JP2011022429A - Writing brush font creating method and writing brush font creating program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method for creating writing brush font in which crack and bleeding are added to the writing brush font without using an image filter. <P>SOLUTION: Graphs which represent both character skeletons of Reference which becomes a reference instance and Target which becomes a processing target are obtained as a set of edges connecting nodes, and correspondence relationship between edges configuring each graph is determined. As shown by 16 and 17 in the figure, an edge of a left-falling stroke in Reference Shape kanji pronounced "ten" is allocated to an edge of a left-falling stroke in Target kanji pronounced "dai". By this allocation, character areas around the edges are associated. The corresponding character area on the Reference (portion of the left-falling stroke of "ten") is extracted as a texture and pasted to the character area of the left-falling stroke in the Target "dai". In this way, allocation of the edge of the Reference Shape corresponding to each edge of the Target, association of the character area around the edge to which the Reference is allocated, and pasting of the texture are executed for all edges configuring the Target, and output font is obtained. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a brush font creation method and a brush font creation program for creating a brush font by adding a blurring feature of a brush brush character as a reference example to a brush font to be processed.

  So far, there have been several reports on creation and processing of brushstroke fonts. The brush blotting expression technology (Patent Document 1, Non-Patent Documents 1 and 2) using a brush cursor and mathematical morphology is a typical example of the research, but it is not suitable for existing brush fonts to be processed. It adds drowning / bleeding design and adds hand-drawn artistic features to the font. This method corresponds to an approach using an image filter.

  On the other hand, although the processing target is not a brush brush font, there is a method of converting a general image into a painting style by texture synthesis. Non-Patent Documents 3 and 4 are typical examples. In addition to general images to be processed, non-patent documents 3 and 4 are prepared using reference images such as painting images of famous painters and using their artistic features. The image is processed and converted into a picture-like image similar to the reference example.

JP-A-10-83177

Takeshi Nakamura, Yuji Mano, Hirohisa Seki, Hidenori Ito, “On the brush font blurring / bleeding processing system”, IPSJ Journal, vol.38, no.5, pp.1008--1015, May. 1997. Takayuki Ichikawa, Tomoyuki Idokawa, Masayoshi Tsutsumi, “Folding and blurring of brush fonts using mathematical morphology”, Journal of Japan Society of Applied Mathematical Sciences, vol.10, no.3, pp.263--272, Sept. 2000. A. Hertzmann, C. Jacobs, N. Oliver, B. Curless, and D. Salesin, `` Image Analogies '', Proc. SIGGRAPH '01, pp.327--340, Aug.2001. G. Ramanarayanan, and K. Bala, `` Constrained Texture Synthesis via Energy Minimization '', IEEE Transactions on Visualization and Computer Graphics, vol.13, no.1, pp.167--178, Jan.2007.

  The problem with those described in Patent Document 1 and Non-Patent Documents 1 and 2 is that an image filter is used without adding a reference example such as a work image of a famous writer in the process of adding a blur or blur to a brush brush font. There is in processing. For this reason, the design of the image filter greatly affects the quality of the brush brush font that is finally processed and output. However, the design of this filter is not easy, and it is difficult to predict what output will be obtained by the filter, so it is not possible to add any blurring or blurring features expected by the designing user to the font. It is almost impossible.

  Non-Patent Documents 3 and 4 can obtain the output image expected by the user because the artistic feature can be executed by the user himself / herself visually confirming the art feature by using the reference example. I can do it. However, this method cannot perform processing based on the structure of a processing target having a structure in which the direction in which the brush advances is clear, such as a brushstroke character. For this reason, even if it is used for font processing, the features of blurring and blurring become unnatural, and it is not possible to reflect features similar to the reference case in the output.

  The present invention has been made in view of the above points, and an object of the present invention is to create a writing brush font in which a blur is added to the writing brush font without using an image filter.

In order to achieve the above object, the invention described in claim 1 is a brush font creation method for creating a brush font by adding a blurring feature of a brush stroke character as a reference example to a target font to be processed. ,
Obtaining a character skeleton of both the reference case and the target font as a graph expressed as a set of edges connecting nodes.
Associating edges of graphs of both the reference case and the target font;
Pasting the blurring feature around the edge of the reference case corresponding to the edge of the target font.

  According to a second aspect of the present invention, there is provided a program for causing a computer to execute the brush font creation method described above.

It is a figure which shows the outline | summary of the writing brush font production method which concerns on one Embodiment of this invention. It is a figure which shows the graph structure which shows the character skeleton of Reference Shape and Reference. It is a figure which shows the graph structure which shows the character skeleton of Target. It is a conceptual diagram of a writing brush font creation method. It is a figure which shows the hardware constitutions which implement the writing brush font production method which concerns on one Embodiment of this invention. It is a flowchart which shows the program for performing the writing brush font production method. It is a figure which shows the conceptual diagram of the edge allocation process in FIG. It is a flowchart which shows the Style evaluation part of an edge allocation process. It is a flowchart which shows the Length and Direction evaluation part of an edge allocation process. It is a flowchart which shows a texture sticking process. It is a conceptual diagram of a texture sticking process. It is a figure which shows the output example of the input / output example of this invention, and a conventional research.

  FIG. 1 shows an outline of a writing brush font creation method according to an embodiment of the present invention. In this writing brush font creation method, Reference 1 shown in FIG. 1 is a reference example, and Target 3 shown in FIG. 1 is a processing target. The output font Output shown in 4 is obtained by adding the blurring feature of Reference to Target. However, in addition to Reference and Target, the Reference Shape shown at 3 in FIG. The reference shape is prepared separately to distinguish the reference character area (the blacked-in and brushed parts) from the background area.

  Reference Shape 3 shown in FIG. 1 and Target 2 shown in FIG. 1 each acquire the character skeleton by thinning processing. Each character skeleton can be represented by the undirected graph structure shown in FIGS. FIG. 2 shows an example of Reference Shape, and FIG. 3 shows an example of Target. In FIG. 2, 21 indicated by ○ represents a terminal node of the graph, and 22 indicated by □ represents a non-terminal node. Further, as indicated by 23 in FIG. 2, a line connecting nodes is defined as an edge. Also for Target in FIG. 3, ◯ and □ shown in 24 and 25 are the same terminal nodes and non-terminal nodes as shown in FIG. 2, and 26 represents an edge. Thus, Reference Shape and Target are originally composed of a character area and a background area, but in addition to this, a graph representing the character skeleton of the character area is added. The character skeleton of Reference Shape and its undirected graph structure also correspond to the skeleton of Reference shown in 1 of FIG.

FIG. 4 shows a processing flow of the present invention. FIG input shown in 13 of 4, Target, Reference, a third image of Reference Shape, 14 in FIG. 4 is a character skeleton G _R of Reference Shape and Reference with graph structure acquired after thinning processing, FIG. 4 15 is a character skeleton G _T of Target obtained in the same manner.

After acquiring the character skeleton, the correspondence between edges constituting each graph is determined. That is, assign one of the edges constituting the G _R ambassador edges constituting the G _T. Allocations may overlap. 16 and 17 in Fig. 4 are examples in which the left-paid edge of Reference Shape “Ten” shown in 14 is decided to be assigned to the left-paid edge of Target “Large”. Put it on. The state in which the edges corresponding to 16 and 17 are associated with the character areas around the edges (areas shown in gray) is shown. 18 in FIG. 4 is Reference, and 19 in FIG. 4 is the corresponding character area (left-handed part of “heaven”) on 18 Reference is extracted as a texture. The state of pasting is shown.

  In this way, the assignment of the reference shape edge corresponding to each edge of the target, the association of the character area around the edge to which the reference is assigned, and the texture pasting are executed for all the edges constituting the target, and the output font (20 output in FIG. 4) is obtained.

  Next, a specific configuration of the above outline will be described. When the outline shown in FIG. 1 is configured as hardware, for example, it is configured as shown in FIG. That is, Target, Reference, and Reference Shape are each taken as an image into the computer 6 by the scanner 5 or the like, processed by a program inside the computer 6, and Output is outputted as an image to the display, the printer 7, or the like.

  A flowchart of a program executed in the computer 6 is shown in FIG. The outline of each process will be described along the process flow of this flowchart.

First, as shown in step 8, the program receives three images of Target, Reference, and Reference Shape as inputs, and executes the thinning processing (for example, thinning of Hilditch) shown in step 9 for two images of Reference Shape. . As a result, as shown in step 10, the character skeletons of Target and Reference Shape are acquired as graph structures G _T and G _R , respectively. The graph structure here refers to a character skeleton expressed as a set of nodes and edges. The character skeleton G _R of Reference Shape is set to handle even a character skeleton of Reference.

Next, initialization is performed with j = 1, and as shown in step 11, the reference edge e _j ^R is assigned to the j-th edge e _j ^T of the target. When the edge assignment is determined, attached by processing the edge e _j ^R near texture Reference ambassador j th edge e _j ^T a character peripheral region of the Target as shown in step 12. This assignment, to end the repeated program the processing of texture pasting only the total number of edges M _T worth of Target.

The details of the processing in steps 10 to 12 among the above-described steps will be described below.
(Process of Step 10)
In Step 10, the Reference Shape and the Target each acquire the character skeleton by thinning processing (for example, Hilditch thinning method) (see FIGS. 2 and 3 for examples of the graph structures G _R and G _T ). Graph G _R, edges constituting the G _T, for each definition of a node is as follows. As illustrated in FIGS. 2 and 3, G _R and G _T are composed of a plurality of nodes and edges connecting the nodes.

Here, as shown in the formula (1) into equation (2), represents G _R the set of nodes with the V _R, the set of nodes with the G _T and V _T. Also as is the equation (1) into equation (2), i-th nodes constituting the V j ^_R, V _j ^T each G _R, and j-th nodes constituting the G _T.

As shown in Expression (3) and Expression (4), two types of terminal nodes (tn) and non-terminal non-terminal nodes (ntn) are set for each node. Furthermore, a set of edges of G _R and G _T is expressed as E _R and E _T as shown in Equations (5) and (6), respectively, and as shown in Equations (5) and (6), G i th edge constituting the _R, the j-th nodes constituting the G _T respectively represent a e _{j ^R,} e _j ^T.

  Also, as shown in Expression (7) and Expression (8), three types of attributes (Style, Direction, Length) are set for each edge. Here, Style represents the edge type (tn or ntn), Direction represents the approximate vector of the edge, and Length represents the length of the edge.

  As shown in Expression (9), there are three types of edge style: an edge connecting tn and tn, an edge connecting tn and ntn, and an edge connecting ntn and ntn. In order to handle undirected graphs, (tn, ntn) and (ntn, tn) are not treated as the same Style.

The edge approximation vector Direction is an edge approximation vector and represents an attribute for evaluating the edge direction. _Here, (see equation 12 for the notation) a Direction of _e ^{j R} _e ^{j R} _vector, representing the Direction of _e ^{j T} a _e ^{j T} vector (see equation 12 for the notation). Edge the length Length is representative of the length of the edge, where the number of pixels constituting the _edge, the Length of ^{_{e j R | e j R |}} , the Length of _e ^{j T} _{| e} ^{j T} _| represent .
(Step 11)
FIG. 7 shows a conceptual diagram of the edge assignment process shown in step 11. An e _j ^R to be allocated is determined for the j-th edge e _j ^T of the target to be processed. First, the similarity between the Style of Style and all references edges e _j ^T evaluated by the Hamming distance. Equation (10) is set with tn = 0 and ntn = 1 in order to evaluate the Style by the Hamming distance. e _j ^T of Style to the e _j ^R as the Hamming distance minimum only (the Hamming distance smallest e _j ^R sometimes there are a plurality) and subsequent processing of the subject.

Chosen for that _e ^{j R,} the Direction and Length evaluated using equation (11), assign _e ^{j R} the evaluation value becomes the maximum _e ^{j T.} In formula (11), r _ij and l _ij are defined by formula (12) and formula (13), respectively, and w _r and w _l satisfy the relationship of formula (14).

  FIGS. 8 and 9 show the processing of FIG. 7 in a flowchart. FIG. 8 corresponds to a portion for performing the similarity evaluation of Style.

First, as the initialization, the i indicating a Reference number of the Shape to 0, for recording the minimum value of Hamming distance implemented as similarity evaluation to D _h, initialized at the third maximum value or values obtained taking the D _h To do.

_{Then, e} ^j the Style of similarity between ^T and _e ^{j R} assessed by criticism Hamming distance, and records the distance d _i as Hamming distance against _e ^{j T} of _e ^{j R.} d _i is repeated a value if smaller d _i than the current minimum value D _h the above processes to be assigned to D _h only total number of edges M _R min of Reference Shape.

FIG. 9 is a continuation of the flowchart shown in FIG. After performing the Style similarity evaluation, initialize as i = n = 1, max = -1. i indicates the reference shape number, n is a variable that records i that maximizes the evaluation value of equation (11), and max is a variable that records the maximum evaluation value of equation (11). After initialization, e _j ^R of e _j ^T d _i which was recorded as the Hamming distance against it is checked whether or not matching the minimum value D _h, only the object of subsequent processing that matches the minimum value D _h .

_Next, Length between _e ^{j T} and _e ^{j R,} the evaluation of the Direction performed by equation (11), recording the evaluation value dmax. If dmax is larger than the current maximum value max, it is recorded in dmax, and the reference shape number i at that time is recorded as n.

By repeating the above processing only the total number of edges M _R min of Reference Shape, ambassador j th edge e _j ^T of Target, acquiring Reference Shape of edge number assigned as n. That is, e _i ^R (i = n) assigned to e _j ^T is determined.
(Process of Step 12)
A flowchart of the texture processing shown in step 12 is shown in FIG. Moreover, the conceptual diagram is shown in FIG.

e _j ^T affixed processed text region near e _j ^T cut texture patch group from the character region of Reference of fixed e _j ^R around the assigned to.

First, sampling points are arranged on the edge of the target. The total number is set to | e _j ^T | +1, and the sampling points are arranged at equal intervals. Sampling points are similarly arranged in Reference Shape and Reference, but the total number is set to | e _j ^R | +1, and sampling points are arranged at equal intervals. Reference numerals 28 and 32 in FIG. 11 denote one of sampling points arranged on each edge (27, 31). Also, m represents the mth sampling point of e _j ^T , and is initialized as m = 0.

The k th sampling point of e _j ^R corresponding to the m th sampling point of e _j ^T is determined by Equation (15). In the example of FIG. 11, the mth sampling point is 32 and the kth sampling point is 28.

a rectangular region around the k-th sampling point e _j ^R is extracted as a texture patch P _k ^R of Reference. In the example of FIG. 11, 30 reference rectangular areas corresponding to 29 rectangular areas indicated by Reference Shape are extracted as P _k ^R.

P _k ^R is subjected to enlargement / reduction rotation processing by Expression (16), and pasting processing is performed in accordance with the shape of the rectangular region centered on the m th sampling point of e _j ^T. For the pasting process, texture synthesis is used to seamlessly obscure joints between texture patches. According to the size of the rectangular region 33 of FIG. 11 deforms by scaling rotating handle 30 of P _k ^R, affixed. Note that α in equation (16) is defined by equation (17). D _{m ^T,} d _k ^R each m-th sampling point in expression (17) represents the shortest distance from the k-th sampling point to the ringleader of the character region and a background region, which is an index indicating the thickness of the character.

By repeating the above process only, the total number of sampling points of the target | e _j ^T | +1 is repeated repeatedly to complete the texture pasting process for e _j ^T.

  The brush font creation method according to the embodiment described above has the following characteristics.

  First, the actual brushstroke character (a work image of a famous calligrapher, etc.), which is a reference example, is used, and the blurring feature that it has is reflected in the output. As a result, the user can visually check the characteristics of the reference case when he / she performs the blurring process on the target brush font, so the design of the output font can be predicted and the design becomes easy.

  In order to add the blurring feature of the reference case to the target font, the skeleton structure of both the reference case and the target font is acquired. This skeletal structure has information on the direction of the brush and the thickness of the characters. This skeleton is represented by a graph structure, the edges of the graphs of both the reference case and the target font are associated with each other, and the blurring feature around the edge of the corresponding reference case is added to the edge of the target font.

  The addition of the blur blur feature is realized by texture synthesis (seamless texture patch application). The texture patch group used in this texture synthesis is obtained by disassembling the reference case brush characters along the skeleton structure. Further, the order relation of the decomposed texture patches is held as an order relation continuous along the skeleton, and the texture patches are assigned and pasted to the edges of the corresponding target font while keeping the order relation. As a result, the blur bleeding process is realized without impairing the blur bleeding characteristics of the reference example.

  FIG. 12 shows an input / output example of the present invention. 34 is Target and 35 is Reference. Reference Shape is omitted because it is created from 35. 36 is an output font example created by the present invention, and 37 is an output example created by the conventional method.

Claims (2)

A brush font creation method for creating a brush font by adding a blurring feature of a brush stroke character as a reference example to a target font to be processed,
Obtaining a character skeleton of both the reference case and the target font as a graph expressed as a set of edges connecting nodes.
Associating edges of graphs of both the reference case and the target font;
And a step of applying a blurring feature around the edge of the reference case corresponding to the edge of the target font. A program for causing a computer to execute a brush font creation method for creating a brush font by adding the blurring feature of a brush brush character as a reference example to the target font to be processed.
Obtaining a character skeleton of both the reference case and the target font as a graph expressed as a set of edges connecting nodes.
Associating edges of graphs of both the reference case and the target font;
A program for causing a computer to execute the step of pasting the blurring feature around the edge of the reference case corresponding to the edge of the target font.