A method for printing texture map topology analysis and vectorization
By processing printed texture images using a multi-vector field model and curve tracking algorithm, the problem of low vectorization accuracy in existing technologies is solved, achieving high-precision vectorization results. This method is suitable for topological analysis and vectorization of natural texture images.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANGHAI UNIV
- Filing Date
- 2023-03-13
- Publication Date
- 2026-05-22
AI Technical Summary
Existing technologies struggle to process vectorized images of complex textures with high precision, especially printed textures, and existing methods perform poorly when processing natural images.
A multi-vector field model of the image foreground points is established by combining the Sobel operator and the LBFGS++ algorithm. A topology map is formed by curve tracing, and Bezier interpolation fitting and smoothing filtering are performed to output the line skeleton as an SVG format vector file.
It improves vectorization accuracy and applicability, enabling it to handle natural texture maps, including printed texture maps, and reduces the impact of breakpoints, thus enhancing image continuity and visibility.
Smart Images
Figure CN116402869B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a vectorization method, and more particularly to a method for topological analysis and vectorization of printed texture maps. Background Technology
[0002] Vector graphics store images in the form of lines and closed shapes. These images retain their sharpness even when enlarged and are easy to edit, but they struggle to represent complex and varied colors, and their storage capacity is significantly affected by the complexity of the image content. Major image processing software such as Adobe Illustrator and CorelDRAW now integrate dedicated vectorization functions like Live Trace and PowerTRACE. Although vectorization has mature applications in the industry, these graphic vectorization algorithms are based on tracing the contours of connected components in the image, ultimately generating vector graphics with surfaces as their core structure.
[0003] The main challenge in vectorizing textured line images lies in extracting the points, lines, shapes, and color information from the raster image and converting them into vector data, i.e., into corresponding vector description codes. Furthermore, since lines in an image often intersect and have complex paths, separating the lines before vectorization becomes a critical challenge. Only by extracting all lines from the textured line image and converting them into independent, unbranched paths can these lines be vectorized and converted into vector data to be written into a vector file.
[0004] Chinese patent CN111462023A discloses an image texture line vectorization system and method, which utilizes neural networks for line separation and then performs Bézier curve fitting on each line. However, this method still suffers from low vectorization accuracy and is limited to processing unnatural images with simple textures.
[0005] Therefore, how to achieve a vectorization method for printed texture maps with high accuracy and wide processing range has become a technical problem that needs to be solved. Summary of the Invention
[0006] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a method for topological analysis and vectorization of printed texture maps.
[0007] The objective of this invention can be achieved through the following technical solutions:
[0008] A method for topological analysis and vectorization of printed texture maps, comprising the following steps:
[0009] 1) Preprocess the initial image;
[0010] 2) Establish a multi-vector field model of the image foreground points using the Sobel operator and the LBFGS++ algorithm;
[0011] 3) Establish a curve formed by connecting the foreground points, and perform curve tracking based on the multi-vector field model to form a topology map;
[0012] 4) Perform Bezier interpolation fitting on each group of curves according to the topology map results, and perform smoothing filtering. Finally, output an SVG format vector file with line skeleton as the primitive information.
[0013] Furthermore, the preprocessing specifically involves: performing denoising and histogram transformation operations on the initial image, and enhancing the foreground image using a dark channel-based dehazing model.
[0014] Furthermore, the denoising process employs a Gaussian filter to achieve linear smoothing, thereby completing the denoising.
[0015] Furthermore, the enhancement of the foreground image using the dark channel-based dehazing model is performed using the following formula:
[0016]
[0017] Where A is the atmospheric light component, t(x) is the transmittance, I(x) is the current image, and t0 is the lower limit of transmittance.
[0018] Furthermore, the foreground point is a point in the image whose grayscale is lower than a set threshold.
[0019] Furthermore, 2) specifically refers to:
[0020] 201) Construct complex equations:
[0021] f(z):=(z 2 -u 2 (z) 2 -v 2 )=z 4 +c2z 2 +c0
[0022] Where u and v represent the tangent vector and normal vector of the foreground point, respectively, z represents a complex variable, and c0 and c2 are constants;
[0023] 202) Set the objective function using the Sobel operator:
[0024]
[0025] Where C represents the plane containing vectors u and v, and θ ɡ θ represents the gradient calculated by the Sobel operator of size 3. τAdd 90 degrees to the gradient, where λ and μ are constants, and x represents the location of the foreground point in the image;
[0026] 203) The objective function is optimized using the LBFGS++ algorithm to obtain c0 and c2 for each foreground point. The tangent vector u and normal vector v of the foreground point are then mapped to obtain the foreground point, and a multi-vector field model of the foreground point is established.
[0027] Furthermore, 3) specifically refers to:
[0028] 301) Search for the foreground point with the closest tangent vector u and normal vector v in the neighborhood, obtain multiple curve bundles, and perform curve tracking. Each foreground point is tracked only once.
[0029] 302) At 8 adjacent foreground points, the matching direction of the multi-vector field is selected using the standard minimum angle matching criterion, and then a curve perpendicular to this matching direction is traced.
[0030] 303) Starting from each seed point, trace a test curve and record its intersection with the curve. Starting from a seed point, group the adjacent points on the test curve that are less than one pixel apart to form a series of curve bundles.
[0031] 304) Associate each curve bundle with a foreground point in the graph, and connect them to form a topological graph if they have at least one pair of intersections adjacent along their shared curve.
[0032] Furthermore, in 302), tracking stops when the curve leaves the narrow band or gets very close to a set distance from a curve with the same tangent.
[0033] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the method described above.
[0034] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.
[0035] Compared with the prior art, the present invention has the following beneficial effects:
[0036] I. This invention proposes a topology analysis and vectorization model based on multi-vector fields. It is the first time that a multi-vector field model has been applied to the field of vectorization. Combined with a curve tracking algorithm for topology extraction, it greatly improves the accuracy and fidelity of vectorization.
[0037] Second, this invention uses a dark channel dehazing model to enhance foreground elements in the image, improving the continuity and visibility of foreground lines and avoiding the impact of numerous breakpoints on the final vectorization result.
[0038] Third, the vectorization method of the present invention has a wide range of applications. It can process all natural texture images acquired by cameras and scanners, including printed texture images, as well as non-natural images. Attached Figure Description
[0039] Figure 1 This is a schematic diagram of the process of the present invention;
[0040] Figure 2 This is a diagram illustrating the dark channel enhancement effect of the present invention.
[0041] Figure 3 This is a comparison diagram of the output vector image and the input image of the present invention. Detailed Implementation
[0042] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.
[0043] Example
[0044] like Figure 1 As shown, a method for topology analysis and vectorization of printed texture maps is described, and the method includes the following steps:
[0045] 1) Preprocess the initial image.
[0046] The initial images are mostly obtained from printed products through scanners. The images contain noise, which will affect the vectorization effect. Denoising and histogram transformation operations are performed on the initial images, and the foreground points are enhanced using a dark channel-based dehazing model. The denoising is achieved by using a Gaussian filter to achieve linear smoothing, and histogram transformation and high-pass filter are used to enhance the spatial and frequency domains of the image, making the grayscale levels of the image more distinct and the edge contours clearer.
[0047] like Figure 2 As shown, the enhancement of the foreground image using the dark channel-based dehazing model is performed using the following formula:
[0048]
[0049] Where A represents the atmospheric light component, t(x) represents the transmittance, I(x) represents the current image, and t0 represents the lower limit of transmittance, typically set to 0.1. The default dark channel window size is 81. Guided filtering is used to obtain the transmittance map, improving the continuity and visibility of foreground lines and avoiding numerous breakpoints that could affect the final vectorization result.
[0050] The foreground point is a point in the image whose gray level is lower than a set threshold.
[0051] 2) A multi-vector field model of the image foreground points is established using the Sobel operator (first-order differential operator) and the LBFGS++ algorithm (optimized quasi-Newton algorithm).
[0052] 201) Construct complex equations:
[0053] f(z):=(z 2 -u 2 (z) 2 -v 2 )=z 4 +c2z 2 +c0
[0054] Where u and v represent the tangent vector and normal vector of the foreground point, respectively, z represents a complex variable, and c0 and c2 are constants;
[0055] 202) Define the objective function, which includes calibration, smoothing, and regularization terms, using the Sobel operator:
[0056]
[0057] Where C represents the plane containing vectors u and v, and θ ɡ θ represents the gradient calculated by the Sobel operator of size 3. τ Add 90 degrees to the gradient, where λ and μ are constants, and x represents the location of the foreground point in the image;
[0058] 203) The objective function is optimized using the LBFGS++ algorithm to obtain c0 and c2 for each foreground point. The tangent vector u and normal vector v of the foreground point are then mapped to obtain the foreground point, and a multi-vector field model of the foreground point is established.
[0059] 3) Establish a curve formed by connecting the foreground points, and perform curve tracking based on the multi-vector field model to form a topology map.
[0060] 301) Search for the foreground point with the closest tangent vector u and normal vector v in the neighborhood, obtain multiple curve bundles, and perform curve tracking. Each foreground point is tracked only once.
[0061] 302) At 8 adjacent foreground points, the matching direction of the multi-vector field is selected using the standard minimum angle matching criterion. Then, a curve perpendicular to this matching direction is traced. When the curve leaves the narrow band or gets very close to a curve with the same tangent at a set distance, tracing is stopped to prevent multiple tracing attempts. The set distance is 0.01. The multi-vector field is expanded using the same procedure each time it moves to a nearby foreground point.
[0062] 303) Starting from each seed point, trace a test curve and record its intersection with the curve. Starting from a seed point, group the adjacent points on the test curve that are less than one pixel apart to form a series of curve bundles.
[0063] 304) Associate each curve bundle with a foreground point in the graph, and connect them to form a topological graph if they have at least one pair of intersections adjacent along their shared curve.
[0064] 4) such as Figure 3 As shown, each group of curves is fitted with Bezier interpolation according to the topology map results, and then smoothed and filtered. The final output is an SVG format vector file with line skeleton as the primitive information.
[0065] The present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the method described above.
[0066] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.
[0067] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.
Claims
1. A method for topological analysis and vectorization of printed texture maps, characterized in that, The method includes the following steps: 1) Preprocess the initial image; 2) Establish a multi-vector field model of the image foreground points using the Sobel operator and the LBFGS++ algorithm; 3) Establish a curve formed by connecting the foreground points, and perform curve tracking based on the multi-vector field model to form a topology map; 4) Perform Bézier interpolation fitting on each group of curves according to the topology map results, and perform smoothing filtering. Finally, output an SVG format vector file with line skeleton as the primitive information. Specifically, 2) refers to: 201) Construct complex equations: in, u, v Let these represent the tangent vector and normal vector of the foreground point, respectively. z Represents a complex variable. c 0 , c 2 It is a constant; 202) Set the objective function using the Sobel operator: in, C Representative vector u,v The plane in which it is located θ ɡ This represents the gradient calculated using the Sobel operator of size 3. θ τ Adding 90 degrees to this gradient, where λ and μ are both constants. x Indicates the location of the foreground point in the image; 203) The objective function was optimized using the LBFGS++ algorithm to obtain each foreground point. c 0 and c 2 This mapping yields the tangent vector of the foreground point. u and normal vector v And establish a multi-vector field model of the foreground point; Specifically, 3) refers to: 301) Search for the tangent vector within the neighborhood. u and normal vector v The closest foreground point is used to obtain multiple curve bundles for curve tracking, with each foreground point being tracked only once. 302) On 8 adjacent foreground points, the matching direction of the multi-vector field is selected using the standard minimum angle matching criterion, and then a curve perpendicular to this matching direction is traced. 303) Starting from each seed point, trace a test curve and record its intersection with the curve. Starting from a seed point, group the adjacent points on the test curve that are less than one pixel apart to form a series of curve bundles. 304) Associate each curve bundle with a foreground point in the graph, and connect them to form a topological graph if they have at least one pair of intersections adjacent along their shared curve.
2. The method for topological analysis and vectorization of printed texture maps according to claim 1, characterized in that, The preprocessing specifically involves: performing denoising and histogram transformation operations on the initial image, and enhancing the foreground image using a dark channel-based dehazing model.
3. The method for topological analysis and vectorization of printed texture maps according to claim 2, characterized in that, The denoising process employs a Gaussian filter to achieve linear smoothing, thus completing the denoising.
4. The method for topological analysis and vectorization of printed texture maps according to claim 2, characterized in that, The enhancement of the foreground image using the dark channel-based dehazing model is performed using the following formula: in, It is an atmospheric light component. t ( x ) represents transmittance. I ( x (This refers to the current image.) t 0 This is the lower limit of transmittance.
5. The method for topological analysis and vectorization of printed texture maps according to claim 1, characterized in that, The foreground point is a point in the image whose gray level is lower than a set threshold.
6. The method for topological analysis and vectorization of printed texture maps according to claim 1, characterized in that, In 302), tracking stops when the curve leaves the narrow band or gets very close to a set distance from a curve with the same tangent.
7. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 6.
8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 6.