;

[0028] 3.2) Obtain the frame difference O_c between the original image O_c before printing after brightness compensation and the global registration result G t =diff c =G-O_c=diff-δ*diff avg ; Prepress original image gradient information after brightness compensation

[0029] 3.3) According to the deformation estimation algorithm, O_c(x,y,t)=O_c(x+Δx,y+Δy,t+Δt), after Taylor expansion, Thus,

[0030] 3.4) Add the gradient of the global registration result to the calculation of the offset to obtain the offset of each pixel (U x , U y ),have:

[0031]

[0032] It is stated is the result of calculation after adding brightness compensation, which is (δ*diff avg –diff) at the point (x,y), O_c x and O_c y are the gradients in the x and y directions at the point (x, y) on the prepress original image after adding brightness compensation, G x and G y are the gradients in the x and y directions at the point (x, y) on the global registration map, respectively. α is a normalization factor used to control the size of the coordinate offset. The larger α is, the smaller the offset is, and vice versa.

[0033] Further, according to the offset (U x , U y ) Perform local deformation optimization and adjustment on the global registration result to obtain the final registration result map.

[0034] The present invention provides a digital printing image registration method integrating feature registration and deformation optimization, which can improve the registration accuracy of digital printing images to cope with various rigid deformations, elastic deformations and brightness differences between the printed images and the pre-press original images that may occur during the printing process and the process of collecting the printed images.

[0035] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the digital printing image registration method of integrating feature registration and deformation optimization as described in any one of the above items.

[0036] An electronic device, comprising:

[0037] one or more processors;

[0038] A memory for storing one or more programs;

[0039] When the one or more programs are executed by the one or more processors, the one or more processors implement the digital printing image registration method integrating feature registration and deformation optimization as described in any one of the above items.

[0040] The beneficial effects of the present invention are as follows:

[0041] 1) In the local deformation optimization stage, the prepress original image is segmented into superpixels. According to the difference information of the corresponding superpixel area between the printed image and the prepress original image, a nonlinear brightness compensation mechanism is introduced to overcome the constraint of brightness on the deformation estimation algorithm;

[0042] 2) Adaptively introduce a reduction factor in the global registration stage to significantly improve the registration efficiency;

[0043] 3) The global registration stage corrects the rigid deformation of the printed image;

[0044] 4) The local deformation optimization stage corrects the elastic deformation of the printed image. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a schematic diagram of the process of the present invention;

[0046] Figure 2 This is the original pre-press image of Lena;

[0047] Figure 3 Print images for lena;

[0048] Figure 4 Schematic diagram of the process of introducing reduction factor for adaptation;

[0049] Figure 5 The relationship between the reduction factor and the number of feature points and the global registration time is shown;

[0050] Figure 6 This is the feature extraction and matching result diagram of the lena pre-press original image and the printed image;

[0051] Figure 7 It is the global registration result image and its local magnified image;

[0052] Figure 8 This is the SCIL superpixel segmentation map of the lena pre-press original image;

[0053] Fig. 9 This is the effect diagram of the deformation estimation algorithm without introducing brightness compensation;

[0054] Fig.10 This is the effect diagram of the deformation estimation algorithm that introduces brightness compensation. DETAILED DESCRIPTION

[0055] The following is a detailed description of an embodiment of the present invention. This embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation method and a specific operation process are given, but the protection scope of the present invention is not limited to the following embodiment.

[0056] This embodiment will use the lena prepress original image with an image resolution of 7795×3840. Figure 2 As shown, the image resolution is 8088×4127 lena printing image, such as Figure 3 As shown. Figure 2 and Figure 3 It can be seen that the printed image has slight elastic deformation and rigid deformation such as scaling, rotation, and translation compared to the pre-press original image, and there is also a certain brightness difference. This embodiment uses the classic correlation coefficient (CC), mean square error (MSE), peak signal-to-noise ratio (PSNR), mutual information (MI), and structural similarity (SSIM) to evaluate the registration effect of the algorithm; at the same time, the obtained registration result map can verify the performance of the registration algorithm. Among these five evaluation indicators, the larger the PSNR and MI values, the better the image registration quality; the closer the SSIM and CC values ​​are to 1, the more similar the images are; MSE is used to detect the stability of the algorithm. The smaller the MSE value, the more stable the proposed algorithm is.

[0057] like Figure 1 As shown, in this embodiment, the digital printing image registration method integrating feature registration and deformation optimization specifically includes the following steps:

[0058] First, the reduction factor is adaptively introduced to achieve global registration of the lena pre-press original image O and the lena printed image P. The specific steps are as follows:

[0059] 1) Select the smaller value of the length and width of the prepress original image, 3840, and divide 240 by this value. The result is the initial reduction factor: 1 / 16;

[0060] 2) Obtain the corresponding reduced affine matrix according to the reduction factor and the corresponding magnified affine matrix The printed image and the pre-press original image are resized according to the reduced affine matrix S. At the same time, the spatial position matrix of the printed image P is denoted as A, and after resizing, it is updated to A ′ =SA;

[0061] 3) Globally aligning the resized lena printed image P with the lena pre-press original image O: In this embodiment, the feature extraction algorithm SIFT algorithm (or other feature extraction algorithms, such as SUFR algorithm) is used to extract feature points of the resized lena pre-press original image O and the lena printed image P, locate the feature points, determine the direction of the feature points, describe the feature points, and perform feature matching;

[0062] 4) According to the feature matching results, the affine transformation coefficients of the resized lena pre-press original image O and the lena printed image P are obtained through random sample consensus (RANSAC), thereby obtaining the application transformation matrix H;

[0063] 5) Through matrix calculation, the spatial position matrix of the lena printed image after global registration is obtained, T = UHA'. According to A' described in 2), T = UHSA can be obtained. Therefore, the final application transformation matrix can be obtained. Finally, the lena printed image P is affine transformed by M to obtain the global registration result G;

[0064] 6) Calculate the correlation coefficient between the global registration result G and the lena prepress original image O The value of CC ranges from -1 to 1, where 1 represents a perfect match, 0 represents no correlation, and -1 represents a complete negative correlation; where N is the total number of pixels, O i is the pixel value of the i-th pixel in the original image before printing, is the average value of all pixel intensity values ​​of the original image before printing, G i is the intensity value corresponding to the i-th pixel in the registration image, is the average value of the intensity values ​​of all pixels in the registered image;

[0065] 7) Set the threshold to 0.9. If the CC value exceeds the set threshold, output the global registration result. If it does not exceed the set threshold, double the reduction factor and repeat steps 2) to 7) until a global registration result G that meets the requirements is output.

[0066] The flowchart of adaptively introducing the reduction factor is as follows Figure 4 As shown in the figure, the relationship between different reduction factors and the number of feature point extraction and global registration time is as follows Figure 5 As shown in the figure, the smaller the reduction factor is, the fewer the number of feature points detected is, and the global registration time is significantly reduced; in this embodiment, after the initial reduction factor of 1 / 16 is adaptively introduced, the feature extraction and matching results of the lena pre-press original image O and the lena printed image P are as follows: Figure 6As shown in the figure, the correlation coefficient CC value between the global registration result G and the lena prepress original image O is 0.97, which is greater than the set threshold of 0.9. Therefore, the global registration result G is output. The global registration result G and the local enlarged image are shown in Figure 7 As shown, rigid deformations such as scaling, translation and rotation have been successfully resolved; however, it can be observed from the local magnified image that the elastic deformation has not been corrected.

[0067] Next, the deformation estimation algorithm is used to perform local deformation optimization on the output global registration result G, and a nonlinear brightness compensation mechanism is introduced. The specific steps are as follows:

[0068] The global registration result G is regarded as a frame of the prepress original image O after slight movement and deformation, and assuming that the brightness of G is consistent with that of O, then the brightness value of the point (x, y) on the prepress original image O of lena at time t is O(x, y, t);

[0069] According to the motion process: O(x,y,t)=O(x+Δx,y+Δy,t+Δt);

[0070] Taylor expansion on the right:

[0071] Since G is regarded as a frame of O after slight movement and deformation, dx, dy and dt are very small, and the ε after Taylor expansion can be ignored:

[0072] remember, but,

[0073] Therefore, the deformation estimation algorithm can use the gradient information of the prepress original image O The difference information O between the prepress original image O and the global registration result G t To calculate the offset between the original image before printing and the printed image The O t The difference between O and G: O t =diff=GO.

[0074] The above derivation process is to regard G as a frame of O after slight movement and deformation, and the brightness is consistent. G is the global registration result, which is slightly deformed relative to O, but there is a nonlinear brightness difference between O and G, such as Figure 2 , Figure 3 As shown, it will cause errors in the deformation estimation algorithm. Therefore, a nonlinear brightness compensation mechanism is introduced into the deformation estimation algorithm to solve this problem. The specific steps are as follows:

[0075] 1) Through the SLIC algorithm: initialization of cluster centers, search area restriction, pixel similarity measurement, iterative clustering, post-processing, the pixels in the prepress original image are combined into larger areas with similar properties, and the SCIL superpixel segmentation of the prepress original image O is as follows: Figure 8 As shown, brightness compensation is performed on regions with similar properties respectively. Specifically, brightness compensation is performed on the prepress original image according to the difference information diff between the prepress original image O and the global registration result G:

[0076] O_c(x,y)=O(x,y)+δ*diff avg (x,y)

[0077] The diff avg (x, y) is the average pixel value of the superpixel region where the point (x, y) is located on diff, δ*diff avg (x, y) is the brightness compensation value of the point (x, y), δ is the compensation coefficient, and the value range is 0.8 to 1.2. In this embodiment, δ is 1.1;

[0078] 2) Obtain the frame difference between the original pre-printed image O_c after brightness compensation and the global registration result G: O_c t =diff c =G-O_c=diff-δ*diff avg ; Gradient information of the prepress original image after brightness compensation:

[0079] 3) According to the deformation estimation algorithm principle, O_c(x,y,t)=O_c(x+Δx,y+Δy,t+Δt), after Taylor expansion, Thus,

[0080] 4) Based on the deformation estimation algorithm, a normalization factor α is added to control the size of the coordinate offset. The larger α is, the smaller the offset is, and vice versa. At the same time, the gradient of the global registration result is added to the offset calculation to obtain the offset of each pixel (U x , U y ), It is stated is the result of calculation after adding brightness compensation, which is (δ*diff avg –diff) at the point (x,y), O_c x and O_c y are the gradients in the x and y directions at the point (x, y) on the prepress original image after adding brightness compensation, G x and G yare the gradients in the x and y directions at the point (x, y) on the global registration map. Finally, according to the calculated offset (U x , U y ) Perform local deformation optimization and adjustment on the printed image after global registration.

[0081] like Fig. 9 , Fig.10 As shown in the figure, there is a comparison of the local optimization effect of the deformation estimation algorithm without introducing brightness compensation and the deformation estimation algorithm with introducing brightness compensation on the global registration result; when there is a brightness difference between the prepress original image O and the printed image P, the deformation estimation algorithm without introducing brightness compensation cannot effectively correct the elastic deformation that still exists in the global registration result, but by introducing nonlinear brightness compensation, it successfully overcomes the limitation of brightness difference on the deformation estimation algorithm, realizes the local deformation optimization of the global registration result, and corrects the elastic deformation.

[0082] Table 1 is a comparative analysis of the registration accuracy of the method of the present invention with a single feature extraction algorithm (SIFT algorithm), a deformation estimation algorithm (Active Demons algorithm), and a feature extraction algorithm (SIFT algorithm) + a deformation estimation algorithm without introducing brightness compensation (Active Demons algorithm) for the same instance. It can be seen that the registration accuracy of the algorithm proposed in the present invention is the best in all five evaluation indicators.

[0083] Table 1: Comparative analysis of registration accuracy

[0084]

[0085] In this embodiment, according to the algorithm of the present invention, the final registration result and the local magnified image are obtained, as shown in FIG. Fig.10 As shown, a high registration accuracy is achieved, and both rigid deformation and elastic deformation are well corrected.

[0086] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0087] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0088] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0089] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A digital printing image registration method integrating feature registration and deformation optimization, characterized in that: include: Firstly, the feature extraction algorithm is used to realize the global registration between the pre-press original image and the printed image, and then the deformation estimation algorithm is combined to perform local deformation optimization. Adaptively introduce a reduction factor in the global registration stage; In the local deformation optimization stage, the prepress original image is segmented into superpixels, and a nonlinear brightness compensation mechanism is introduced based on the difference information of the corresponding superpixel areas between the printed image and the prepress original image.

2. The digital printing image registration method integrating feature registration and deformation optimization according to claim 1 is characterized in that: This method specifically The following steps are involved: The first step is to adaptively introduce a reduction factor of 1 / n, and obtain the corresponding reduced affine matrix according to the reduction factor. and the corresponding magnified affine matrix In the second step, after resizing the prepress original image O and the printed image P by reducing the affine matrix S, the global registration of the printed image P and the prepress original image O is completed by the feature extraction algorithm; In the third step, the global registration result obtained in the second step is locally deformed and optimized through the deformation estimation algorithm. A nonlinear brightness compensation mechanism is introduced into the deformation estimation algorithm to overcome the brightness constraint on the deformation estimation algorithm.

3. The digital printing image registration method integrating feature registration and deformation optimization according to claim 2, characterized in that: In the first step, the reduction factor is adaptively introduced, and the specific steps include: 1.1) Select the smaller value of the length and width of the pre-press original image, divide it by 240, and the result is the initial reduction factor; 1.2) Introducing the reduction factor, the printed image to be registered and the pre-press original image are resized and then globally registered; 1.3) Calculate the correlation coefficient CC value between the global registration result and the prepress original image. If the CC value exceeds the set threshold, output the global registration result; if it does not exceed the set threshold, double the reduction factor, and then repeat steps 1.2) and 1.3) until the CC value exceeds the set threshold, and output a global registration result that meets the requirements.

4. The digital printing image registration method integrating feature registration and deformation optimization according to claim 3, characterized in that: The correlation coefficient CC value is specifically: The value of CC ranges from -1 to 1, where 1 represents a perfect match, 0 represents no correlation, and -1 represents a complete negative correlation; where N is the total number of pixels, O i is the pixel value of the i-th pixel in the pre-press original image, is the average value of all pixel intensity values ​​of the original image before printing, G i is the intensity value corresponding to the i-th pixel in the global registration result, It is the average value of all pixel intensity values ​​in the global registration result.

5. The digital printing image registration method integrating feature registration and deformation optimization according to claim 2, characterized in that: The second step specifically includes: 2.1) Using the reduced affine matrix S, resize the pre-press original image O and the printed image P, and at the same time, update the printed image spatial position matrix, where A′=SA, A is the spatial position matrix of the printed image before resizing; 2.2) Extract feature points of the resized pre-press original image and the printed image through a feature extraction algorithm, locate the feature points, determine the direction of the feature points, describe the feature points, and perform feature matching; 2.3) Obtain the affine transformation coefficients of the resized pre-press original image and the printed image through random sample consensus (RANSAC), and obtain the application transformation matrix H; 2.4) Through matrix calculation, the spatial position matrix of the printed image after global registration is obtained, T = UHA′, that is, T = UHSA, thus obtaining the final application transformation matrix Finally, the printed image P is affine transformed through M to obtain the global registration result G.

6. The digital printing image registration method integrating feature registration and deformation optimization according to claim 2, characterized in that: The deformation estimation algorithm adopts the Active Demons algorithm. The global registration result G is regarded as a frame of the prepress original image O after slight movement and deformation. The brightness value of the point (x, y) on the prepress original image O at time t is O(x, y, t), then That is, through the gradient information of the prepress original image O The difference information O between the prepress original image O and the global registration result G t Ability to calculate the dot offset between the prepress original and the printed image The O t The difference between O and G: O t =diff=GO, a nonlinear brightness compensation mechanism is introduced to eliminate the error in deformation estimation caused by the nonlinear brightness difference between the prepress original image O and the global registration result G.

7. The digital printing image registration method integrating feature registration and deformation optimization according to claim 6, characterized in that: The nonlinear brightness compensation mechanism specifically includes: 3.1) The pixels in the prepress original image are combined into larger areas with similar properties through the superpixel segmentation algorithm, and the brightness of the areas with similar properties is compensated. Specifically, according to the difference information diff between the prepress original image O and the global registration result G, the brightness of the prepress original image is compensated, O_c(x,y)=O(x,y)+δ*diff avg (x,y), where the diff avg (x, y) is the average pixel value of the superpixel region where the point (x, y) is located on diff, δ*diff avg (x, y) is the brightness compensation value of point (x, y), δ is the compensation coefficient, and the value range is 0.8~1.2; 3.2) Obtain the frame difference O_c between the original image O_c before printing after brightness compensation and the global registration result G t =diff c =G-O_c=diff-δ*diff avg ; Prepress original image gradient information after brightness compensation 3.3) According to the deformation estimation algorithm, O_c(x,y,t)=O_c(x+Δx,y+Δy,t+Δt), after Taylor expansion, Thus, 3.4) Add the gradient of the global registration result to the calculation of the offset to obtain the offset of each pixel (U x , U y ), in is the result of calculation after adding brightness compensation, which is (δ*diff avg –diff) at the point (x,y), O_c x and O_c y are the gradients in the x and y directions at the point (x, y) on the prepress image after adding brightness compensation, G x and G y are the gradients in the x and y directions at the point (x, y) on the global registration map, and α is a normalization factor used to control the size of the offset.

8. The digital printing image registration method integrating feature registration and deformation optimization according to claim 7, characterized in that: According to the offset (U x , U y ) Perform local deformation optimization adjustment on the global registration result G to obtain the final registration result map.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the digital printing image registration method integrating feature registration and deformation optimization as described in any one of claims 1 to 8 is implemented.

10. An electronic device, characterized in that: The device comprises: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the digital printing image registration method integrating feature registration and deformation optimization as described in any one of claims 1 to 8.

Images