Tensor decomposition based low-quality fingerprint inpainting method, system, medium and device

Abstract

The application belongs to the technical field of fingerprint repair, and provides a low-quality fingerprint repair method, system, medium and equipment based on tensor decomposition. The method comprises the following steps: obtaining a low-quality fingerprint image, and dividing the low-quality fingerprint image into a plurality of non-overlapping image blocks in advance; searching for the most similar image blocks from different directions around the image blocks in parallel, and constructing a similar block group of each image block; performing denoising and enhancement processing on the similar block group of each image block to obtain an enhanced similar block group of each image block; directly stacking the similar blocks in the enhanced similar block group of each image block along a third dimension to form a similar tensor, performing completion on the similar tensor, and then performing image block aggregation; dividing the fingerprint image formed by the image block aggregation into a plurality of non-overlapping image blocks again, re-construction the similar block group of each image block, and sequentially performing denoising and enhancement, completion operation and image block aggregation until the iteration reaches a preset number of times, so that a final repaired fingerprint image is obtained.

Description

Technical Field

[0001] This invention belongs to the field of fingerprint repair technology, and particularly relates to a method, system, medium and device for repairing low-quality fingerprints based on tensor decomposition. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Fingerprint recognition technology has wide applications in criminal investigation, immigration, personal consumption, security, and finance. Generally speaking, collected fingerprints can be divided into two categories: one is the imprint obtained under controlled conditions; the other is latent fingerprints left unintentionally when manipulating objects, which are very useful at criminal investigation scenes. Latent fingerprints are usually incomplete and distorted images, showing broken lines and noisy backgrounds. Due to their low quality, it is difficult to achieve clear, distinct, and sufficient feature reflection, feature quantity, and feature quality, which increases the risk of fingerprint recognition errors and seriously hinders the performance and application of fingerprint recognition technology. Improving the recognition rate of low-quality fingerprints and removing and effectively repairing noise in low-quality fingerprint images have significant guiding significance in practical applications.

[0004] Traditional methods for enhancing and restoring low-quality fingerprint images primarily focus on fingerprint preprocessing, but they fail to restore incomplete texture details and may even introduce new spurious feature points during denoising. Deep learning-based methods for enhancing and restoring low-quality fingerprint images often suffer from insufficient training data, making it difficult to achieve satisfactory results. Furthermore, the networks exhibit poor generalization ability, leading to a significant performance degradation when applying trained models to real latent fingerprints.

[0005] In conclusion, existing low-quality fingerprint image restoration methods are ineffective and cannot meet the needs of practical applications. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a method, system, medium, and device for repairing low-quality fingerprints based on tensor decomposition, which can obtain fast and accurate low-quality fingerprint image repair results and improve the accuracy of low-quality fingerprint recognition.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] The first aspect of the present invention provides a method for repairing low-quality fingerprints based on tensor decomposition.

[0009] In one or more embodiments, a low-quality fingerprint repair method based on tensor decomposition is provided, comprising:

[0010] Obtain low-quality fingerprint images and pre-divide them into several non-overlapping image blocks;

[0011] For each image patch, the most similar image patch is searched in parallel from different directions around it to construct a similar patch group for each image patch; among them, the most similar image patch in each different direction is found based on the maximum similarity coefficient, where the similarity coefficient is the product of the geometric similarity coefficient and the statistical similarity coefficient;

[0012] Denoising and enhancement processing is performed on the similar block groups of each image block to obtain the enhanced similar block groups of each image block;

[0013] The similar blocks in the enhanced similar block group of each image block are directly stacked along the third dimension to form a similarity tensor. After the similarity tensor is completed, the image blocks are then aggregated.

[0014] The fingerprint image formed by aggregating image blocks is then re-divided into several non-overlapping image blocks. Similar block groups are reconstructed for each image block, and noise reduction, enhancement, completion, and image block aggregation are performed sequentially until a preset number of iterations are completed to obtain the final repaired fingerprint image.

[0015] A second aspect of the present invention provides a low-quality fingerprint repair system based on tensor decomposition.

[0016] In one or more embodiments, a low-quality fingerprint repair system based on tensor decomposition is provided, comprising:

[0017] The low-quality fingerprint image acquisition module is used to acquire low-quality fingerprint images and pre-divide them into several non-overlapping image blocks;

[0018] The similarity block group construction module is used to find the most similar image blocks in parallel from different directions around each image block to construct a similarity block group for each image block; wherein, the most similar image blocks in each different direction are found based on the maximum similarity coefficient, and the similarity coefficient is the product of the geometric similarity coefficient and the statistical similarity coefficient;

[0019] The denoising and enhancement processing module is used to perform denoising and enhancement processing on the similar block group of each image block to obtain the enhanced similar block group of each image block;

[0020] The image patch aggregation module completes the image patch aggregation module. It is used to directly stack the similar blocks in the enhanced similar block group of each image patch along the third dimension to form a similar tensor. After completing the similar tensor, the image patch aggregation is then performed.

[0021] The iterative fingerprint image restoration module is used to re-divide the fingerprint image formed by aggregating image blocks into several non-overlapping image blocks, reconstruct similar block groups for each image block, and sequentially perform denoising enhancement, completion operations, and image block aggregation until the preset number of iterations is reached to obtain the final restored fingerprint image.

[0022] A third aspect of the present invention provides a computer-readable storage medium.

[0023] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the low-quality fingerprint repair method based on tensor decomposition as described above.

[0024] A fourth aspect of the present invention provides an electronic device.

[0025] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the low-quality fingerprint repair method based on tensor decomposition as described above.

[0026] Compared with the prior art, the beneficial effects of the present invention are:

[0027] This invention utilizes a multi-directional similar block search method driven by both fingerprint image data and direction to adapt to the structural characteristics of fingerprint images, obtaining more diverse similar blocks and improving subsequent fingerprint enhancement and repair effects. It employs a kernel Wiener filtering method to solve the problem of noise interference in low-quality fingerprint images. Furthermore, it leverages low-rank approximation and tensor decomposition methods to address the problem of rapid repair of incomplete fingerprint images, obtaining repair results suitable for low-quality fingerprint images and improving the accuracy of low-quality fingerprint recognition. Attached Figure Description

[0028] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0029] Figure 1 This is a flowchart illustrating the low-quality fingerprint repair method based on tensor decomposition according to an embodiment of the present invention.

[0030] Figure 2 This is a flowchart of the relative total variation kernel Wiener filtering method according to an embodiment of the present invention;

[0031] Figure 3 This is a flowchart of the low-rank tensor approximation algorithm according to an embodiment of the present invention;

[0032] Figure 4 This is a schematic diagram of the low-quality fingerprint repair system based on tensor decomposition according to an embodiment of the present invention;

[0033] Figure 5 This is a schematic diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0034] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0035] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0036] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0037] Figure 1 This is a flowchart illustrating a low-quality fingerprint repair method based on tensor decomposition in an embodiment of the present invention, as shown below. Figure 1 The low-quality fingerprint repair method based on tensor decomposition in this embodiment may include:

[0038] S101: Obtain a low-quality fingerprint image and pre-divide it into several non-overlapping image blocks.

[0039] Low-quality fingerprint images here include fingerprint images that are incomplete or distorted.

[0040] In the specific implementation of step S101, the low-quality fingerprint image can be divided into several non-overlapping image blocks, such as 2, depending on the actual situation. n A block, where n is a positive integer greater than or equal to 1.

[0041] S102: For each image patch, search for the most similar image patch in parallel from different directions around it to construct a similar patch group for each image patch; wherein, based on the maximum similarity coefficient, find the most similar image patch of each image patch in each different direction, and the similarity coefficient is the product of the geometric similarity coefficient and the statistical similarity coefficient.

[0042] In some specific implementations, the geometric similarity coefficient represents the directional similarity of image patches and is characterized by an oriented Gaussian weight function.

[0043] For example, the expression for the geometric similarity coefficient is:

[0044]

[0045] in, For image blocks and The geometric similarity coefficient; r i With r j It is an element in the image patch. and They are respectively in r i With r j The image patch centered at [x1, x2] represents the coordinate vectors in the corresponding directions, h is the smoothing factor, and the direction angle θ is in radians; T -1 These are intermediate parameters.

[0046] In some specific implementation processes, the statistical similarity coefficient is obtained by calculating the generalized similarity ratio using maximum likelihood estimation.

[0047] Based on the Neyman-Pearson theorem in statistical decision-making, the similarity of data blocks is calculated using the generalized similarity ratio criterion. In sparse and low-rank data filtering methods, similarity block calculation is of great significance.

[0048] For example, the statistical similarity coefficient is expressed as:

[0049]

[0050] in, It is an image block and The statistical similarity coefficient; r i With r j It is an element in the image patch. and They are respectively in r i With r j The image patch centered on α is a constant coefficient.

[0051] Similarity coefficient in this embodiment for:

[0052]

[0053] To adapt to the characteristics of fingerprint image stripe structure and obtain more diverse similar blocks, facilitating subsequent low-quality fingerprint image restoration, a multi-directional search strategy is designed: for reference blocks... Using point p as the center, in N-1 different directions around it, find the N-1 most similar blocks using the similarity block calculation criteria described above.

[0054] q∈Γ i ,Γ i={p+tu:0 <t<<r},u=(cosθ i sinθ i ) T

[0055] Where θ i = 2πi / N, i = 1, ..., N-1. Similar blocks in different directions are independent of each other, so multi-directional search can be computed in parallel, satisfying the requirements of fast and efficient computation.

[0056] Find the current image patch The N-1 most similar blocks form a similar block group. i = 1, ..., N.

[0057] S103: Perform denoising and enhancement processing on the similar block group of each image block to obtain the enhanced similar block group of each image block.

[0058] In one or more embodiments, such as Figure 2 As shown, the optimal kernel Wiener filter is used for denoising and enhancement of similar block groups for each image block. The process is as follows:

[0059] Singular Value Decomposition (SVD) is performed on similar block groups for each image patch to obtain eigenvalue coefficients;

[0060] After shrinking the eigenvalue coefficients, an inverse SVD operation is performed to obtain a preliminary estimated image patch;

[0061] The kernel of each image block in the preliminary estimated image block is calculated, and then the kernel Wiener filter with windowing is used to estimate the value of each pixel. The images are then aggregated to obtain the denoised image block.

[0062] This embodiment uses kernel Wiener filtering, which can better preserve the texture and detailed structure of fingerprint images and improve the recovery results.

[0063] It should be noted that Gaussian filtering, median filtering, mean filtering, etc., can also be used to perform noise reduction and enhancement processing on similar block groups of each image block.

[0064] S104: Stack the similar blocks in the enhanced similar block group of each image block directly along the third dimension to form a similarity tensor. After completing the similarity tensor, perform image block aggregation.

[0065] like Figure 3 As shown, during the process of completing the similar tensor, the rank of the similar tensor is minimized to make the recovered tensor as close as possible to the tensor formed by the observed data, thus completing the task of tensor completion. Since the same data point may belong to different data blocks, a weighted average aggregation operation is performed to obtain the result after low-quality fingerprint repair.

[0066] Tensor completion can be transformed into solving the following minimization problem:

[0067]

[0068] in, This represents the recovered tensor. The tensor representing the observed data. This indicates that the kernel tensor is decomposed by TR. The generated estimated tensor, P Ω It represents the set consisting of all observed elements.

[0069] Solving minimization problems using the alternating iterative multiplier method: introducing auxiliary variables. Alternately update the solution for the augmented Lagrangian function.

[0070] This invention introduces auxiliary variables and incorporates the equality constraints of the auxiliary variables into the Lagrange equation to obtain the augmented Lagrange function of the model.

[0071]

[0072] in It is a Lagrange multiplier tensor, where <·,·> denote the tensor inner product. Update them separately.

[0073] S105: The fingerprint image formed by aggregating image blocks is re-divided into several non-overlapping image blocks, and similar block groups of each image block are reconstructed. Then, denoising enhancement, completion operation and image block aggregation are performed in sequence until the preset number of iterations are completed to obtain the final repaired fingerprint image.

[0074] Figure 4 This is a schematic diagram of a low-quality fingerprint repair system based on tensor decomposition in an embodiment of the present invention. This embodiment is similar to... Figure 1 Corresponding to the low-quality fingerprint repair method based on tensor decomposition, such as Figure 4 As shown, the low-quality fingerprint repair system based on tensor decomposition in this embodiment may include:

[0075] The low-quality fingerprint image acquisition module 401 is used to acquire low-quality fingerprint images and pre-divide them into several non-overlapping image blocks;

[0076] The similarity block group construction module 402 is used to search for the most similar image block in parallel from different directions around each image block to construct a similarity block group for each image block; wherein, the most similar image block in each different direction is found based on the maximum similarity coefficient, and the similarity coefficient is the product of the geometric similarity coefficient and the statistical similarity coefficient;

[0077] The denoising and enhancement processing module 403 is used to perform denoising and enhancement processing on the similar block group of each image block to obtain the enhanced similar block group of each image block;

[0078] The image block aggregation module 404 is used to directly stack the similar blocks in the enhanced similar block group of each image block along the third dimension to form a similar tensor, and then perform image block aggregation after completing the similar tensor.

[0079] The iterative fingerprint image repair module 405 is used to re-divide the fingerprint image formed by aggregating image blocks into several non-overlapping image blocks, reconstruct similar block groups for each image block, and sequentially perform denoising enhancement, completion operations and image block aggregation until the preset number of iterations is reached to obtain the final repaired fingerprint image.

[0080] Wherein: in the similar block group construction module, the geometric similarity coefficient is the directional similarity of image blocks, which is characterized by an oriented Gaussian weight function; or / and the statistical similarity coefficient is obtained by calculating the generalized similarity ratio using maximum likelihood estimation.

[0081] It should be noted here that, Figure 4 The various modules in the tensor decomposition-based low-quality fingerprint restoration system, and... Figure 1 The steps in the tensor decomposition-based low-quality fingerprint repair method are the same and their specific implementation processes are identical, so they will not be repeated here.

[0082] Reference Figure 5 A schematic diagram of an electronic device is provided. It should be noted that... Figure 5 The electronic device 500 shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0083] like Figure 5 As shown, the electronic device 500 includes a central processing unit (CPU) 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 502 or a program loaded from a storage section 508 into a random access memory (RAM) 503. The RAM 503 also stores various programs and data required for system operation. The CPU 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0084] The following components are connected to I / O interface 505: an input section 506 including a keyboard, mouse, etc.; an output section 507 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a local area network (LAN) card, modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to I / O interface 505 as needed. A removable medium 511, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 510 as needed so that computer programs read from it can be installed into storage section 508 as needed.

[0085] When the central processing unit 501 in the electronic device of this embodiment executes the program, it achieves the following: Figure 1 The steps in the low-quality fingerprint repair method based on tensor decomposition are shown.

[0086] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including functions for executing... Figure 1 The program code for the method shown. In such an embodiment, the computer program can be downloaded and installed from a network via communication section 509, and / or installed from removable medium 511. When the computer program is executed by central processing unit 501, it performs the various functions defined in the apparatus of this application.

[0087] in, Figure 1 The computer program instructions corresponding to the method shown may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in the process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0088] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0089] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for repairing low-quality fingerprints based on tensor decomposition, characterized in that, include: Obtain low-quality fingerprint images and pre-divide them into several non-overlapping image blocks; For each image patch, the most similar image patch is searched in parallel from different directions around it to construct a similar patch group for each image patch; among them, the most similar image patch in each different direction is found based on the maximum similarity coefficient, where the similarity coefficient is the product of the geometric similarity coefficient and the statistical similarity coefficient; Denoising and enhancement processing is performed on the similar block groups of each image block to obtain the enhanced similar block groups of each image block; The similar blocks in the enhanced similar block group of each image block are directly stacked along the third dimension to form a similarity tensor. After the similarity tensor is completed, the image blocks are then aggregated. The fingerprint image formed by aggregating image blocks is then re-divided into several non-overlapping image blocks. Similar block groups for each image block are reconstructed, and noise reduction, enhancement, completion, and image block aggregation are performed sequentially until a preset number of iterations are completed to obtain the final repaired fingerprint image. The geometric similarity coefficient represents the directional similarity of image patches and is characterized using an oriented Gaussian weighting function; the expression for the geometric similarity coefficient is: ， ； in, For image blocks and The geometric similarity coefficient; r i and r j It is an element in the image patch. and They are respectively based on r i and r j The image patch centered on; x =[ x 1, x [2] represents the coordinate vector in the corresponding direction. h For smoothing factor, direction angle Take the radius in radians; For intermediate parameters; The statistical similarity coefficient is obtained by calculating the generalized similarity ratio using maximum likelihood estimation; the expression for the statistical similarity coefficient is: in, It is an image block and Statistical similarity coefficient; r i and r j It is an element in the image patch. and They are respectively based on r i and r j The image patch centered on; It is a constant coefficient.

2. The low-quality fingerprint repair method based on tensor decomposition as described in claim 1, characterized in that, For each image patch, similar block groups are subjected to denoising and enhancement processing using optimal kernel Wiener filtering.

3. A low-quality fingerprint repair system based on tensor decomposition, characterized in that, include: The low-quality fingerprint image acquisition module is used to acquire low-quality fingerprint images and pre-divide them into several non-overlapping image blocks; The similarity block group construction module is used to find the most similar image blocks in parallel from different directions around each image block to construct a similarity block group for each image block; wherein, the most similar image blocks in each different direction are found based on the maximum similarity coefficient, and the similarity coefficient is the product of the geometric similarity coefficient and the statistical similarity coefficient; The denoising and enhancement processing module is used to perform denoising and enhancement processing on the similar block group of each image block to obtain the enhanced similar block group of each image block; The image patch aggregation module completes the image patch aggregation module. It is used to directly stack the similar blocks in the enhanced similar block group of each image patch along the third dimension to form a similar tensor. After completing the similar tensor, the image patch aggregation is then performed. The iterative fingerprint image restoration module is used to re-divide the fingerprint image formed by aggregating image blocks into several non-overlapping image blocks, reconstruct similar block groups for each image block, and sequentially perform denoising enhancement, completion operations and image block aggregation until the preset number of iterations is reached to obtain the final restored fingerprint image. The geometric similarity coefficient represents the directional similarity of image patches and is characterized using an oriented Gaussian weighting function; the expression for the geometric similarity coefficient is: ， ； in, For image blocks and The geometric similarity coefficient; r i and r j It is an element in the image patch. and They are respectively based on r i and r j The image patch centered on; x =[ x 1, x [2] represents the coordinate vector in the corresponding direction. h For smoothing factor, direction angle Take the radius in radians; For intermediate parameters; The statistical similarity coefficient is obtained by calculating the generalized similarity ratio using maximum likelihood estimation; the expression for the statistical similarity coefficient is: in, It is an image block and Statistical similarity coefficient; r i and r j It is an element in the image patch. and They are respectively based on r i and r j The image patch centered on; It is a constant coefficient.

4. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the low-quality fingerprint repair method based on tensor decomposition as described in any one of claims 1-2.

5. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the low-quality fingerprint repair method based on tensor decomposition as described in any one of claims 1-2.

Images