JP4107659B2 - Handwritten font output system - Google Patents

Description

The present invention relates to a handwritten character font output system that automatically generates a handwritten character font that reflects the individuality of each person for use in a word processor or the like.

  A method for extracting and quantifying a writer's handwritten character from off-line input user's handwritten character and generating a user's original handwritten character font (hereinafter referred to as “character font with individuality”) reflecting the character has already been disclosed in Patent Document 1. However, in this method, when outputting a character font with individuality, it is assumed that uniform width characters are output at equal intervals.

JP 2003-58142 A

  However, handwritten characters written by a person using a writing instrument such as a pen, a ballpoint pen, or a pencil vary considerably in character width (width of the character itself) and the space between characters depending on the writer.

  FIG. 3 is a diagram illustrating an example of handwritten text. FIG. 3 shows an example in which four people A to D are naturally written as “having a big influence”. Actually, the character width varies from person to person, and the way of taking the space between characters varies from person to person. This difference suggests that it is a big element of handwriting. For this reason, even if it is a handwritten-style character font, if uniform-width characters are output in a grid pattern at regular intervals, “Gikochinasa” will be conspicuous in appearance, which is unsatisfactory as a handwritten style .

In view of the above problems, an object of the present invention is to provide a handwritten-style character font output system, a program, and a recording medium that enable character segmentation to extract handwritten inter-character space information.

The handwritten character font output system of the present invention includes a first memory for storing standard m (m: natural number) character fonts, and n (n: natural number, 2 ≦ n <m) characters inputted as an image. a second memory for storing a series of fonts, a tilt correcting means for correcting the tilt of the input character font stored in the second memory, the character corresponding to the character fonts inclination correction by inclined-out correction means The character fonts are cut out from a series of character fonts that have been input by comparing the two read out from the first memory and the character fonts are cut out to extract the spacing between adjacent characters . character segmentation means for registering quantified by adjacent type of character combinations, cut out by the character segmentation unit Displacement information extracting means for comparing each character font with a standard corresponding character font stored in the first memory and extracting the displacement of the corresponding point to extract the statistical information, and the displacement information extraction Displacement information adding means for changing the character font stored in the first memory in accordance with the output of the means and outputting the character font output by the displacement information adding means by the character segmentation means Inter-character space information adding means for outputting at a character interval according to the interval.

The handwritten-style character font output system of the present invention includes a first memory for storing standard m (m: natural number) character fonts, and n (n: natural number, 2 ≦ n <m) input image. A second memory that stores a series of character fonts, an inclination correction unit that corrects an inclination of the input character font stored in the second memory, and a character font that is inclined by the inclination correction unit The character fonts to be read out from the first memory and compared with each other are extracted from the series of character fonts that have been input and corrected to extract the spacing between adjacent characters by cutting out each character font. Character segmentation means for quantifying and registering each character according to the combination of the projection complexity of adjacent characters, and the character segmentation Displacement information extracting means for comparing each character font cut out by the image means with the standard corresponding character font stored in the first memory, extracting the displacement of the corresponding points, and extracting the statistical information thereof Displacement information adding means for changing and outputting the character font stored in the first memory in accordance with the output of the displacement information extracting means; and character font output by the displacement information adding means for the character segmentation. And an inter-character space information adding means for outputting at character intervals corresponding to the character intervals registered by the means.

According to the present invention, a handwritten character can be accurately cut out even with respect to a diagonally inclined character string. Can be output.

  Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a block diagram showing a configuration of a handwritten character font output system according to an embodiment of the present invention. The handwritten character font output system 1 is used as a control unit 11 for controlling the entire handwritten character font output system 1, an input unit 12 including a keyboard for inputting characters and an image scanner for inputting character fonts, and a work area. RAM (Random Access Memory) 13 which is a memory, ROM (Read Only Memory) 14 which stores a control program and predetermined constants, a display unit 15 which is a computer display, and character fonts are printed on paper. The printing unit 16 is configured to be connected to each other. Each of these components is basically the same as each component in a normal computer, and general-purpose components can be used.

FIG. 2 is a functional block diagram showing functions of the handwritten style font output system according to the present embodiment. The functions are large and can be divided into two functions: a personality information extraction / registration function 2 and a character font generation function 3 with personality information.

  In the individuality information extraction / registration function 2, a plurality of characters (hereinafter referred to as registered characters), which are predetermined sentences written by the user on ruled line paper, are input to a computer using an image scanner, and the registered character font is used. And the standard character font are extracted, the registered characters are cut out, the interval between adjacent registered characters, the geometric displacement of the registered character font relative to the standard character font, and the size of the registered character are analyzed, and the statistical information Are registered as feature quantities (hereinafter referred to as individuality information) representing the individuality of the user.

  First, in the standard character font group storage function 21, several types of pen font fonts (hereinafter referred to as standard character fonts) are preliminarily stored in the computer for extracting registered characters, extracting user personality information, and creating character fonts with personality information. Is installed. It is better to have more types, and it is desirable to have as many variations as possible in the way the characters are broken. For example, there are a Chiba pen font, a pen line font, a Hakushu pen font, a Soei pen font, a sabishi sword, and a pen font.

  Next, in the handwritten character input storage function 22, the user writes a sentence of about 100 to 200 characters on the ruled line paper, which is input to the computer as an image and stored. Sentences to be written are predetermined. As the ruled line paper to be used, a stationery of a red ruled line (a red color is just an example and any color other than monochrome or blue) may be used (see FIG. 4). For this stationery, the user may print out and use a stationery form with built-in application software. Furthermore, when the user enters a writing sample for extracting personality information, a sentence (predetermined about 100 to 200 characters in total) with a registration mark (+) on the top is used. (Specified text) should be entered (use black or blue ballpoint pen, fountain pen or pencil). In order to reliably perform character segmentation (to be described in detail later), the number of characters of about 10 characters per line is appropriate. If the number of characters is increased, if an error occurs in the comparison between the projection of registered characters and the projection of the standard font, there is a possibility that the error is expanded to a formula that can be determined. Therefore, increasing the number of characters is not a good idea. On the other hand, it is also important to preserve the flow of letters in sentences that are written by the user. For this reason, the number of characters in one line cannot be reduced so much that the user can enter characters in a natural manner. Accordingly, the appropriate number of characters in one line is “about 10 characters”. By doing so, it eliminates the restriction that one character must be written in one square (if you write one character at a time, it will hinder the flow of writing and it will be ugly), and the continuity between characters and the writer It is possible to obtain a sentence (character group) in which unique handwriting (including the state of character width and inter-character space) is maintained. Then, the sentence written as described above is input to the computer by the image scanner and stored. As mentioned above, the registration form is stamped on the character entry form, and the ruled line has a color different from the character written by the user. Can be done. That is, no matter how the entry sheet is set on the image scanner surface, it is possible to automatically determine the orientation of the entry sheet, and it is possible to acquire character data independent of the orientation of the sheet.

  The row or column extraction function 23 extracts a row in the case of a character group written in horizontal writing or a column in the case of a character group written in vertical writing. Here, the case of a group of characters written horizontally will be described as an example.

  First, as preprocessing, binarization processing is performed on the entire acquired image. Since the character group written on the letter writing note is arranged along the red ruled line, after erasing the red ruled line, the projection is taken in the direction of the red ruled line, and the projection image (projected image) is taken. By creating, the line is cut out from the projections and depressions of the projection image.

  FIG. 5 is a diagram illustrating an example of a row projection image. An example of the acquisition image of the designated sentence written by the user on the ruled line paper is shown on the left side, and the projection image in the row direction is shown on the right side. A row can be extracted by cutting out the row at the “valley” portion of the projection image.

  The standard character font selection function 24 corrects the inclination of each character string extracted by the row or column extraction function 23, and the character corresponding to the registered character font previously installed in the computer by the standard character font group storage function 21. The character strings of several kinds of standard character fonts are compared, and the standard character font most similar to the registered character font as a whole is selected.

  First, with respect to the extracted character string in each line, character inclination detection is performed on a character string basis, and character string inclination correction is performed. Some users write diagonally inclined character strings, so that character segmentation can be performed even for diagonally inclined character strings by detecting and correcting the inclination of the character string. It is. Two methods are proposed as methods for detecting and correcting the inclination of the character string: (a) using the dispersion value of the projection in the vertical direction of the character string, and (b) using the stroke direction of the black pixels of the character string.

  First, the former method (a) will be described. A projection in the vertical direction is created for the character string in each line cut out in one line (see FIG. 6). In the created vertical projection, if the character string is not inclined, the variation in the projection value is large (the variance value is large), and if the character string is inclined, the variation in the projection value is small (the variance value is Becomes smaller). If the character string written perpendicular to the ruled line is used as the inclination of the reference character string, the inclination of the character string written by the user is within ± 45 ° from the inclination of the reference character string at most. it is conceivable that. Accordingly, the dispersion value of the projection in the vertical direction of the character string is calculated for all directions within ± 45 ° from the inclination of the reference character string, and the character string is calculated when the dispersion value of the projection is the largest. The inclination of the character string is corrected by inclining the character string in the opposite direction by the detected inclination.

  Next, the latter method (A) will be described. The stroke direction (line segment constituting the character) in all the black pixels constituting the character string is calculated with respect to the character string of each cut out line, and a stroke direction histogram is created. As for the stroke direction of the character string, when the character is written perpendicular to the ruled line, the stroke in the direction perpendicular to the ruled line is the largest. That is, the stroke direction histogram has the highest probability in the direction perpendicular to the ruled line. Therefore, the direction in which the histogram of the stroke direction of the character string written by the user has the highest value is the stroke direction of the character string. Thus, by using the stroke direction perpendicular to the ruled line as a reference, a deviation from the reference stroke direction to the stroke direction of the character string written by the user is detected as the inclination of the stroke of the character string, and thus is detected. The inclination of the character string can be corrected by inclining in the opposite direction by the amount of inclination.

  The inclination detected by the processes (a) and (b) is stored. Through the processing described above, the character string written by the user becomes a character string arranged perpendicular to the ruled line.

  Next, all the registered character strings arranged perpendicular to the ruled lines, and character strings of several types of standard character fonts which are installed in the computer in advance by the standard character font group storage function 21 and correspond to the registered character strings. By comparison, the standard character font most similar to the registered character font as a whole is selected.

  The character segmentation function 25 cuts out each character from the character string by searching for a corresponding point between the vertical projection of each character string in which the inclination of the character string is corrected and the vertical projection of the corresponding character string of the standard character font.

  Then, the vertical projection of the character string of the selected standard character font is calculated, and the vertical projection of the character string of the standard character font and the vertical direction of the registered character string written and corrected by the user are corrected. Search for corresponding points with the projection. The technique used here is by minimizing the energy functional E (DX).

E (DX) = P (DX) + λS (DX) ... (1)
P (DX) = Σ (f (x + DX (x)) - g (x)) 2 ··· (2)
S (DX) = Σ (∂DX (x) / ∂x) ² (3)
here,
P (DX): Corresponding error between projections when displacement is applied,
S (DX): corresponding smoothness,
λ: parameter for weighting,
f (x): vertical projection of a character string written by the user,
g (x): vertical projection of standard character font,
DX (x): A displacement amount in the x direction at the position (x).

  The displacement amount DX (x) at the projection position (x) is obtained by repetitively solving the Euler equation that minimizes E (DX) shown in Equation 1 (see FIG. 7). Since the standard font stored in advance already has information on where to correspond to the projection of one character, based on the information, from the displacement amount DX (x) at the measured position (x), A range in which registered characters exist, that is, a character width can be extracted. Therefore, it is also possible to extract information about the space between adjacent characters, that is, the space between characters (see FIG. 8). This is nothing other than character segmentation. The extracted inter-character space information is quantified and stored as the writer's personality.

  Actually, when this character segmentation method was applied to a large number of handwritten characters, character segmentation could be performed with a success rate of 95% or more (see FIG. 9). This shows that handwritten characters can be segmented character by character with a high success rate by this method. In the case of a character group written in vertical writing, if the same processing is performed, characters can be segmented for each character in the same way as in a character group written in horizontal writing.

  The size information extraction function 26 first restores the inclination of each registered character from the inclination of the character string stored by the standard character font selection function 24. Then, the standard character font selected by the standard character font selection function 24 and the size of the registered character font corresponding to the registered character are compared with the size of the registered character font (vertical direction and horizontal direction). Is quantified and memorized as the writer's personality. For the difference in size, the average of all the registered character fonts may be stored, or the registered character font of kanji and the registered character font of alphanumeric characters may be averaged and stored separately.

  The displacement information extraction function 27 compares the standard character font corresponding to the registered character with the standard character font selected by the standard character font selection function 24 and the registered character font, and displays the registered character font screen for the standard character font. The amount of geometric displacement (with direction and size information) at each position is extracted by two-dimensional search for corresponding points, quantified and stored as the writer's personality.

  In the character font generation function 3 with individuality information, the individuality information extracted by the individuality information extraction / registration function 2 is added to the standard character font to generate and output a character font with individuality information of an arbitrary character.

  First, the standard character font calling function 31 calls the standard character font selected in the individuality information extraction / registration function 2. Different fonts are used by users.

  Next, the keyboard input function 32 is specific to a user called by the standard character font call function 31 from a plurality of standard character fonts installed in the computer in advance according to the character code input by the keyboard which is the input unit 12. The image information of the standard character font is temporarily stored in the memory. Each standard character font contains information about all characters, like other character fonts such as Mincho and Gothic.

  The displacement information adding function 33 adds displacement information to each character of the standard character font stored in the memory. That is, the displacement corresponding to the average geometric displacement amount at each position in the user image registered in the displacement information extraction function 27 is added to the standard character font.

  The size information adding function 34 determines and outputs the size of the standard character font to which the displacement information is added by the displacement information adding function 33. In other words, the size of the standard character font to which the displacement information is added is changed at a magnification corresponding to the average size of the user's character registered in the size information extraction function 26.

  The inter-character space information adding function 35 determines and outputs the inter-character space of the standard character font whose size is determined by adding displacement information. In other words, the character segmentation function 25 opens the space between the characters registered for the user and outputs each character.

There are the following methods for adding user-specific character space information.
(1). Register the character segmentation function 25 value calculated by averaging all of the inter-character space information extracted by as a user-specific inter-character space information, output by the inter-character space information adding function 35 To do.
(2). A value calculated by averaging from all the inter-character space information extracted by the character segmentation function 25 is registered, and the inter-character space information adding function 35 outputs a change with random numbers.
(3) The character segmentation function 25 quantifies and registers the character of the space between characters for each combination of adjacent character types such as hiragana, kanji and punctuation marks. For example, the inter-character space is quantified and registered for each combination of adjacent character types such as between hiragana and hiragana, between hiragana and kanji, between hiragana and punctuation. Then, the inter-character space information adding function 35 adds inter-character spaces according to combinations of adjacent character types such as between hiragana and hiragana, between hiragana and kanji, between hiragana and punctuation.
(4) The character segmentation function 25 quantifies and registers the character of the space between characters according to the complexity of projection in the vertical direction (for example, the size of dispersion). For example, dividing the projection complexity into large, medium, and small, the space between characters according to the projection complexity of adjacent characters such as between large and large, between large and medium, between large and small, etc. Quantify and register . Then, the inter-character space information addition function 35 adds inter-character spaces according to the combination of the projection complexity of adjacent characters such as between large and large, between large and medium, and between large and small. To do.

  The individual character font output function 36 displays the individual character font to which the displacement information, the size information, and the inter-character space information are added on the display unit 15 or prints it on the printing unit 16.

  The present invention is not limited to the above embodiment.

  Although it is desirable to add both the displacement information and the size information, the personality information may be only one.

  It is desirable to be able to control the amount of personality information added according to user preferences.

  The standard character font is preferably a proportional font.

  In the above embodiment, the displacement information extraction function has been described for extracting the geometric displacement amount by two-dimensional corresponding point search. However, the displacement information extracting function extracts the displacement of the corresponding point of the registered character font with respect to the standard character font. Any method may be used.

The handwritten character font output system of the present invention is also realized by a program for causing a computer to function as the handwritten character font output system. This program may be stored in a computer-readable recording medium.

  The recording medium on which this program is recorded may be the ROM 14 itself shown in FIG. 1, or a program reading device such as a CD-ROM drive is provided as an external storage device, and the recording medium is inserted therein. It may be a CD-ROM or the like that can be read.

  The recording medium may be a magnetic tape, a cassette tape, a flexible disk, a hard disk, an MO / MD / DVD, or a semiconductor memory.

It is a block diagram which shows the structure of the handwritten style character font output system by one embodiment of this invention. It is a functional block diagram which shows the function of the handwritten style character font output system by this Embodiment. It is a figure which shows the example of a handwritten sentence. It is a figure which shows the example of the notepaper used by this Embodiment. It is a figure which shows the example of the projection image of a line. It is a figure which shows the example of the projection image of a vertical direction. It is a figure explaining the corresponding point search of a character projection. It is a figure which shows the relationship between character width information and the space information between characters. It is a figure which shows the success rate of character segmentation.

Claims (2)

A first memory for storing standard m (m: natural number) character fonts;
A second memory for storing a series of n (n: natural number, 2 ≦ n <m) character fonts that are input as images;
Inclination correcting means for correcting the inclination of the input character font stored in the second memory ;
A character font corresponding to the character font whose inclination has been corrected by the inclination correction means is read out from the first memory and compared with the character fonts. Character segmentation means for extracting an interval and quantifying and registering the extracted character interval for each combination of adjacent character types ;
Displacement information for comparing each character font cut out by the character segmentation means with a standard corresponding character font stored in the first memory, extracting the displacement of the corresponding point, and extracting its statistical information Extraction means;
Displacement information adding means for changing and outputting the character font stored in the first memory in accordance with the output of the displacement information extracting means;
Handwritten-style character font output, comprising: character space information adding means for outputting the character font output by the displacement information adding means at character intervals corresponding to the character intervals registered by the character segmentation means. system. A first memory for storing standard m (m: natural number) character fonts;
A second memory for storing a series of n (n: natural number, 2 ≦ n <m) character fonts that are input as images;
Inclination correcting means for correcting the inclination of the input character font stored in the second memory;
A character font corresponding to the character font whose inclination has been corrected by the inclination correction means is read out from the first memory and compared with the character fonts. A character segmentation means for extracting an interval and quantifying and registering the extracted character interval for each combination of projection complexity of adjacent characters;
Displacement information for comparing each character font cut out by the character segmentation means with a standard corresponding character font stored in the first memory, extracting the displacement of the corresponding point, and extracting its statistical information Extraction means;
Displacement information adding means for changing and outputting the character font stored in the first memory in accordance with the output of the displacement information extracting means;
Handwritten-style character font output, comprising: character space information adding means for outputting the character font output by the displacement information adding means at character intervals corresponding to the character intervals registered by the character segmentation means. system.

Images