JP2689380B2 - Line figure folding device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Outline] The present invention relates to a line-figure polygonalization device for polygonalizing a line figure, and when performing line-approximation to a thinned figure that has been thinned, a thinned point sequence forming an arc. In order to solve the problem that the whole line figure is undesirably distorted by connecting the points where the distance between the line segment that forms the arc and the line segment that forms an arc becomes smaller than a certain threshold value in sequence, the total length of the point sequence to be thinned Also, an optimal threshold value is determined based on the length of a line segment that connects both ends of the point sequence and the folding process is performed to generate a polygonal line with good quality. [Field of Industrial Application] The present invention is a line figure folding device configured to determine a threshold value based on the total length of a point sequence to be thinned and the length of a line segment connecting both ends of the point sequence to perform line folding. It is about. [Problems to be Solved by Conventional Techniques and Inventions] In order to automatically recognize pattern information such as documents and drawings and accurately input it to a computer, a polygonal line existing in an image is easy to handle for a computer. Need to convert to data format. Conventionally, a thinning / polygonal approximation method has been widely used as this polygonalization processing. This is because after performing thinning processing on a line figure, when performing polygonal approximation to this, the distance comparison between the thinning point sequence forming an arc and the line segment forming the chord is performed using a certain threshold value. Was there. For example, when the threshold value is set large for the character "6B", it is shown in FIG. 9 (a).
As shown in Fig. 9, the whole line figure is undesirably coarsely approximated. On the other hand, when this threshold value is set to be small, the straight line is distorted as shown in Fig. 9 (b), and unnecessarily finely approximated. It will be done. As described above, there are problems in that the shape of the entire line figure is distorted or the number of broken lines becomes larger than necessary. In making a polygonal line of this line figure, it is desirable to perform the polygonal line approximation separately for a portion such as a long straight line portion to be roughly approximated to a polygonal line and a portion such as a symbol or a character to be accurately approximated. [Means for Solving Problems] In order to solve the above problems, the present invention detects a feature point including an end point, a branch point, and a bending point of a thinned thinned figure using a predetermined mask. Feature extraction mechanism section 3
And a line tracking mechanism unit that traces the feature points detected by the feature extraction mechanism unit 3, stores distance information about the feature points in the point counter 6, and stores position information about the feature points in the point table 7. 5, an approximation accuracy determination mechanism unit 8 that determines the approximation accuracy based on the information stored in the point counter 6 and the point table 7, and a distance calculation mechanism unit that calculates the distance for the point sequence registered in the point table 7. 9
Is provided, the distance calculation result is calculated by the distance calculation mechanism unit 9 until the result becomes smaller than the approximation precision determined by the approximation precision determination mechanism unit 8, and the result is output. There is. FIG. 1 shows a principle configuration diagram of the present invention. The image memory 1 in the figure stores thinned figures thinned by the thinning mechanism unit 10. The initial setting mechanism unit 2 sets various initial values and threshold values. The feature extraction mechanism unit 3 detects a feature point consisting of an end point, a branch point, and a bending point from the point sequence of the thinned figure read from the image memory 1, and associates the detected feature point information with the corresponding point sequence. It is something to store. The mask memory 4 stores a mask for detecting feature points. The line tracking mechanism unit 5 tracks the point sequence of the thinned figure, stores the distance information of the point sequence in the point counter 5, and stores the position information of the characteristic points in the point sequence in the point table 7. . The point counter 6 stores distance information for calculating the distance between the start point and the end point of the point sequence. The point table 7 stores position information of feature points and the like. The approximation precision determination mechanism unit 8 determines in a mode of selecting one of the predetermined approximation precision (threshold value) initialized from the values stored in the point counter 6 and the point table 7. The distance calculation mechanism unit 9 calculates that the calculated distance from each point sequence is
The linear distance is calculated by repeatedly calculating the distance until it becomes smaller than the approximation accuracy determined by the approximation accuracy determining mechanism unit 8. The thinning mechanism unit 10 performs known thinning processing on an image generated by a scanner or the like. [Operation] Next, the operation will be described. In FIG. 1, the feature extraction mechanism unit 3 includes an image memory 1
Each point sequence of the thinned figure read out from the inside is collated with the mask read out from the mask memory 4, and characteristic points consisting of end points (starting point and ending point are hereinafter referred to as “end points”), branch points and bending points are identified. It is extracted and this information is stored in association with each point sequence. The line tracing mechanism unit 5 traces the point sequence of the thinned figure, and starts from the start point P ₀ to the end point of the traced point sequence.
The distance information up to P _n is stored in the point counter 6, and the position information such as the bending point is stored in the point table 7. The approximation accuracy determination mechanism unit 8 uses the distance information and the position information stored in the point counter 6 and the point table 7 to calculate the approximation accuracy (threshold value at the time of polygonalization) for each thinned figure to be subjected to linearization processing. ) Is determined. The distance calculation mechanism unit 9 calculates the distance by referring to the point table 7, sequentially performs the polygonal line approximation processing until the distance becomes smaller than the determined approximation accuracy, and outputs the result. As described above, the distance d between the start and end points (between the start point and the end point, which is the end point, is referred to as “start and end point” below) of the point sequence of the thinned figure for which the linearization processing is to be performed, and the straight line distance L between the start and end points.
The approximation accuracy is determined based on the approximation accuracy, and iterative linearization processing is performed until the distance of each point sequence becomes smaller than this approximation accuracy, so that the approximation accuracy suitable for the shape of the thinned figure is used to perform the polygonalization. It becomes possible. [Embodiment] Next, the configuration and operation of one embodiment of the present invention will be described in detail with reference to FIG. 2 to FIG. FIG. 2 shows an example of a thinned graphic. This is the thinning point sequence to be broken that has already been thinned by the known method.
Let _i (i = 0 to n) and the endpoints of these ends be P ₀ and P _n . _Let L be the distance along the point sequence of this thinned point sequence P _i , and d be the straight line distance between the start point P _P and the end point P _n . FIGS. 3 (a), (b) and (c) show mask examples for detecting feature points consisting of end points, branch points and bending points, respectively. When the black area in the figure corresponds to the point sequence of the thinned figure, it is determined that they match. This mask pattern is stored in the mask memory 4. It should be noted that these masks are rotated by 90 ° and matching is also performed on the point sequence of the mirror image pattern. FIG. 4 shows a display example of feature points. This is because the feature extraction mechanism unit 3 performs the mask matching shown in FIGS. 3A to 3C on the point sequence of the thinned figure and stores the result for each point sequence. is there. Pixel value “0” in the figure
Indicates a white pixel in the background of the point sequence of the thinned figure stored in the image memory 1. The pixel value "1" indicates a black pixel that does not correspond to a feature point. The pixel value "2" indicates an end point, for example, a black pixel at P ₀ and P _{n in} FIG. The pixel value “3” indicates a black pixel at the branch point. The pixel value “4” indicates a black pixel at the bending point. FIG. 5 shows an example of a point counter. This is because the line tracing mechanism unit 5 traces the point sequence of the thinned figure, increments n1 when the next point sequence of a certain point sequence is in the vertical and horizontal directions, and When the point sequence is in the 45 ° direction, the n2 part is incremented and the distance L of the point sequence (distance L in FIG. 2)
For calculating. FIG. 6 shows an example of a point table. This is because the line tracking mechanism unit 5 traces the point sequence of the thinned figure, and when a bending point (pixel value “4”) appears, the coordinates of this bending point are set to X _i and Y _i.
Is to be registered. Here, 1 is set at the start point P ₀ (X ₀ , Y ₀ ) and the end point P _n (X _n , Y _n ). When an end point (pixel value “2”) and a branch point (pixel value “3”) appear, tracking is stopped here and this point P _n is registered as an end point. Thus, in the point table 7, only the end points, the branch points, and the inflection points are registered, and the second table continuing in the same direction is registered.
Since the point sequence after the th is not registered, the amount of data is reduced compared to the case where all the points forming an arc are recorded and calculated,
It is possible to perform arithmetic processing at high speed. Next, the operation will be described in detail with reference to FIG. In FIG. 7, the figure shows a state in which initial setting is performed. This is the initial value k ₁ from the initial setting mechanism unit 2 in FIG.
It means to input and set k ₂ , k ₃ , h ₁ , and h ₂ . The figure shows a state in which an image is read. this is,
Scan a document with a scanner to generate a binary image,
It means storing in the image memory 1. The figure shows a state in which thinning is performed. This means that a known thinning process is performed on the line figure in the binary image stored in the image memory 1. This allows
Each line figure is represented by a series of points P ₀ to P _n as shown in FIG. The figure shows a state in which feature extraction is performed. As described above, the feature extraction mechanism unit 3 performs matching with the mask shown in FIG. 3 for the thinned figure represented by the point sequences P ₀ to P _n , and the result is This means storing at each point using pixel values as shown in FIG. The figure shows a state in which line tracing is performed. This is because the line tracking mechanism unit 5 uses the point sequence P ₀ to P ₀ to
P _n is tracked and the distance information between these point sequences is stored in the point counter 6 in the incrementing manner as described above, and the position information of the start point P ₀ , the bending point P _i , the end point P _n , etc. are stored. This means registration in the point table 7 as shown in FIG. The figure shows a state in which the approximation accuracy is determined. First, the length d of the line segment connecting the start point P ₀ and the end point P _n is calculated using the following formula. _{d = {(X 0 -X n} ) 2 + (Y 0 -Y n) 2} 1/2 ...... (1) where, (X _0, Y ₀₎ is the starting point P ₀ coordinates, (X _n, Y _n ) represents the coordinates of the end point P _n . Secondly, the length L of the arc between the start point P ₀ and the end point P _n is calculated from the values of n ₁ and n ₂ held in the point counter 6 using the following formula.
Is calculated. L = n ₁ + (2) ^1/2 × n ₂ ...... (2) where n ₁ is the number of _consecutive points in the vertical and horizontal directions, as already mentioned, and n ₂ is the point in the 45 ° direction. Represents the number of consecutive columns. Thirdly, the approximation accuracy (threshold value) H for making a broken line is determined using the condition of the following equation. L ≦ k ₁ (3) d ≦ k ₂ (4) d / L ≦ k ₃ (5) When the above formulas (3) to (5) are satisfied, the approximation accuracy H = h ₁ To decide. When the above equations (3) to (5) are not satisfied, it is determined that the approximation accuracy H = h ₂ (h ₂ ≧ h ₁ ). The figure shows a state in which distance calculation is performed. This means that the folding process is repeatedly performed until the distance calculated for the point sequence registered in the point table 7 becomes smaller than the approximation accuracy H determined in the figure. 8th below
A specific description will be given with reference to the drawings. First, the starting point P ₀ and the ending point P _n registered in the point table 7 are connected by a line segment, a perpendicular line is drawn from each point to this line segment, and the maximum distance of this perpendicular line, for example, the point P _{i in} FIG. Find the distance h _{max of} the vertical line drawn from.
When the obtained h _max is smaller than the approximation accuracy H determined in the figure, the line segment connecting the start point P ₀ and the end point P _n is the broken line to be obtained. On the other hand, when the distance h _max is larger than the approximation accuracy H, “1” is stored at the position in the point table 7 corresponding to this point P _i , and the start point P _{0 and} this newly registered point are stored.
The P _i with obtaining a distance h _max as an end point, obtaining the same manner distance h _max against and this point P _i and the ending point P _n. Similarly, the process is repeated until the obtained distance h _max becomes smaller than the approximation accuracy H. Then, a desired polygonal line is obtained by sequentially connecting the points registered with "1" registered in the point table 7. Next, the initial values k1, k2, k3, h1, and h2 will be described in detail with reference to (a) and (c) of FIG. 10 using specific numerical examples. As shown in FIG. 10 (a), the processing target is an image which has already been subjected to the thinning process and has a width of 1 pixel. The purpose is to approximate the lower straight line portion (point sequence C) by a line segment that is as smooth as possible, and the portion of the character "6" (point sequence A, B) by a line segment that is as fine as possible. In Fig. 10 (a), the point sequence A is 10 points in total from points A1 to A10, and the point sequence B is 18 points in total from point B1 (same as point A10) to returning to point B1. Row C is assumed to be composed of a total of 24 points from point C1 to point C24. FIG. 10 (b) shows a polygonal line approximation result (represented by a thick line) according to the prior art. Here, the distance threshold H (described on page 12, line 19 of the specification) for controlling the approximation is fixed at 2.5 (pixels) as an example. In the prior art, without taking into consideration the complexity of figures and the length of point sequences,
Since the approximation is performed only by the threshold value H, the point sequence C is a smooth line segment, while the character sequences of the point sequences A and B are collapsed. FIG. 10 (c) shows the result of processing on FIG. 10 (a) using the technique of the present invention. As an example, the following initial values were set here (corresponding to the description on page 10, line 18 to line 20, and page 12, line 18 to page 13, line 6 of the specification).・ Point sequence length threshold k1 = 20 (pixels) ・ Point sequence length threshold k2 = 20 (pixels) ・ Curticity threshold k3 = 0.9 ・ Approximation accuracy h1 = 1.0 (pixels) ・ Approximation accuracy h2 = 2.5 (pixels) As shown on the right side of (a) in Fig. 10, ・ About point sequence A ・ Arc length (point sequence length) L = 10.2 ・ Distance between end points d = 8.48 ・ Curveness d / L = 0.83 ・ Point sequence About B ・ Arc length (point sequence length) L = 19 ・ Distance between end points d = 15.23 ・ Curveness d / L = 0.80 ・ About point sequence C ・ Arc length (point sequence length) L = 24.6 ・ End point Distance d = 23 · Curvature d / L = 0.93 Apply the above data to the conditions (3), (4) and (5) from page 12 line 18 to page 13 line 6 of the specification, When the respective values are set, the point sequence A satisfies the conditions (3), (4), and (5), so it is determined that the approximation accuracy H = h1 = 1.0. Since the point sequence B satisfies the conditions (3), (4), and (5), it is determined that the approximation accuracy H = h1 = 1.0. Since the point sequence C does not satisfy the conditions (3), (4), and (5), it is determined that the approximation accuracy H = h2 = 2.5. Each point sequence A, with the approximation accuracy determined in this way,
When B and C are subjected to the polygonal line approximation processing, the points A4, A7, B5, B10 and B13 are detected as bending points and the desired polygonal line shape is obtained as if they were connected by the thick straight line in (c) of FIG. . [Effects of the Invention] As described above, according to the present invention, the optimum threshold value (approximation accuracy) is determined based on the total length of the thinned point sequence and the length of the line segment connecting both ends of the point sequence, Since a configuration that calculates the distance until a distance smaller than this threshold is obtained, it is possible to easily discriminate between places where rough line approximation is desired, such as a long straight line portion, and where fine line approximation is desired, such as characters. However, it is possible to obtain a polygonal line with good quality by performing polygonal line formation using threshold values suitable for each. Also,
Since the end point, the branch point, and the bending point are detected as the characteristic points and only those points are registered in the point table 7 to be the object of the distance calculation, the polygonalization processing can be performed at a high speed with a small amount of calculation processing. Can be done.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram of the principle of the present invention, FIG. 2 is an example of a thinned figure,
FIG. 3 is an example of a feature point detection mask, FIG. 4 is an example of feature point display,
FIG. 5 is an example of a point counter, FIG. 6 is an example of a point table, FIG. 7 is a flowchart for explaining the operation of the present invention, FIG. 8 is a broken line explanatory view, FIG. 9 is an operation explanatory view of a conventional device, and FIG. Shows an explanatory diagram of a specific example of the linearization processing. In the figure, 1 is an image memory, 2 is an initial setting mechanism unit, 3 is a feature extraction mechanism unit, 4 is a mask memory, 5 is a line tracking mechanism unit, 6 is a point counter, 7 is a point table, and 8 is an approximate accuracy determination mechanism unit. , 9 represents a distance calculation mechanism unit, and 10 represents a thinning mechanism unit.

   ────────────────────────────────────────────────── ─── Continuation of front page    (72) Inventor Noichi Ito               Fujitsu, 1015 Ueodanaka, Nakahara-ku, Kawasaki-shi               Inside the corporation                (56) References JP-A-58-134747 (JP, A)                 JP-A-60-215283 (JP, A)                 JP 60-201475 (JP, A)                 JP-A-60-49483 (JP, A)                 JP 61-208585 (JP, A)

Claims (1)

(57) [Claims] In a line figure folding device for thinning a line figure, a feature extraction mechanism section (3) for detecting a feature point consisting of an end point, a branch point, and a bending point of the thinned figure that has been thinned using a predetermined mask, The point sequence of the feature points detected by the feature extraction mechanism unit (3) is tracked, and the point sequence length information that is the length of the arc formed by the point sequence is stored in the point counter (6).
A line tracking mechanism unit (5) that stores point position information of each point in each column in a point table (7), and a line tracking mechanism unit (5) that moves from a start point to an end point obtained by the point sequence length information stored in the point counter (6). The length L along the point and the length d of the line segment connecting the start point and the end point in the point position information stored in the point table (7) are obtained, and based on L, d, d / L For the approximation accuracy determination mechanism (8) that determines the approximation accuracy and the point sequence stored in the point table (7), 1
Calculate the vertical distance from a straight line (hereinafter referred to as a reference line) that connects the start point and the end point of a continuous point sequence that forms one arc to each point that forms the arc, and the point with the maximum distance in the point sequence and its Distance settlement mechanism unit (9) to detect the maximum distance value when
And a line-figure polygonalization device, wherein the reference line is determined as an approximate line segment based on the approximation accuracy.