JPH0613212B2 - Character image data processing method - Google Patents

Description

Description: TECHNICAL FIELD OF THE INVENTION The present invention relates to a method for processing character image (hereinafter referred to as “character”) data, and in particular, the contour of a character developed on x, y coordinates is represented by a variable x. Is divided into a plurality of blocks [P ₁ , Pn] of a single-valued function, data for specifying the shape of each block is created, and a set of this block data is stored as compressed data of one character. The present invention relates to a method of processing character image data which is compressed.

[Background of the Invention] It is well known that binary data obtained by dot decomposition of characters is data with extremely high redundancy. Therefore, various data compression methods have been conventionally proposed in order to reduce the redundancy.

One of them is a data compression method, which is a so-called contour method, in which the shape of a character is grasped by a contour and data for specifying the contour is stored so as to compress the data amount.

As a data compression method by the contour method, a straight line (vector) approximation method as shown in FIG. 1 and an nth-order curve approximation method as shown in FIG. 2 have already been proposed.

The linear approximation method illustrated in FIG. 1 is disclosed in JP-A-54-1495.
22 and JP-A-55-79154, the outline of which is to approximate the outline 1 of an arbitrary character shown by a dotted line with a set of vectors 2 shown by a solid line, and calculate each vector. Data can be compressed by specifying the specified information (starting point position, length and inclination, or horizontal / vertical direction components) as encoded data.

Further, the n-th order curve approximation method illustrated in FIG. 2 is disclosed in Japanese Patent Application No. 55-116160 (Japanese Patent Publication No. 62-33948).
No.), the outline of which is to compress the amount of data by storing the coordinates of a group of points P set appropriately on the contour of an arbitrary character, and to calculate an arbitrary continuous (n + 1) points. It is a set of n-th order curves 3 (n = 2 in the figure) that are connected to each other to approximate a desired contour.

These contour-based data compression methods deal with reproduction of character images of various magnifications by performing interpolation processing, thinning-out processing, or vector magnification conversion processing when decoding the compressed data and reproducing the character image. It has the feature that it can.

[Problems of background art] However, on the other hand, in these conventional methods, for example, the end point P of each vector in FIG. 1 or the left and right centering on the connection point Pc of each n-th degree curve 3 in FIG. As is clear from the fact that the tangent inclination angle δ becomes discontinuous in any case, an essential defect that the optimum result is not guaranteed for the smoothness of the contour (continuity of the contour inclination). Had.

On the other hand, in general, the contour shape of a character has a linear portion and a curved portion, and not only the contour portion itself is continuous, but also the intersection portion of character drawing lines and ""
The first derivative (contour slope) has a characteristic that it continuously changes, except for singular points such as the tip of.

Therefore, in the conventional data compression method based on the contour method, not only is it difficult to obtain compressed data that faithfully specifies the character contour, but also unnaturalness (discontinuity of inclination) of the character image reproduced based on the data is removed. I had the problem that I could not.

In order to solve such a problem, the applicant of the present invention has filed Japanese Patent Application No.
We have already applied for the system of 7-16884 (Japanese Patent Laid-Open No. 58-134745).

However, since the data compression method disclosed herein attempts to approximate all the sample sections in one arbitrary block at a time, in the portion where the curve portions are connected to each other, there is a deviation from the contour in the vicinity of the connection point. Roughness is likely to occur. In addition, since the portion that should be reproduced as a straight line is approximated by the influence of the approximate curve of the sample section before and after that, the reproducibility of the straight line is deteriorated.

Furthermore, it is also clear that when a sample (sample) point is newly established because each sample section is tried to be approximated at once, the approximation curve of the other sample (sample) section also changes due to this. became. Then, each time a sample (sample) point is newly established, it is necessary to re-calculate the approximate curves for all sample (sample) sections on any one block, and moreover, complicated calculation is required for encoding, so that the data to be calculated is required. There was also the problem that it took time to create.

[Object of the Invention] The present invention has been made in view of the above-mentioned problems of the prior art. Data having a small deviation from a desired contour and capable of smoothly reproducing the contour and having a high compression rate can be obtained. It is designed to be able to calculate at high speed.

In order to achieve such an object, the present invention provides x, y
The character image contour developed on the coordinates is divided into blocks [P ₁ , Pn] (P ₁ , Pn are start and end points of arbitrary 1 block) of a monovalent function having x as a variable, and the contour shape of each block In the method of processing character image data in which block data for specifying a block data is created and the set of block data is stored as compressed data of one character image, m contours forming blocks of the divided contours. Point Q
_For a contour section whose start point and end point are any one of _j (j = 1 to m), the deviation amount from each contour point in the contour section is equal to or less than a predetermined allowable value, and the length li is maximum. To replace the vector with a vector L i, a process of comparing the length l i of the vector with a preset length L, and a crossing angle θ i between the vector and a vector adjacent to the vector and a preset angle θ. In the process of comparing and when the comparison result is i> L, the contour section is identified as a straight line portion, and an n-th order polynomial f (x) that linearly approximates the contour section based on the coordinates of the start point and the end point of the contour section In the process of obtaining (n = 1), and when the comparison result is li ≦ L, the contour section is identified as a curved line portion,
Further, when θ ≦ θ _i, the above-mentioned comparison is repeated for the next vector, and when l _i > L, or θ _i <
When θ is reached, a process of identifying the contour section up to that point as one continuous curve section, a step of setting a sample section in the contour section of the curved section, and coordinates of the start point and end point of the set sample section Based on the inclination, a process of obtaining an n-th order polynomial f (x) (n = 2, 3) that approximates the sample section with a curve, starting point coordinates of a straight line portion existing in the arbitrary 1 block, and a curve portion The starting point coordinates of each set sample section, and the coefficient and order of an n-degree polynomial approximating each sample section of the straight line portion and the curved line portion are obtained for each block and stored as compressed data of one character image. Is characterized by.

[Outline of the Invention] An outline of data processing according to the present invention will be described with reference to the flowchart shown in FIG.

The input character image is dot-decomposed in an x, y matrix (30), and the contour of the dot-decomposed character image is extracted (3).
1). The extracted contour is linearly approximated by a vector such that the deviation amount from the contour is less than the allowable error (32). The contour that is approximated to a straight line is distinguished into a contour that is imitated as a straight line portion and a contour that is imitated as a part of a curved portion based on the length of the vector.
(33). Further, points for dividing the contour are obtained based on the intersection angle of the adjacent vectors (34). For the straight line portion, the section is defined as a straight line sample section and is represented by an n-th degree polynomial (n = 1). For the curved part, an arbitrary plurality of sample sections are set in the section, and the contour shape in the section is defined by an nth degree polynomial (n =
As a preparation for curve approximation in (2, 3), first, the slope at each contour point forming the contour is obtained (35). The n-th degree polynomial approximating the sample section is obtained from the coordinates of the contour points corresponding to the start point and the end point of the sample section and the previously obtained inclinations there. Then, while extending the section, a sample section having the longest deviation amount between an n-th degree polynomial approximating the section and the contour within a permissible error is obtained, and the start point and the end point of the section are determined as sample points. Do (36).

In this way, the sample sections are sequentially obtained, and when the n-th degree polynomial that approximates each sample section is obtained, then the first derivative of each n-th order polynomial is obtained, and the slope of each contour point is calculated again (37). Then, sample points are determined in the same manner as described above on the basis of the newly obtained inclination and coordinate value of each contour point, and an n-th degree polynomial more faithful to the contour is calculated (38). Next, the coefficient, the degree, etc. in the n-th degree polynomial thus obtained are coded (39). Furthermore,
If each block data is stored (40) in the order in which the distance between the start point and the end point of one block is long in the x direction, the subsequent character contour can be efficiently restored. Then, the stored compressed data of the character contour is subjected to distributed decoding (41) by a plurality of decoders, and a desired character image is restored (42) based on the decoding result.

[Embodiment of the Invention] Next, the data processing of each unit described in FIG. 3 will be described in detail.

[Image input (30)] x, y character image by raster scanning of a scanner device or the like.
The dot pattern is decomposed into a matrix, and the bit pattern data obtained by this is supplied as the character data to be processed.

[Contour extraction (31)] The dot position at which binary data corresponding to the decomposed character data changes from "0" to "1" or from "1" to "0" in the x or y direction (contour point Q ), The contour is obtained. Then, the obtained contour is divided into blocks of a monovalent function having x as a variable to form a set of a plurality of blocks.

FIG. 4 is an explanatory diagram when one character is expressed as a set of blocks thus obtained, and “◯” is the start point and end point of each block.

[Linear approximation (32)] In any one block on the contour, linear approximation is performed by a large number of vectors set so that the length li thereof becomes maximum within a range in which the deviation amount from the contour falls within a predetermined allowable error. To do.

FIG. 5 shows an arbitrary block [P ₁ , Pn] (P ₁ , Pn
n is the straight line approximation of the character outline 50 shown by the dotted line with a set of vectors 51 at the start and end points of an arbitrary block. Connection point P ₁ of each vector 51,
P ₂ , ..., P _n−1 , Pn are stored in the sample point coordinate storage unit 60, which will be described later, as sample (sample) points.

[Identification of Linear and Curved Part of Contour (33)] As described above, the contour shape of a character generally has a linear part and a curvilinear part. In one case, since it was attempted to approximate the sample intervals in one block at a time, many sample points near the connection point are set in the part where the curve part and the curve part are connected. Since it has to be set and divided into many approximate expressions to be expressed, the amount of data in this portion increases, and the compression rate decreases.

To deal with this, [Identify straight and curved parts of the contour (3
[3]], first identify whether the contour shape is a straight line portion or a curved line portion based on the vector length li obtained by the above-described straight line approximation, and process the identified straight line portion and curved line portion separately. Now, it is possible to faithfully reproduce even characters with straight and curved contours.

FIG. 6 is a block diagram showing an embodiment of a configuration for realizing a method of identifying a straight line portion and a curved portion of a character contour. In the figure, 60 is a sample point coordinate storage unit that stores the sample points (connection points) P obtained by the straight line approximation (32) in block units, and 61 is the straight line identification vector length that presets the vector length L to be identified. A setting unit 62 is a curve division point identification angle setting unit 63 that sets an angle θ for identifying a point that divides the curve.
Is the coordinate of the i + 1-th sample point P _{i + 1} , that is, the next sample point for holding the coordinates (x _{j + s} , y _{j + s} ) of the contour point Q _{j + s} corresponding to the point P _{i + 1.} Coordinate register, 64 is i-th sample point P
The current sample point coordinate register for holding the coordinates (x _j , y _j ) of the contour point Qj corresponding to i, and 65 is the coordinate of the contour point Q _ju corresponding to the _i− 1th sample point P _i−1 ( x _ju ,
y _ju ), the pre-sample point coordinate register, 66
Is a vector length calculation unit for calculating the vector length li of the sample section [xj, x _{j + s} ] from each seat of the next sample point coordinate register 63 and the current sample point coordinate register 64, and 67 is the next sample point Coordinate register 63, current sample point coordinate register 64
And a vector-to-vector angle calculation unit that calculates the vector-to-vector angle θi at the sample point Pi from the coordinates of the previous sample point coordinate register 65, and 68 is the vector length L set by the straight line identification vector length setting unit 61. A vector length comparison unit for comparing the vector length li calculated by the calculation unit 66, 69
Is the angle θ set by the curve division point identification angle setting unit 62.
And an inter-vector angle θi calculated by the inter-vector angle calculation unit 67, the angle comparison units 70 and 71 store the straight line section or the curved line section based on the comparison result of the vector length comparison unit 68. A straight line storage unit and a curved line storage unit, and 72 is a curve division point coordinate storage unit that stores the i-th sample point Pi as a curve division point when θ> θi in the angle comparison unit 69.

Next, the operation will be described. First, the straight line identification vector length setting unit 61 sets the identification vector length L, and the curve division point identification angle setting unit 62 sets the angle θ. Next, the coordinates of the starting point P ₁ of any one block [P ₁ , Pn] are sent from the sample point coordinate storage unit 60 to the next sample point coordinate register 63. At this time, since nothing is stored in the current sample point coordinate register 64, the vector length cannot be obtained by the vector length calculation unit 66 described later. Next, the sample point coordinates stored in the next sample point coordinate register 63 are transferred to the current sample point coordinate register 64, and the coordinates of the next sample point P ₂ are newly stored in the next sample point coordinate register 63. After that, the sample point coordinates of the current sample point coordinate register 64 are shifted and stored in the previous sample point coordinate register 65, and the sample point coordinates of the next sample point coordinate register 63 are shifted and stored in the current sample point coordinate register 64, respectively.
The next sample point coordinate register 63 stores the next sample point coordinates from the sample point coordinate storage unit 60.

The vector length calculation unit 66 uses the next sample point coordinate register 6
The coordinates (x _{j + s} , y _{j + s} ) of the i + 1-th sample point P _{i + 1} of 3
From the coordinates (x _j , y _j ) of the i-th sample point Pi of the current sample point coordinate register 64, To calculate. The vector length comparison unit 68 compares the calculated vector length li with the identification length L preset by the straight line identification vector length setting unit 61. Here, when li> L, the section of [Pi, P _{i + 1} ] approximated by the vector is determined to be a straight line portion, and the start point coordinates and end point seats of the section are set to 1
The information about one straight line portion is stored in the straight line portion storage unit 70.

When li ≦ L, the section [Pi, P _{i + 1} ] is determined to be the curve portion, and the coordinates of the start point and the end point of the section are temporarily stored in the storage unit 71 as information about the curve portion, The sections [P _{i + 1} , P _{i + 2} ] are sequentially identified in the same manner.

In this way, the data of the straight line portion and the curved line portion that form _one arbitrary block [P ₁ , Pn] are stored in the respective storage units 70 and 71, and thereafter, similar processing is repeated to form the following characters. Identify all blocks.

[Division of contour (curved portion) (34)] On the other hand, in the inter-vector angle calculation unit 67, the sample points P _i- are respectively obtained from the next sample point coordinate register 63, the current sample point coordinate register 64, and the previous sample point coordinate register 65. ₁ , Pi, P _{i + 1}
Coordinates (x _ju , y _ju ), (x _j , y _j ), (x _{j + s} ,
y _{j + s} ) is read out, and the inter-vector angle (crossing angle between the vector and vectors adjacent to the vector) θi shown in FIG. 5 is calculated. The calculated inter-vector angle θi is calculated by the angle comparison unit 69 in the curve division point identification angle setting unit 62.
Compare with the preset angle θ. Here, when θ> θi, a curve division signal is generated, and the coordinates are stored in the curve division point coordinate storage unit 72 with the sample point Pi as a new curve division point.

Normally, there are many sample points near the curve dividing point, such as the intersection of the character image described above, and it took time to process the approximated curve to be obtained. By dividing the contour at the dividing points, the curve is individually approximated before and after the curve dividing point, so the processing time is fast,
Moreover, the approximate curve can be easily obtained.

FIG. 7 is an example in which the above-described [straight line approximation (32)] to [contour division (34)] processing is performed on the character contour shown in FIG. Is shown.

In the figure, ◯ is the start and end points of one block, Δ is the sample point obtained by linear approximation, and ● is the curve dividing point. Also, * indicates a straight line portion, and no mark indicates a curved line portion.

[Calculation of Inclination of Each Contour Point (35)] The contour shape of the above-obtained curve portion is appropriately divided into one or a plurality, and these are approximated using n-th degree polynomials (where n = 2 and 3), respectively. Each n-th degree polynomial is uniquely determined if the coordinate value and the slope of two points are determined. Therefore, first, the inclination at each contour point on the contour is obtained.

As for the inclination at each contour point, a predetermined number of contour points before and after the contour point may be extracted, and the inclination of a line segment connecting the contour point for calculating the inclination and each extracted contour point may be obtained.

Next, the calculation method will be described according to the example of the curved portion in FIG. First, in order to obtain the slope t ₁ at the starting point P ₁ (contour point Q ₁ ) of an arbitrary block [P ₁ , Pn] on the contour as shown in FIG. 8A, the contour existing after the starting point P ₁ is determined. An arbitrary number of points are extracted, the inclination of the line segment connecting the starting point P ₁ and the extracted contour point is calculated, and the starting point P ₁ is calculated from the inclination of each of these line segments.
The slope t _{1 of} is calculated by the formula described later.

For example, when two gradient points on the rear side of the contour point Q ₁ whose gradient is to be calculated are extracted and the gradient is calculated, first, the gradient m _{1 of} the line segment Q ₁ Q ₂ and the line segment Q ₁ Q ₃ are calculated. Find the slope m ₂ .

The slope m ₁ of the line segment is the coordinates (x ₁ , y ₁ ) of two points,
From (x ₂ , y ₂ ), Can be found at. In addition, in the case of m _2, the same is obtained.

When each line of slope _m 1, _{m 2} is obtained, the slope _{t 1} at the contour point _{Q 1} is, Determined by. It should be noted that the end point Pn (Qn) can be obtained in the same manner as the start point P ₁ (Q ₁ ).

Next, the case of obtaining the slope t ₂ at the _second contour point Q ₂ will be described with reference to FIG. In this case, since there is only one point of Q _{1 on} the front side of the contour point Q ₂ , one point before and after is extracted, the inclinations of the line segments connecting each contour point from the contour point Q ₂ are obtained, and the contour point Q ₂ The slope is t ₂ .
This slope t ₂ is calculated by the following equation.

If the desired number of contour points designated in this way does not exist before and after, the maximum number of contour points within the designated range is extracted to obtain the slope.

Then, the slope _{t 3} at the contour point _{Q 3} by FIG. 8 (c)
A case will be described below. The contour point Q ₃ are, there is a predetermined number of contour points before and after. Therefore, the inclinations m ₁ , m ₂ , m ₃ , m _{4 of} each line segment connecting the contour point Q ₃ and another contour point
Respectively, and the gradient t ₃ at the contour point Q ₃ is Ask more. Thereafter, the inclination at each contour point is calculated in the same manner.

[Determination of sample point (36)] When the slope at each contour point forming the curved line portion is obtained as described above, one or more sample sections are set appropriately on the contour corresponding to the curved line portion. Then, two contour points are found on the contour by the procedure described later, then an approximate curve that approximates the two contour points is found, and the deviation amount ε between the approximate curve and the contour is sequentially found for each contour point.

When each deviation amount ε thus obtained is less than or equal to the allowable error Δ, the contour point is moved forward by one step, the deviation amount ε is calculated again between the enlarged contour points, and the same processing as above is repeated. The contour is divided for each sample section while obtaining the longest section (hereinafter referred to as a sample section) within a range in which even one deviation amount ε does not cause the allowable error Δ. The n-th degree polynomial determined by the sample section thus obtained is determined as one of the approximate curves necessary for expressing the contour.

Hereinafter, the calculation method of the approximated curve of the nth degree polynomial will be described with reference to FIG. 9, and then the calculation of the deviation amount will be described.

First, based on the coordinate values stored in the straight line storage unit 70 and the curved line storage unit 71, it is determined whether the approximated section is a straight line portion or a curved line portion. Then, if the section is a straight line portion, the starting point coordinates of the section and the n-th degree polynomial (n = 1) indicating the section are calculated, and the coding described later is performed.

Next, a case where the section is a curved portion will be described.

In FIG. 9A, Q _j is the j-th sample point of interest and Q _{j + r} is an arbitrary sample candidate point. A section having both points as a start point and an end point is called a sample candidate section [Qj, Qj _{+ r} ].

The sample point Qj is set as the starting point of the curved portion as an initial condition, and the sample point candidate Q _{j + r} is set at a position separated from the sample point Qj by r contour points.

Therefore, first, these sample points Qj and sample point candidates Q
_{Based on} the coordinates (x _j , y _j ) and (x _{j + r} , y _{j + r} ) of _{j + r} , and their inclinations t _j and t _{j + r} , an nth-order polynomial f (x ) And the order to obtain an approximate curve. Further, it is determined whether or not each deviation amount ε between all contour points existing in the sample candidate section [Qj, Q _{j + r} ] and the approximate curve obtained by the n-th degree polynomial is less than or equal to a predetermined allowable error Δ. Compare
If ε ≦ Δ in all cases, it is judged that the obtained approximated curve is acceptable.

By the way, the longer the section that can be represented by one approximate curve, in other words, the smaller the number of approximate curves required to represent a given arbitrary contour, the smaller the amount of data and the more efficient compression. You can

Therefore, a sample candidate section [Qj, which is determined by using the initially set arbitrary contour point Q _{j + r} as a sample point candidate
Q _{j + r} ], the deviation amount ε between the approximate curve obtained as described above and each contour point of the sample candidate section [Q _j , Q _{j + r} ] is calculated and compared with the allowable error Δ. If all the results are ε ≦ Δ, the current sample point candidate Q _{j + r} is extended by one and the adjacent contour point Q _{j + r + 1} is used as a new sample point candidate for the comparison.

After that, as long as all the comparison results are ε ≦ Δ, the sample point candidates are sequentially updated and the comparison is repeated. When the contour point Q _{j + r + p} is set as a new sample point candidate, any one of the deviation amounts ε When ε> Δ, the contour point Q immediately before that
_{With j + r + p-1} as the second sampling point, the sampling interval [Q
j, Q _{j + r + p-1} ]. This sample section [Q
j, Q _{j + r + p−1} ], is determined as one of the approximation curves required to represent a given arbitrary contour.

On the other hand, the first set sample candidate section [Qj,
Q _{j + r} ], and the sample candidate section [Q
j, Q _{j + r} ], the displacement amount ε of each contour point is calculated. If even one result is ε> Δ, the sample candidate section is reduced by one and the contour point Q _{j + r- The} above comparison is performed with ₁ as a new sample point candidate. After that, the sample candidate sections are successively reduced until the comparison results are all ε ≦ Δ, and the above comparison is repeated. When all ε ≦ Δ at the contour points Q _{j + rp} , the sample point candidates Q _{j + rp} are A sample section [Qj, Qj _{+ rp} ] is set as the second sample point. The approximated curve in this sample section [Qj, Qj _{+ rp} ] is determined as one of the approximated curves necessary for expressing a given arbitrary contour.

By such processing, the contour is divided into sample sections while determining the sample section where the displacement amount ε is the longest within the range of the allowable error Δ. Then, the contour is approximated by the nth degree polynomial determined in this sample section.

As another embodiment, the sample candidate section [Qj, Q
_{[j + rp} ] exceeds the allowable error, Q _{j + r + p} is tentatively stored as the sample candidate point at that time, and the deviation amount ε of the deviation amount ε is further stored for sample candidate sections up to several points ahead. Make an evaluation. If the allowable error is exceeded even if a predetermined number of points are reached, the sample section may be determined using the Q _{j + r + p−1} as the sample point.

Of course, in this case, when the condition of the allowable error is satisfied in the sample candidate section several points ahead, the candidate section is updated to a new sample section, and the evaluation process is repeated while advancing the sample point from there. Run. By performing such prefetching, the sample section can be further lengthened and the data compression rate is improved.

Next, a method of calculating the deviation amount will be described.

In FIG. 9 (a), 90 is a contour, B is a contour point on the contour 90, 91 is an approximate curve of the n-th degree polynomial f (x), and 92 is a starting point Qj of the sample candidate section [Qj, Q _{j + r} ]. And the end point Q
_A straight line connecting two points of _{j + r} , 93 is a perpendicular line from the point B to the straight line 92, C is an intersection of the approximated curve 91 and the perpendicular line 93, B
A is the deviation amount εx between the contour 90 and the approximate curve 91 at the point B in the x direction, BC is the deviation amount ε between the contour 90 and the approximate curve 91, and BD is the deviation between the contour 90 and the approximate curve 91 in the y direction. The amount is εy.

Hereinafter, a case of a cubic curve will be described as an example. The gradients t _j , t _{j + r} and the coordinate values (x _j , y _j ), (x _{j + r} , y _{j + r} ) of the already calculated starting point Qj and sample candidate point Q _{j + r} are converted into the following cubic expression. substitute.

f (x) = y _j + b _j (x−x _j ) + c _j (x−x _j ) ² + d _j (x−x _j ) ³ b _j = t _j When each coefficient for determining the approximate curve f (x) is obtained as described above, the deviation amount ε between the approximate curve f (x) and each contour point is obtained.

In the figure, the deviation amount ε (= B of the contour 90 and the approximate curve 91)
In obtaining C), the present invention enables high-speed processing by approximating the deviation amount ε.

That is, the displacement amount ε is the distance between BCs, and is the amount of separation between the contour point and the approximate curve. Usually, when measuring the distance between a point and a curve, it is measured in the direction orthogonal to the tangent to the curve. When a plurality of points on a curve exist in the normal direction, the shortest distance among them is generally the distance between the point and the curve.

Similarly when measuring the distance ε between the nth-order curve and the contour point as in the present application, first, the equation of the normal line orthogonal to the tangent line of the nth-order curve and the equation of the straight line passing through the contour point are made simultaneous. The measured point on the nth degree curve must be determined by solving the above equation and then the distance between that point and the contour point of interest must be calculated. A huge amount of calculation is required to carry out the above calculation.

Therefore, in the present invention, the deviation amount ε is obtained by the following approximate calculation.

FIG. 9B shows the contour 90 and the approximate curve 9 in FIG.
The displacement amount ε with 1 is enlarged. In the figure, it is assumed that the line segment CA is parallel to the straight line 92, and the slope m of the line segment CA is the slope of the straight line 92. Then, the amount of separation εx between the contour point B and the approximated curve 91 in the x direction and the y direction,
By obtaining εy, the desired deviation amount ε is Can be approximately calculated with.

In this way, the approximate deviation amount ε and the allowable error Δ are compared for the determination, the sample point candidate is selected based on the result, the sample section is determined, and the curve portion is divided into a plurality of parts.

[Recalculation of Inclination of Each Contour Point (37)] As described above, when the contour is approximated by a cubic approximation curve in each of a plurality of sample sections, the starting point coordinates of the sample section and the approximation that approximates the sample section are approximated. A compressed code can be obtained by storing the degree and coefficient of the curve as contour data.

However, since the inclination of each contour point obtained in [Calculation of inclination of each contour point (35)] is the average value of the inclination with a predetermined number of contour points before and after that, the arrangement of the contour points is large. The inclination may differ from the actual inclination due to the fact that it changes or the data of the contour point is originally discrete digital data.

Therefore, in order to more closely approximate the continuity of the inclination of the contour, the inclination is calculated again as follows. That is, the slope obtained from the first derivative f ′ (x) of the n-th order polynomial f (x) in each sample section is set as the slope at each contour point of the sample section.

[Re-determination of sample points (38)] When the gradient at each contour point is recalculated by the above [Recalculation of gradient of each contour point (37)], the obtained gradient at each contour point is used to The same processing as [Determination of Point (36)] is performed, and the sample section is determined again while determining the sample point again. The sample section obtained again in this way is approximated by an n-th order polynomial f (x) that is more faithful to the contour.

[Coding (39)] The coordinates of the start point and the end point of one straight line portion stored in the straight line storage unit 70 are regarded as one sample section, and an nth-order polynomial f (x) that linearly approximates the section based on these two coordinates. ) (N =
The coefficient of 1) is calculated and coded. This is one encoded data.

On the other hand, in the case of a curved part, the starting point coordinates of each sample section,
Coding the coefficient of the polynomial f (x) of degree n, the degree, etc.,
Further, by storing the contour as a set of block data organized in block units, compressed data faithful to the contour can be obtained.

FIG. 10 is a diagram showing an example of a preferable data storage format of arbitrary one block data. In the format shown in the figure, the block header stores the start point / end point coordinates of one block and the number of sample sections of the straight line portion and the curved line portion existing in one block, and the segment header stores the start point X, Y coordinates of one sample section and nth degree polynomial f
The order of (x) is stored, and the segment information stores each coefficient of the n-th order polynomial f (x) obtained by the order.

Then, one sample section coded data is organized by the segment header and the segment information, and further, the sample section coded data corresponding to the number of the sample sections is sequentially arranged, and one block character data is formed as a whole. ing.

[Memory of compressed data (40)] By storing each block data coded in the above format in order of long decoding processing time in each block, for example, in units of blocks, it is possible to efficiently restore the character contour. It is possible to restore.

For example, FIG. 11A shows a case where compressed data is obtained for each block (for the sake of explanation, each block is numbered 1 to 20) for the character "A". Then, it is stored in order from the one having the longest decoding time (for example, the one having a long distance between the start point and the end point of each block in the x direction). First
In FIG. 1A, it is the block 12 that takes the longest time to decode one block.
Following block 16, block 3 is the shortest. Therefore, the contour compressed data for one character finally obtained by this example is stored as a set of block data as shown in FIG. 11 (b).

[Distributed Decoding (41)] Next, a method of restoring the original contour will be described. In this case, in order to restore the contours at high speed, the block data for each block is sequentially transferred to a plurality of decoders for processing, and the next block data is processed by the processed decoder. To

Therefore, as shown in FIG. 11 (b), compressed data is stored in order from the one having the longest time for decoding one block, and processed by a plurality of decoders, whereby the decoder can be used efficiently.

Hereinafter, description will be given with reference to FIGS. 12 and 13. First
FIG. 2 is a block diagram showing an embodiment for optimum contour restoration. In the figure, reference numeral 120 is a compressed data storage unit that stores each block data in order from the longest time required for decoding one block, and reference numeral 121 is a selector 1 or 12 that transfers one block data by selecting a decoder described later.
2 is a magnification storage unit that stores a separately input desired magnification, 12
3 is n for restoring the block data to the contour corresponding to the magnification.
A decoder group consisting of a plurality of decoders, 124 is a decoder group 123 which selects a decoder which has been processed and transfers the decoded contour pixel data to a one-character storage unit described later. Selector 2 (12
4) a one-character storage unit for storing the contour data sent from
An output device 126 prints or displays a character based on the outline pixel data for one character which has been stored in the one-character storage unit 125.

Next, the operation will be described. Compressed data storage unit 120
The data is transferred to the selector 1 (121) in the order of longer time required for decoding one block (for example, the order of block 12, block 10, block 16 ... In FIG. 11). The selector 1 (121) transfers the transferred block data by selecting decoders in order from the decoder 1 of the decoder group 122. The decoder, which has completed the transfer of the block data, starts the contour restoration processing based on the block data and the desired magnification data separately stored in the magnification storage unit 122. The selector 2 (124) sequentially selects the decoders for which the restoration processing has been completed, and transfers the restored contour pixel data to the one-character storage unit 125. The decoder that has completed the restoration processing for one block requests the selector 1 (121) to transfer the block data, and the block data for the next one block is transferred.

FIG. 13 is a timing chart showing an example of the processing state of each decoder group 123 shown in FIG.

In the figure, T ₁ is a time for transferring block data from the selector 1 (121) to each decoder, and T ₂ is for transferring contour pixel data decoded by each decoder to the one-character storage unit 125 via the selector 2 (124). Time, and T ₃ is a block data decoding processing time in each decoder. Further, T ₄ indicates the processing time for one character.

When n blocks of block data are transferred to the decoders 1 to n and the processing is started, the (n + 1) th block data is transferred to the decoder (decoder 3 in this case) having the shortest processing time, and the subsequent processing is performed. Go.

Moreover, since each block data to be transferred is arranged in the order in which the processing time is gradually shortened as described above, the respective decoders are arranged in the same manner as shown in FIG.
Finish decoding the characters. Therefore, the decoders operate almost equally, and the decoding of the character data is completed promptly without wasting idle time.

[Contour Restoration (42)] As described above, one block of contour data restored by each decoder is sequentially stored in the one-character storage unit 125.
The contour data for the character is completed. Then, the outline data for one character stored in the one-character storage unit 125 is supplied to an output device 126 such as a laser beam printer, a CRT phototypesetting machine or a display device, and a desired character is restored.

EFFECTS OF THE INVENTION As described above, the present invention distinguishes the contour into a straight line portion and a curved line portion, and the straight line portion is represented by a straight line and the curved line portion is appropriately represented by a plurality of curve approximations. It is possible to faithfully reproduce a character image with high quality with a smaller amount of data as compared with the case of using only one of approximation and curve approximation. Further, the curve part is sequentially divided into sample sections that can be optimally approximated by a curve. An n-th degree polynomial that approximates each sample section that is appropriately determined on these curve sections is a start point and an end point of each sample section. Since the calculation is performed for each sample section in consideration of the slope of, the connection of the curved portion can be smoothly reproduced. Further, since the portion reproduced as a straight line is not affected by the approximate curves before and after it, the reproducibility of the straight line can be maintained. Further, according to the present invention, since the contour is approximated by an n-th degree polynomial, when the character is greatly enlarged and reproduced, the disadvantage of the conventional vector method that the originally curved portion is reproduced in a polygonal line is naturally solved. You can

As described above, the present invention has a great effect that there is little deviation from the desired contour, the contour can be smoothly reproduced, and data with a high compression rate can be obtained at high speed.

As a result of applying the character image data processing method according to the present invention described above to the Mincho type Hiragana character “a” consisting of 800 × 800 dots and verifying that the allowable error for a desired character image is 1 dot , The data amount could be compressed to 1.21%.

Further, in the above description, the character "a" has been described as an example, but other than that, various kanji and various marks, symbols,
It is obvious that images such as line drawings can be handled in the same way.

[Brief description of drawings]

1 and 2 are diagrams for explaining conventional contour method data compression, FIG. 3 is a flow chart showing an outline of the present invention, and FIG.
FIG. 6 is a diagram showing an example in which the contour is divided into blocks of a monovalent function with x as a variable, FIG. 5 is a diagram illustrating linear approximation, and FIG.
FIG. 7 is a block diagram showing an embodiment of a method of discriminating between a straight line portion and a curved portion, FIG. 7 is a diagram showing a discrimination result of a straight line portion and a curved portion, and FIG. Figure,
FIG. 9 is a diagram illustrating a method of determining sample points, FIG. 10 is a diagram illustrating an example of a storage format of arbitrary block data, FIG. 11 is a diagram illustrating an example of storage of block data,
FIG. 12 is a block diagram showing an embodiment for restoring block data, and FIG. 13 is a timing chart showing the operation of FIG. 1, 50, 90 ... Character outline 2, 51 ... Two-dimensional vector 3 ... m-order curve element 60 ... Sampling point coordinate storage unit 61 ... Straight line identification vector length setting unit 62 ... Curve division point identification angle setting unit 63 ... Next Sample point coordinate register 64 ... Current sample point coordinate register 65 ... Previous sample point coordinate register 66 ... Vector length calculation unit 67 ... Inter-vector angle calculation unit 68 ... Vector length comparison unit 69 ... Angle comparison unit 70 ... Straight line storage unit 71 ... Curve section storage section 72 ... Curve division point coordinate storage section 91 ... n-degree polynomial approximation curve 120 ... Compressed data storage section 121 ... Selector 1, 122 ... Magnification storage section 123 ... Decoder group, 124 ... Selector 2 125 ... 1 character storage Part, 126 ... Output device

 ─────────────────────────────────────────────────── --- Continuation of the front page (56) References JP-A-54-149522 (JP, A) JP-A-58-134745 (JP, A) "bit" Vol. 13, No. 10 (September 1981-Volume 169) (Sho 56-9-1) Kyoritsu Publishing P.P. 1218-1225 IEICE Transactions Vol. J62-D, No. 10 (Sho 54-10-25) P. 665-672

Claims (3)

[Claims] 1. A character image contour expanded on x and y coordinates,
A block of monovalent functions with x as a variable [P ₁ , Pn] (P
₁ and Pn are divided into arbitrary block start and end points), block data specifying the contour shape of each block is created, and a set of this block data is stored as compressed data of one character image. In the image data processing method, a contour section whose start point and end point are any of m contour points Q _j (j = 1 to m) forming the divided contour block A step of replacing the vector with a deviation amount with respect to each contour point equal to or smaller than a predetermined allowable value and a length l i of which is the maximum, and a step of comparing the vector length l i with a preset length l And a step of comparing an intersection angle θi between the vector and a vector adjacent to the vector with a preset angle θ, and when the comparison result is li> L, the contour section is identified as a straight line portion, and Of the contour section A process of obtaining an n-th degree polynomial f (x) (n = 1) that linearly approximates the contour section based on the coordinates of the points and the end points; and when the comparison result is li ≦ L, the contour section is identified as a curved portion. Further, when θ ≦ θ _i, the above-mentioned comparison is repeated for the next vector, and when li> L or θ _i <θ, the contour section up to that point is identified as one continuous curved line portion. A step, a step of setting a sample section in the contour section of the curved section, and an n-th degree polynomial f (x) for curve-approximation of the sample section based on the coordinates of the start point and the end point of the set sample section and their inclinations. A process of obtaining (n = 2, 3), starting point coordinates of a straight line portion existing in the arbitrary 1 block,
The starting point coordinates of each sample section set in the curved line section, and the coefficient and order of the nth degree polynomial approximating each sample section of the straight line section and the curved line section are calculated for each block and stored as compressed data of one character image. Method of processing character image data that has been set. 2. The step of obtaining an n-th degree polynomial f (x) (n = 2, 3) that approximates the curve of the sample section based on the coordinates of the start point and the end point of the sample section and the inclination thereof, Mean value t ₁ of the slopes of the line segments connecting the start point Q ₁ and the end point Qr of
And seeking _{t r,} defines it as the slope of the curve at the start point _{Q 1} and the end point Qr, an inclination _{t 1} and _{t r} in the determined start point _{Q 1} and the end point Qr, the coordinate origin _{Q 1 (x 1,} y ₁ ) and the coordinates (x _r , y _r ) of the end point Qr, the n-th degree polynomial f (x) (where n = 2, 3) of the following formula (1) is obtained, and f (x) = y ₁ + b ₁ (x-x ₁ ) + c ₁ (x-x ₁ ) ² + d ₁ (x-x ₁ ) ³ (1) where b ₁ = t ₁ Then, the slope of the contour at each contour point of the curve approximation section is recalculated based on the first derivative f ′ (x) of the n-th degree polynomial f (x), and based on the first derivative f ′ (x) Similar to the equation (1), an n-th order polynomial f (x) (where n = 2, 3) is recalculated from the calculated inclination of each contour point and the coordinate value of each contour point, and the recalculation is performed. The desired n-th order polynomial f (x)
The method of processing character image data according to claim (1), which is a process of making a polynomial of degree f (x). 3. The process of setting a sample section in the contour section of the curved line section, wherein the contour point Q _j is the first sample point, and an arbitrary contour point Q is set.
A sample candidate section [Qj, where _{j + r} is a sample point candidate]
Q _{j + r} ], and the coordinates (x _j , y) of both sample points are set.
_j ), (x _{j + r} , y _{j + r} ) and their slopes t _j , t _{j + r} , an n-th degree polynomial f for curve-approximating the sample candidate section.
(X) (n = 2, 3), the deviation amount ε between the curve approximated by the n-th degree polynomial f (x) and each contour point in the sample candidate section (However, m is the inclination of a straight line connecting the start point and the end point of the sample candidate section, and εx and εy are displacement amounts in the x direction and the y direction of each contour point existing in the sample candidate section and the approximate curve) A step, a step of comparing the deviation amount ε with a preset allowable error Δ, and the comparison result ε ≦ Δ at all contour points of the sample candidate section, the contour points adjacent to the sample point candidate Using Q _{j + r + 1} as a new sample point candidate, the updated sample candidate section [Q _j , Q _{j + r + 1} ] is compared in the same manner as described above to postpone the sample point candidate, and the contour point Q _{j + r + p}
When ε> Δ at, the new sample section [Qj, Q _{j + r + p-1} ] is set at the contour point Q _{j + r + p-1} immediately before that. The character image data processing method according to claim (1).