JPH01246677A - Segment length simplified calculation system - Google Patents

Abstract

PURPOSE:To accurately obtain a segment length by tabling square root data beforehand, retrieving the square root data based on the absolute value of the coordinate difference of the X axis and the Y axis of a segment vector, multiplying a regularization coefficient and obtaining a segment length. CONSTITUTION:When starting points A(Xs, Ys) and terminations B(Xe, Ye) of the segment are given, the absolute value of the coordinate difference between the X axis and the Y axis is calculated by difference computing elements 1 and 2 and absolute value computing elements 3 and 4. X, Y direction coefficient computing elements 5 and 6 divide the absolute value of the coordinate difference with respective reference values and obtain direction coefficients Kx and Ky. A regularization coefficient K is obtained by a maximum value comparator 7. The absolute value of the coordinate difference between the X axis and the Y axis is divided by the regularization coefficient K, the value is made into a key, a two-dimensional reference segment length table memory 9 is retrieved. The square root data are stored in the reference segment length table memory 9. K is multiplied to the retrieved value and a segment length is obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION A. Field of Industrial Application The present invention relates to a method for calculating line segment vectors when recognizing character or graphic patterns using an input device such as an image scanner, and in particular to a method for calculating line segment vectors. This paper relates to a simple line segment length calculation method for calculating length.

B1 Summary of the Invention The present invention uses pre-calculated square root data as m a reference line segment length table memory stored in a table of m to g, and a normalization coefficient calculation means for making the absolute value of the coordinate difference between the X axis and the y axis of the line segment vector correspond to the number of divisions of the reference line civil table; With that normalization coefficient,
means for comparing the reference line segment length table with each direction reference value obtained by dividing the absolute value of the coordinate difference between the axes and the y-axis, and normalizing the reference line segment length table to the square root data adopted from the reference line segment length table memory. The present invention discloses a technique for improving the method of calculating the length and direction of a line segment vector, especially the line segment length, by multiplying by a coefficient, and performing an accurate approximation calculation of the square root.

C3 Conventional technology When recognizing patterns such as characters and figures, the length and direction of a line segment vector connecting two coordinate points are usually treated as feature quantities. For example, black and white images obtained by scanning documents, drawings, etc. with an input device such as an image scanner
Value image data is often discussed using a series of line segment vectors along the contours of characters or figures, and in that case, it is necessary to calculate the length and direction as many as the number of line segment vectors.

As shown in FIG. 3, the length a and direction θ of a line segment vector starting from coordinate point A (xs, ys) and ending at coordinate point B (xeSye) can be determined as follows.

Calculations of square roots and arctangents such as the above equation 0 are complex and require a large amount of calculation on the main processor, so usually a sub-processor such as a floating-point arithmetic unit is used as an auxiliary processor in addition to the main processor. Therefore, methods are being used to reduce the computational burden on the main processor and at the same time increase processing speed.

D0 Problems to be Solved by the Invention However, the use of a sub-processor increases costs, increases the size of the device, and impedes the spread of the system. Therefore, although sub-processors are not always used in machines of all ranks, if a sub-processor is not provided, the load on the main processor for calculations will be heavy.

Generally, in cases such as contour vectors of characters or figures, there are many vectors, and the calculations become proportionally larger, so if there is no subprocessor, there is no need to perform square root calculations or arctangent calculations. Reduce the burden on the main processor by performing highly accurate approximate calculations such as
It is necessary to devise ways to speed up the processing.

An example of approximate calculation regarding the length of a line segment vector is shown below.

The following equation represents the relationship between the length of the line segment vector and the coordinate point.

Q≦1xe-xsl+lye ysi-...■In image processing that often deals with local areas, based on this formula, only when the starting point and end point are close, Q=lxe-xsl+lye ysl - In some cases, as in -■, the left and right sides are treated as equivalent, and the calculation is simple and the error is often treated as within an allowable range. However, in this 0 formula, if the starting point and the ending point are far apart, ( 2n-1)
It is obvious that the closer it gets to π/4, the larger the error is, which is beyond the allowable range, and the accuracy cannot be guaranteed at all.

The present invention was created in view of these problems, and
Among the methods of calculating the length and direction of line segment vectors, the purpose of this invention is to improve the method of calculating length in particular, and to provide a simple line segment length calculation method that performs accurate square root approximation calculations.

91 Means for Solving the Problems In the present invention, the means for solving the above problems are as follows:
In the line segment length simple calculation method that calculates line segment lengths on X-axis and y-axis coordinates that are orthogonal to each other, there is a reference line segment length table memory that stores square root data calculated in advance in m x m tables, and a line segment length table memory that stores square root data calculated in advance in m x m tables. means for calculating a normalization coefficient that makes the absolute value of the coordinate difference between the X-axis and the y-axis of the vector correspond to the number of divisions of the reference line segment length table; means for comparing the reference line segment length table with each direction reference value obtained by dividing the base line segment length table, and multiplying the square root data adopted from the base line segment length table memory by the normalization coefficient to obtain a line segment vector. A simple line segment length calculation method is used to obtain the line segment length.

F1 action The present invention is a means for preparing a table of square root data in advance, setting a normalization coefficient in order to make the absolute value of the coordinate difference between the X-axis and y-axis of a line segment vector correspond to the table, and calculating this normalization coefficient. and means for calculating direction reference values for both axes of the table using a normalization coefficient, search for square root data using the direction reference value, and multiplying the square root data by the normalization coefficient to obtain a line segment length.

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

FIG. 1 is a block diagram showing an example of a simple line segment length calculation device embodying the present invention. In Figure 1, l is an X-direction difference calculator, 2 is a y-direction difference calculator, 3 is an X-direction absolute value calculator, 4 is an X-direction absolute value calculator, and 5 calculates the normalization coefficient Kx in the X-direction. 6 is a y-direction coefficient calculator that calculates the normalization coefficient Ky in the y-direction, 7 is a maximum value comparator that selects the larger of both coefficients, and 8 is a line segment that calculates the line segment length. The length calculator 9 is a reference line segment length table memory. The reference line segment length table memory 9 stores values obtained by calculating square roots in advance and converting them into integers as m x m tables as shown in FIG. The length Q of the line segment vector is determined using the following equation.

First, start point A (xs, ys) and end point B (xeS
ye) are given, their X components (xe%xs)
is input to the X-direction difference calculator 1 and the y component (ye, ys)
is input to the y-direction difference calculator 2.

Difference calculators l and 2 calculate the coordinate difference between the end point and the start point of the line segment vector for the X-axis and the y-axis, respectively, and
The axis absolute value calculator 3.4 calculates the absolute value of the coordinate difference 1xe-xs
Output l, lye ysl.

The normalization coefficient is used to make the absolute value of the coordinate difference between the X-axis and y-axis of a line segment vector correspond to the number of divisions in the reference line segment length table 9, and depending on the length of the vector, the reference line segment length table 9
In order to decide which position of the data should be retrieved,
This is a coefficient for normalizing to table space. K is a natural number and is determined as follows.

K=max(l xe-xs l/m+1゜l ye-
ys l/m+1)-■ In this embodiment, the X-direction normalization coefficient Kx is calculated by the X-direction coefficient calculator 5 as Kx-1xe-xs l/m+1, and the y-direction normalization coefficient Ky is calculated as By the coefficient calculator 6, Ky= i ye-ys l/rn engineering 1
They are compared by the maximum value comparator 7, and the larger of both coefficients is selected.

At this time, the reference value i in the X direction on the horizontal axis of the table and y on the vertical axis
The direction reference value j is determined based on K as follows.

Here, i and j are positive integers, 0≦i≦m-1, 0≦J≦
It is m-1.

When the line segment length calculator 8 selects the value Ci, j at the (i, j) position of the basic line segment length table 9 based on these i, j, the line segment length calculator 8 calculates the length Q as follows.

Q=ni-・C1,j ・・・・・・・・・■
In this way, the present invention can easily determine the length Q of the line segment vector.

Incidentally, when evaluating the calculation error in this method, at most it is within the range of ((difference between a value at a certain position in the table and an adjacent numerical value) x (k-1)).

For example, the starting point A of the line segment is (!0.12) and the ending point B is (6
L30. ), the line segment length is 54.08 (5 digits to the 4th decimal place), but according to the method of the present invention, = 4.
1=12, j=4, and a length of 52 is obtained,
The error is at most 2.08. If real number data is provided in which the increments of the reference line segment length table are finer than the example shown in FIG. 2, the error will be further reduced.

When the above numerical values are applied to the equation (2) shown as a conventional simplified method, the result is 51+18=69, and the error is 14.92.It is clear that the error has been improved by the present invention.

As described above, in the method of the present invention: (1) The length of a line segment can be easily obtained without supplementary use of a sub-processor such as a floating point unit.

(2) The number of calculations m is small, and the burden on the main processor is small.

(3) Since the processing only refers to a table, it does not require repeated calculations and is fast.

(4) Since the error can be evaluated in advance, the table can be constructed after determining whether the processing is within the allowable range.

H. Effects of the Invention As explained above, according to the present invention, among the calculations of the length and direction of line segment vectors, the length calculation method is particularly improved, and the square root of the line segment can be calculated with high precision by approximating the square root. A long and simple calculation method can be provided.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of the present invention, FIG. 2 is a sample diagram of a reference line segment length table, and FIG. 3 is an explanatory diagram of line segment vectors. 1...X-direction difference calculator, 2...X-direction difference calculator, 3...X-direction absolute value calculator, 4...X-direction absolute value calculator, 5...X-direction coefficient calculator , 6... Y-direction coefficient calculator, 7... Maximum value comparator, 8... Line segment length calculator, 9... Language line segment length table memory. Figure 2

Claims (1)

[Claims] In the line segment length simple calculation method that calculates line segment lengths on mutually orthogonal x-axis and y-axis coordinates, there is a reference line segment length table memory that stores square root data calculated in advance in m x m tables, and a line segment length table memory that stores square root data calculated in advance in m x m tables. means for calculating a normalization coefficient that makes the absolute value of the coordinate difference between the x-axis and y-axis of the vector correspond to the number of divisions of the reference line segment length table; means for comparing the reference line segment length table with each direction reference value obtained by dividing the line segment vector by multiplying the square root data adopted from the reference line segment length table memory by the normalization coefficient. A simple line segment length calculation method characterized by obtaining the line segment length.