Oracle bone fracture surface contour conjugation method based on feature priority

Abstract

The invention provides an oracle bone fracture surface contour conjugation method based on feature priority, which comprises the following steps: S1, taking an oracle bone fracture surface contour diagram as a matched object, and obtaining feature similarity by adopting a multi-fractal spectrum; s2, combining the feature similarity obtained through the multi-fractal spectrum with the feature information of the gray histogram to obtain a matching cost parameter value, determining the section matching priority in a product form, and carrying out the conjugation operation of the oracle bone; and S3, completing conjugation, determining the judgment priority of edge matching in a product form by utilizing the splicing coincident edge information ratio and the gradient amplitude mean value, and carrying out judgment operation on whether the conjugation of the oracle bone is proper or not. According to the method, the oracle bone fragments are regarded as three-dimensional information, three-dimensional information and two-dimensional information depend on each other, more effective information is provided for the conjugation method, and the oracle bone conjugation process is more perfect and effective.

Description

Technical Field

[0001] The invention belongs to the field of cultural relics digitization research, and specifically relates to an oracle bone section contour splicing method that uses a priority determined by fusing multiple fractal spectrum features with histogram features for matching. Background Art

[0002] Oracle bone script is a script carved by the royal family of the Shang Dynasty on tortoise shells or animal bones for divination and record keeping. It is also the earliest mature Chinese character to date, with a history of more than 3,000 years, and contains profound historical and cultural value. Oracle bone rubbings are mostly made of tortoise shells and animal bones, but due to environmental and material reasons, they are prone to physical damage and become oracle bone fragments, affecting their preservation and inheritance.

[0003] The splicing of oracle bone fragments is a basic and important research in oracle bone archaeology. Due to the huge number of fragments and their similarity, manual splicing is far from meeting the needs. Therefore, the splicing of digital oracle bone fragments with the help of image processing and computer-aided technology can not only avoid secondary damage but also improve efficiency and accuracy, which is of great research significance.

[0004] Since oracle bone fragments are relatively thin, most of the splicing work is to process and study the oracle bone rubbings, mainly focusing on the edge and contour features of the rubbings. At present, the splicing of oracle bone rubbings mostly uses two-dimensional rubbings images to match fragments through inscriptions, semantics, context, edges and other related information. However, the fracture surface contour of oracle bone fragments contains rich information and also plays a vital role in splicing. Therefore, this patent regards the oracle bone fragments as three-dimensional entities, obtains the fracture surface contour and matches it with the priority determined by the multi-fractal spectrum features and the histogram features; then, the edge matching judgment priority is determined in the form of a product using the splicing overlap edge information ratio and the gradient amplitude mean, and the judgment operation of whether the splicing of oracle bones is appropriate is performed. Summary of the invention

[0005] In view of the problems existing in the prior art, the present invention intends to achieve the following research objectives through an oracle bone section contour splicing method based on the fusion of multifractal spectrum features and histogram features: (1) taking the oracle bone section contour map as the matching object, using multifractal spectrum to obtain feature similarity, and using local feature similarity and global feature similarity between sections to form feature similarity; (2) using the feature similarity obtained by multifractal spectrum and the grayscale histogram feature information to obtain the matching cost parameter value to determine the section matching priority in the form of product, and perform the oracle bone splicing operation; (3) after the splicing is completed, using the splicing overlap edge information ratio and the gradient amplitude mean in the form of product to determine the edge matching judgment priority, and perform the oracle bone splicing operation to determine whether the oracle bone splicing is appropriate.

[0006] The present invention adopts the following technical solution to solve the problems existing in the prior art: a method for joining the contour of the fractured surface of an oracle bone based on feature priority, comprising the following steps:

[0007] S1, taking the oracle bone cross-section contour image as the matching object, using multi-fractal spectrum to obtain feature similarity, and using the local feature similarity between the cross-sections and the global feature similarity to form feature similarity;

[0008] S2, combining the feature similarity obtained by the multi-fractal spectrum with the grayscale histogram feature information to obtain a matching cost parameter value, using the product form to determine the priority of the cross-section matching, and performing the oracle bone concatenation operation;

[0009] S3. After the splicing is completed, the priority of edge matching is determined by multiplying the splicing overlap edge information ratio and the gradient amplitude mean value, and a judgment operation is performed to determine whether the splicing of the oracle bone is appropriate.

[0010] The step S1 specifically includes the following steps:

[0011] S1.1. Preprocess the image of the oracle bone fragment to obtain the contour map of the cross section: The preprocessing process is based on the fact that the contour map of the cross section is a set of points, which can be called the target shape point map. The target shape point map can be represented by a finite set of points. The target shape is piecewise smooth. Determine a point set, P = {p 1 ,...,p n} i ∈R 2 , when n is large enough, an approximation of the target shape that is basically continuous can be obtained;

[0012] S1.2. Determine the multifractal spectrum of the target shape point diagram and obtain the feature similarity value determined by the local singularity parameter value and the global singularity parameter value. The multifractal spectrum is a more detailed measure of the complexity, irregularity and unevenness of the fractal structure. There are two feature quantities in the multifractal spectrum: the holder index α and the spectrum function f(α). α represents the local singularity of the image, and f(α) represents the global singularity of the image. Since the direct calculation method of α~f(α) based on the measurement space has the advantage of high efficiency, this method is used to obtain the parameters α and f(α). Specifically, for any point p in the target shape, i , taking this as the center point, select a pixel block of size m×m, denoted as V(i), then we can get p i The holder index α of the point is denoted as α(i), as shown in formula (1):

[0013]

[0014] Based on the obtained α(i), its maximum and minimum values ​​are obtained, which are denoted as α max (i) and α min (i) min (i),α max (i)] is divided into N regions, and the number of boxes to which α(i) belongs, N(α(i)), is counted to obtain f(α(i)), as shown in formula (2):

[0015]

[0016] On the basis of obtaining α(i) and f(α(i)), the multifractal spectrum of the target shape point graph α(i)-f(α(i)) is obtained. On the basis of obtaining α(i) and f(α(i)), the similarity of the contour graph holder index H(p,q) and the multifractal spectrum similarity M(p,q) of the contour graph are obtained. The mathematical expressions are as follows:

[0017]

[0018] Among them, π represents permutation, which means the constraint is a one-to-one correspondence of contour matching, c 1 and c 2 It represents a very small constant to prevent the denominator from being zero. Therefore, for the contour section group to be matched, MTSS(p,q) is used to represent their multifractal spectrum similarity, as shown in formula (5):

[0019] MTSS(p,q)=H(p,q)×M(p,q) (5)

[0020] Among them, MTSS(p,q) represents the multifractal spectrum similarity of the target shape point graph. The smaller the value, the higher the matching similarity, and vice versa.

[0021] The step S2 specifically includes the following steps:

[0022] Step S2.1, determine the grayscale histogram feature information of the pixel complement point of the target shape point map, and obtain the matching cost parameter value. Any sampling point and the remaining points can form a set of vectors to describe the relative position of the entire shape and the reference point. For any point p in the target shape i The histogram h of the relative coordinates of the remaining n-1 points can be calculated i , this figure is point p i The grayscale histogram feature map of the pixel complement point is shown in formula (6):

[0023] h i (k)=#{q≠p i :(qp i )∈bin(k)} (6)

[0024] Point p i The coordinates of the remaining points are measured as the starting point, and the number of sampling points in each small area is calculated using polar coordinates. The maximum black part in the histogram indicates that there are the most sampling points in this area;

[0025] Therefore, for any point p in the target shape point graph i and the corresponding point q in the shape point map to be matched j , using C (i，j) Represents the matching cost between pixels, as shown in formula (7):

[0026]

[0027] In the above formula, Represents p i ,q j The k-bin of the normalized histogram is expressed as follows:

[0028]

[0029] Among them, π represents permutation, which means the constraint is a one-to-one correspondence of contour matching. The smaller the value of the matching cost C(p,q), the higher the matching similarity, and vice versa;

[0030] Step S2.2, determine the matching priority of the target shape point map. The matching priority is determined by two parameters: multifractal spectrum similarity and matching cost. The definition is shown in formula (9):

[0031] Matching priority(p,q)=MTSS(p,q)×C(p,q) (9)

[0032] In the above formula, the smaller the value of matching priority (p, q), the higher the matching similarity, and the priority matching. Conversely, the lower the matching similarity, the later matching. The obtained matching priority order is used to perform the oracle bone splicing operation.

[0033] The step S3 specifically includes the following steps:

[0034] Step S3.1, obtaining the overlapping edge of the two-dimensional image of the two pieces of spliced ​​fragments P and Q with the highest priority matching degree obtained in step S2, using a pixel block centered on pixel point o to obtain the occupancy and gradient along the overlapping edge, setting c(P) as the information ratio of fragment P in the pixel block, c(Q) as the information ratio of fragment Q in the pixel block, g(P) as the edge gradient of the overlapping edge fragment P in the pixel block, and g(Q) as the edge gradient of the overlapping edge fragment Q in the pixel block;

[0035] Step S3.2, optimal splicing judgment, for any pixel point on the overlapping edge, the judgment condition is expressed by the following mathematical expression as follows:

[0036] Determine priority(P,Q)=c(P,Q)×GMS(P,Q) (10)

[0037] The judgment priority is composed of the fragment information ratio and the gradient amplitude similarity, where:

[0038]

[0039] In the above formula, c(P,Q) represents the edge information ratio and has a value range of 0-1. The larger the value, the higher the edge line overlap rate, and the more matched the splicing is; GMS(P,Q) represents the edge gradient amplitude similarity. The larger the value, the higher the edge line overlap rate, and the more matched the splicing is. Among them, M(g(P) i ) and M(g(Q) i ) represents the gradient amplitude, Determine priority (P, Q) is used as the index for judging the conjugation matching degree, and the larger the value, the higher the matching degree.

[0040] The present invention has the following advantages:

[0041] The method of the present invention regards the oracle bone fragments as three-dimensional solids. The three-dimensional information and the two-dimensional information rely on each other, providing more effective information for the splicing method, making the oracle bone splicing process more complete and effective. On the one hand, the multi-fractal spectrum similarity of the fragment cross-sectional contour is integrated with the matching cost based on the histogram feature information to obtain priority and perform cross-sectional matching splicing. On the other hand, on the basis of splicing, the fragment information ratio of the pixel blocks on the overlapping edges and the gradient amplitude similarity priority are used to judge whether the splicing method is reasonable. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic diagram of the cross-sectional profile of the oracle bone fragments of the present invention;

[0043] Figure 2 It is a schematic diagram of the edge splicing of the cross-section of oracle bone fragments of the present invention. DETAILED DESCRIPTION

[0044] The technical solution of the present invention is further specifically described below through embodiments and in conjunction with the accompanying drawings. A method for combining oracle bone section contours based on the fusion of multi-fractal spectrum features and histogram features comprises the following steps:

[0045] Step 1: Joining the outline of the oracle bone section:

[0046] like Figure 1As shown in the figure, the oracle bone section contour concatenation model proposed in this paper consists of four steps: obtaining data sets, extracting and sorting oracle bone section contours, calculating concatenation matching priority, and concatenating;

[0047] Step 1.1, the image of the oracle bone fragment needs to be preprocessed to obtain the contour map of the cross section: the contour map of the cross section is a set of points, which can be called the target shape point map. The target shape point map can be represented by a finite set of points. The target shape is piecewise smooth. Determine a point set, P = {p 1 ,...,p n} i ∈R 2 , when n is large enough, an approximation of the target shape that is basically continuous can be obtained;

[0048] Step 1.2, determine the multifractal spectrum of the target shape point diagram, and obtain the feature similarity value determined by the local singularity parameter value and the global singularity parameter value. The multifractal spectrum is a more detailed measure of the complexity, irregularity and unevenness of the fractal structure. There are two feature quantities in the multifractal spectrum: the holder index α and the spectrum function f(α). α represents the local singularity of the image, and f(α) represents the global singularity of the image. Since the direct calculation method of α~f(α) based on the measurement space has the advantage of high efficiency, this method is used to obtain the parameters α and f(α). Specifically, for any point p in the target shape, i , taking this as the center point, select a pixel block of size m×m, denoted as V(i), then we can get p i The holder index α of the point is denoted as α(i), as shown in formula (1):

[0049]

[0050] Based on the obtained α(i), its maximum and minimum values ​​are obtained, which are denoted as α max (i) and α min (i) min (i),α max (i)] is divided into N regions, and the number of boxes to which α(i) belongs, N(α(i)), is counted to obtain f(α(i)), as shown in formula (2):

[0051]

[0052] On the basis of obtaining α(i) and f(α(i)), the multifractal spectrum of the target shape point graph α(i)-f(α(i)) is obtained. On the basis of obtaining α(i) and f(α(i)), the similarity of the contour graph holder index H(p,q) and the multifractal spectrum similarity M(p,q) of the contour graph are obtained. The mathematical expressions are as follows:

[0053]

[0054] Among them, π represents permutation, which means the constraint is a one-to-one correspondence of contour matching. 1 and c 2 It represents a very small constant to prevent the denominator from being zero. Therefore, for the contour section group to be matched, MTSS(p,q) is used to represent their multifractal spectrum similarity, as shown in formula (5):

[0055] MTSS(p,q)=H(p,q)×M(p,q) (5)

[0056] Among them, MTSS(p,q) represents the multifractal spectrum similarity of the target shape point graph. The smaller the value, the higher the matching similarity, and vice versa.

[0057] Step 1.3, determine the grayscale histogram feature information of the pixel complement point of the target shape point map, and obtain the matching cost parameter value. Any sampling point and the remaining points can form a set of vectors, which can describe the relative position of the entire shape and the reference point; for any point p in the target shape i The histogram h of the relative coordinates of the remaining n-1 points can be calculated i , this figure is point p i The grayscale histogram feature map of the pixel complement point is shown in formula (6):

[0058] h i (k)=#{q≠p i :(qp i )∈bin(k)} (6)

[0059] Point p i The coordinates of the remaining points are measured as the starting point, and the number of sampling points in each small area is calculated using polar coordinates. The maximum black part in the histogram indicates that there are the most sampling points in this area;

[0060] Therefore, for any point p in the target shape point graph i and the corresponding point q in the shape point map to be matched j , using C (i，j) Represents the matching cost between pixels, as shown in formula (7):

[0061]

[0062] In the above formula, Represents p i ,q j The k-bin of the normalized histogram is expressed as follows:

[0063]

[0064] Among them, π represents permutation, which means the constraint is a one-to-one correspondence of contour matching. The smaller the value of the matching cost C(p,q), the higher the matching similarity, and vice versa.

[0065] Step 1.4: Determine the matching priority of the target shape point graph. The matching priority is determined by two parameters: multifractal spectrum similarity and matching cost. The definition is shown in formula (9):

[0066] Matching priority(p,q)=MTSS(p,q)×C(p,q) (9)

[0067] In the above formula, the smaller the value of matching priority (p, q), the higher the matching similarity, the higher the matching priority. Conversely, the lower the matching similarity, the later the matching.

[0068] Step 2: Best conjugation determination:

[0069] The splicing is performed by the splicing method determined by the cross-section contour priority, and the characteristic quantity of the splicing overlap edge is used to judge whether the splicing method is appropriate. The specific steps are as follows:

[0070] Step 2.1, obtain the coincident edge of the two-dimensional images of the concatenated fragments P and Q, and use an l×l pixel block centered at o to obtain the occupancy and gradient along the coincident edge, as shown in Figure 2 As shown, c(P) represents the information ratio of fragment P in the pixel block, c(Q) represents the information ratio of fragment Q in the pixel block, g(P) represents the edge gradient of the coincident edge fragment P in the pixel block, and g(Q) represents the edge gradient of the coincident edge fragment Q in the pixel block;

[0071] Step 2.2, optimal splicing judgment: For any pixel point i on the overlapping edge, the judgment condition is expressed by the following mathematical expression as follows:

[0072] Determine priority(P,Q)=c(P,Q)×GMS(P,Q) (10)

[0073] The judgment priority is composed of the fragment information ratio and the gradient amplitude similarity, where

[0074]

[0075] From the above formula, we can see that c(P,Q) represents the edge information ratio and has a value range of 0 to 1. The larger the value, the higher the edge line overlap rate, and the more matched the splicing is; GMS(P,Q) represents the edge gradient amplitude similarity. The larger the value, the higher the edge line overlap rate, and the more matched the splicing is. Among them, M(g(P) i ) and M(g(Q) i ) represents the gradient amplitude.

[0076] The protection scope of the present invention is not limited to the above-mentioned embodiments. Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the scope and spirit of the present invention. If these changes and modifications fall within the scope of the claims of the present invention and their equivalents, the present invention is intended to include these changes and modifications.

Claims

1. A method for assembling the contour of oracle bone fracture surface based on feature priority, characterized in that: The steps include: S1, taking the oracle bone cross-section contour image as the matching object, and using multi-fractal spectrum to obtain feature similarity; S2, combining the feature similarity obtained by the multi-fractal spectrum with the grayscale histogram feature information to obtain a matching cost parameter value, using the product form to determine the priority of the cross-section matching, and performing the oracle bone concatenation operation; S3. After the splicing is completed, the priority of edge matching is determined by multiplying the splicing overlap edge information ratio and the gradient amplitude mean value, and a judgment operation is performed to determine whether the splicing of the oracle bone is appropriate.

2. The method for assembling the contour of an oracle bone fracture surface based on feature priority as claimed in claim 1, characterized in that: The step S1 specifically includes the following steps: S1.1, preprocessing the oracle bone fragment image to obtain the cross-section contour map, i.e., the target shape point map; S1.

2. Determine the multifractal spectrum of the target shape point diagram, and obtain a feature similarity value determined by a local singularity parameter value and a global singularity parameter value.

3. The method for combining the contours of oracle bone fracture surfaces based on feature priority as claimed in claim 2, characterized in that: The preprocessing process in step S1.1 is based on the fact that the contour map of the cross section is a set of points, that is, the target shape point map, which is represented by a finite set of points. The target shape is piecewise smoothed, and a point set is determined, P = {p 1, ...,p n } i ∈R 2 , when n is large enough, a basically continuous approximation of the target shape is obtained.

4. The method for assembling the contour of an oracle bone fracture surface based on feature priority as claimed in claim 2, characterized in that: The multifractal spectrum in step S1.2 is a more detailed measure of the complexity, irregularity and unevenness of the fractal structure. There are two characteristic quantities in the multifractal spectrum: the holder index α and the spectrum function f(α). α represents the local singularity of the image, and f(α) represents the global singularity of the image. Since the direct calculation method of α~f(α) based on the measurement space has the advantage of high efficiency, this method is used to obtain the parameters α and f(α). Specifically, for any point p in the target shape, i , taking this as the center point, select a pixel block of size m×m, denoted as V(i), then we can get p i The holder index α of the point is denoted as α(i), as shown in formula (1): Based on the obtained α(i), its maximum and minimum values ​​are obtained, which are denoted as α max (i) and α min (i) min (i),α max (i)] is divided into N regions, and the number of boxes to which α(i) belongs, N(α(i)), is counted to obtain f(α(i)), as shown in formula (2): On the basis of obtaining α(i) and f(α(i)), the multifractal spectrum of the target shape point graph α(i)-f(α(i)) is obtained. On the basis of obtaining α(i) and f(α(i)), the similarity of the contour graph holder index H(p,q) and the multifractal spectrum similarity M(p,q) of the contour graph are obtained. The mathematical expressions are as follows: Among them, π represents permutation, which means the constraint is a one-to-one correspondence of contour matching, c1 and c2 represent very small constants to prevent the denominator from being zero. Therefore, for the contour section group to be matched, MTSS(p,q) is used to represent their multifractal spectrum similarity, as shown in formula (5): MTSS(p,q)=H(p,q)×M(p,q) (5) Among them, MTSS(p,q) represents the multifractal spectrum similarity of the target shape point graph. The smaller the value, the higher the matching similarity, and vice versa.

5. The method for combining the contours of oracle bone fracture surfaces based on feature priority as claimed in claim 1, characterized in that: The process of confirming the matching cost parameter value in step S2 is as follows: S2.

1. Determine the grayscale histogram feature information of the pixel complement point of the target shape point map and obtain the matching cost parameter value. Any sampling point and the remaining points can form a set of vectors to describe the relative position of the entire shape and the reference point. For any point p in the target shape i Calculate the histogram h of the relative coordinates of the remaining n-1 points i , this figure is point p i The grayscale histogram feature map of the pixel complement point is shown in formula (6): h i (k)=#{q≠p i :(q-p i )∈bin(k)} (6) Point p i The coordinates of the remaining points are measured as the starting point, and the number of sampling points in each small area is calculated using polar coordinates. The maximum black part in the histogram indicates that there are the most sampling points in this area; Therefore, for any point p in the target shape point graph i and the corresponding point q in the shape point map to be matched j , using C (i，j) Represents the matching cost between pixels, as shown in formula (7): In the above formula, Represents p i ,q j The k-bin of the normalized histogram, the contour matching cost expression is as follows: Among them, π represents permutation, which means the constraint is a one-to-one correspondence of contour matching. The smaller the value of the matching cost C(p,q), the higher the matching similarity, and vice versa.

6. The method for combining the contours of oracle bone fracture surfaces based on feature priority as claimed in claim 5, characterized in that: The process of performing oracle conjugation based on matching priority in step S2 is as follows: S2.2, determine the matching priority of the target shape point map. The matching priority is determined by two parameters: multifractal spectrum similarity and matching cost. The definition is shown in formula (9): Matching priority(p,q)=MTSS(p,q)×C(p,q) (9) In the above formula, the smaller the value of matching priority (p, q), the higher the matching similarity, and the priority matching. Conversely, the lower the matching similarity, the later matching. The obtained matching priority order is used to perform the oracle bone splicing operation.

7. The method for combining the contours of oracle bone fracture surfaces based on feature priority as claimed in claim 1, characterized in that: The step S3 specifically includes the following steps: S3.

1. Obtain the overlapping edge of the two-dimensional image of the two pieces of spliced ​​fragments P and Q with the highest priority matching degree obtained in step S2, use a pixel block centered on pixel point o to obtain the occupancy and gradient along the overlapping edge, set c(P) as the information ratio of fragment P in the pixel block, c(Q) as the information ratio of fragment Q in the pixel block, g(P) as the edge gradient of the overlapping edge fragment P in the pixel block, and g(Q) as the edge gradient of the overlapping edge fragment Q in the pixel block; S3.2, the best stitching judgment, for any pixel point on the overlapping edge, the judgment condition is expressed by the following mathematical expression as follows: Determine priority(P,Q)=c(P,Q)×GMS(P,Q) (10) The judgment priority is composed of the fragment information ratio and the gradient amplitude similarity, where: In the above formula, c(P,Q) represents the edge information ratio and has a value range of 0-1. The larger the value, the higher the edge line overlap rate, and the more matched the splicing is; GMS(P,Q) represents the edge gradient amplitude similarity. The larger the value, the higher the edge line overlap rate, and the more matched the splicing is. Among them, M(g(P) i ) and M(g(Q) i ) represents the gradient amplitude, Determine priority (P, Q) is used as the index for judging the conjugation matching degree, and the larger the value, the higher the matching degree.

Images