JPS63182780A - Picture processing method for drawing reading device - Google Patents

Abstract

PURPOSE:To surely recognize patterns by performing the second scan of a bit pattern after line thinning process to detect a bit array that disturbs the following pattern recognizing process and correcting the bit array. CONSTITUTION:The line thinning data is checked again in a TV scan (raster scan) mode of a (3X3)-mask M1 before the bit pattern data undergone a necessary line thinning process for the pattern recognizing process of the next stage is compressed by the linear approximation. When an inter-bit connection (a lump of picture elements) that disturbs the following pattern recognizing process is detected through said scan, this lump is eliminated for conversion into a bit pattern free from faults. Thus it is always possible even for a binarized pattern (bit pattern) obtained after a line thinning process carried out by the Deutsch algorithm to track the lines thereafter.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention is a screen reading device for reading character and graphic information drawn on maps and drawings and inputting the information into a computer. The present invention relates to a method of processing image data that is processed in advance within a drawing reading device so as to be convenient for graphic recognition processing in a subsequent computer.

[Conventional technology]

In general, a drawing reading device includes an imaging device that optically scans text and graphic information drawn on a map or drawing and outputs it as a grayscale signal, and a binarization device that binarizes the grayscale signal and outputs it as a bit pattern. , a thinning processing means for performing thinning processing on the binary bit pattern, and a data compression means for compressing the bit pattern data subjected to the thinning processing by linear approximation. In order to reduce the required memory capacity, the read character and graphic information is compressed as much as possible before being input into the computer.

Here, the line thinning process will be briefly explained.

In graphic and image processing, processing of line figures plays an important role, and since characters and drawings are originally composed of line drawings, it is essential to decompose them into line segments and process them. When analyzing a grayscale image structurally, the geometric shape is often analyzed from target features obtained through edge detection before describing the relationships between the basic segments that make up the image.

When handling lines in this manner, it is desirable to ideally process thin lines rather than wide lines, and various methods for thinning lines have been proposed and used since ancient times.

In this sense, the above-mentioned thinning process is a process that is performed on a wide bit pattern that has just been binarized to make the line thinner.

Algorithms for such thinning processing include Rutovitz's algorithm and Hildit's algorithm.
Representative examples include the ch algorithm, 1) the eutsch algorithm, the Yokoi algorithm, and the date algorithm. In the thinning process, it is necessary to select and use an appropriate algorithm from among the various algorithms described above, depending on the target figure.

When dealing with maps, circuit diagrams, etc., Deutsch
The algorithm is considered appropriate.

As mentioned above, line thinning processing is a preprocessing process for recognizing characters and figures, which involves processing binary characters with line widths,
This is done to convert graphic data into skeleton lines to facilitate line tracing (method for recognition). In the case of circuit diagrams, it is necessary to separate connection lines and symbols, and to extract so-called feature points such as intersections, branch points, end points, and starting points of connection lines. Since it is necessary for recognition, in order to extract it, the binarized bit pattern that has been thinned is usually scanned with a 3×3 mask of 3 bits each in the vertical and horizontal directions.

In other words, by scanning and tracing the binarized bit pattern after the thinning process using a 3x3 mask of 3 bits each in the vertical and horizontal directions, we can detect the intersection points, branching points, and end points of the connecting cotton. , feature points such as the starting point are extracted, and the target figure is recognized based on them.

[Problem that the invention seeks to solve]

However, when thinning a bit pattern using the Deutsch algorithm described above, depending on the shape of the binary figure (bit pattern), the intersection point (
A cluster of pixels (beep 1 cluster) where it should be a feature point)
occurs, making it difficult to detect intersection points (feature points) by line tracing, and even making them unrecognizable.

This will be specifically explained below with reference to the drawings.

Figure 4 (a) shows the binarized figure (bit pattern) after being thinned using Deutsch's algorithm.
It is an explanatory view showing an example of. In the same figure, Ml is 3×3
It's a mask.

FIG. 4A shows identification numbers ■ to ■ in order to identify the nine bit positions constituting the 3×3 mask M1.
FIG.

In FIG. 4(A), the bit positions ■, ■, ■, ■ in the 3×3 mask M1 are all 1, and it is recognized that a cluster of pixels (a cluster of bits 1) has occurred. Will. When a cluster of pixels like this (a cluster of bit 1) occurs, even if you try to detect feature points by line tracing, a situation will occur where it will not be possible to extract the tracing route from the cluster, and the feature will be lost. Points cannot be detected. In this case, as shown in FIG. 4(b), it is necessary to loosen the lumps in order to enable subsequent line tracing.

In FIG. 5(a) as well, it can be seen that a pixel cluster (a cluster of bits 1) similarly occurs. In FIG. 6(a), no pixel cluster (bit 1 cluster) occurs, but originally, as seen in FIG. 6(b), 3
×3 Mask Ml becomes a cross-shaped bit pattern, and bit position ■ is supposed to become a feature point, but due to an algorithmic problem during the thinning process, l of bit position ■ is shifted to bit position ■. 3×3 squares M1
As a result, the bit position (2) does not become a feature point, which means that the feature point cannot be detected, which poses a problem in figure recognition.

Therefore, in the present invention, the above-mentioned disadvantages are solved and De
An object of the present invention is to provide an image processing method in a drawing reading device that always enables subsequent line tracing even in a binarized figure (focus pattern) that has been subjected to line thinning processing using the Utsch algorithm.

[Means for solving problems]

In order to solve the above problems, the present invention provides an imaging means that optically scans character and graphic information drawn on a map or drawing and outputs it as a grayscale signal, and a system that binarizes the grayscale signal and outputs it as a bit pattern. a binarization means; a thinning processing means for performing line thinning processing on the binarized bit pattern as one of image data processing required for subsequent figure recognition processing; In a drawing reading device comprising data compression means for compressing the processed bit pattern data by linear approximation, the bit pattern after being subjected to line thinning processing is scanned again.

[Effect]

As a result of scanning, if a connection between bits (clump of pixels) is found that becomes an obstacle during the subsequent graphic recognition process, this connection (clump) is resolved and converted into a bit pattern in which the obstacle is removed.

〔Example〕

The present invention will now be described in detail with reference to the drawings.

FIG. 1 is a chart showing the processing flow of an embodiment of the present invention.

Drawing reading devices read characters drawn on maps and drawings,
Imaging processing that optically scans graphic information and outputs it as a grayscale signal; Binarization processing that binarizes the grayscale signal and outputs it as a bit pattern;
Thinning processing that is performed after preprocessing such as waveform shaping (valued image data), data compression processing that compresses data by performing line tracing and linear approximation on the resulting thinned data, etc. As described above, in the present invention, as indicated by the symbol S in FIG. 1, the thinning correction process is performed on the thinning data before line tracing and linear approximation conduct.

FIG. 2 is a flowchart showing details of this thinning correction process.

FIG. 3 is an explanatory diagram showing a 3×3 mask M1 and a 5×5 mask M2 used in the thinning correction process.

The thinning correction process will be explained with reference to FIGS. 2 and 3.

In step Sl, rechecking of the thinning data in TV scanning (rusk scanning) using the 3×3 mask M1 is started. In step S2, the next scanning line is scanned.

Next, in step S3, as a result of scanning with the 3×3 mask M1, r4・r3・r5・rO=1...(1) holds, or X゛=[Σ''K-1l rm rh+t I)
/ 2≧3 ... (2
) is established, it is determined that a link between bits that becomes an obstacle in the subsequent figure recognition process has been found, and the process proceeds to the next step S4, where a correction process as described below is performed.

Here, the above equation (1) will be explained.

In the 3×3 mask Ml in FIG. 3, bit position ffrO
, r3. r4. r5 are located adjacent to each other and form a cluster as a whole. and these bit positions r
o, r3. r4. Since each possible value of r5 is 1 or 0, the fact that the above equation (1) holds means that bit position rQ.

r3. r4. This means that each value of r5 is all 1. That is, this shows that the pixel cluster (bit 1 cluster) shown in FIG. 4(A) and FIG. 5(A) is generated.

Further, in the above equation (2), since the possible values of r and rk+1 are 1 or 0, calculation can be made according to the bit positions shown in the 3×3 mask M1 of FIG. As a result, the fact that the above formula (2) holds true means that a bit pattern that causes trouble in subsequent figure recognition is generated, as shown in Figure 6 (a). be.

If the above equation (1) or (2) is established, the process proceeds to the next step S4 and a correction process is performed.This correction process will be described below.

One of the conditional expressions used in the correction process is shown below as (a).

(a) If R1+R2+R3>0, set r1=1; otherwise, set r1=0.

In other words, by performing rask scanning of the bit pattern using the 5×5 mask M2 shown in FIG. This is why the correction processing is performed.
゛The conditional expressions to be executed are listed below.

(b) If R5+R6+R7>0, set r3=1; otherwise, set r3=0.

(c) If R9+R10+R11>O, set r5=1, otherwise set r5=0.

(d) If R13+R14+R15>O, r7=1
Otherwise, let r7=0.

(e) If R3・R4・R5=1, set r2=1,
Otherwise, let r2=0.

(If f'JR?-R8-R9=1, set r4=1; otherwise, set r4=0.

(g) If R11・R12・R13=1, then R6−
1, otherwise let r6=0.

(h) If R15・R16・R17=1, then r8=
1, otherwise let r8=0.

(i) If rO・R3・R4・r5=1, then rO=0
Let it be.

(j) If r 1 + r 7=0, then r7=1
Let it be.

FIG. 4(b) shows the result of performing the above correction processing on the bit pattern shown in FIG. 4(a). Similarly,
Figure 5 (B) shows the result of performing the above correction process on the bit pattern shown in Figure 5 (A), and the result of performing the above correction process on the bit pattern shown in Figure 6 (A). is shown in Figure 6 (b).

As a result of the correction process, the Deutsc used in the thinning process
A block of pixels (bit 1
It will be understood that bit arrays that would be a hindrance to subsequent figure recognition processing, such as clusters of , have been eliminated.

〔Effect of the invention〕

As explained above, according to the present invention, in the drawing reading device, after the line thinning process, the bit pattern is scanned again using the 3×3 mask M1, and the block of pixels (clump of bit 1), etc. Since a bit array that is a hindrance in the figure recognition process is detected, and then a correction process is performed using the 5×5 mask M2, there is an advantage that the subsequent figure recognition can be performed reliably.

[Brief explanation of the drawing]

FIG. 1 is a chart showing the processing flow of an embodiment of the present invention, FIG. 2 is a flowchart showing details of the thinning correction process, and FIG. 3 is a 3×3 mask Ml and a 5×5 mask used in the thinning correction process. An explanatory diagram showing the mask M2, FIG. 4(A) is an explanatory diagram showing an example of a binarized figure (bit pattern) after thinning processing is performed using the eutsch algorithm, FIG. 4(B) 4A is an explanatory diagram showing the bit pattern after the correction process, FIG. Figure 5 (a) is Deutsch
An explanatory diagram showing another example of a binarized figure (bit pattern) after thinning processing using the algorithm, Fig. 5 (
B) is an explanatory diagram showing the bit pattern after the correction process,
Figure 6 (A) is an explanatory diagram showing still another example of a binary figure (bit pattern) after thinning processing is performed using Deutsch's algorithm, and Figure 6 (B) is an explanatory diagram showing the bit pattern after the correction processing. It is an explanatory diagram showing a pattern. Description of symbols Ml...3x3 mask, M2...5x5 mask Agent Patent attorney Akio Namiki Agent Patent attorney Kiyoshi Matsuzaki 1 Figure 2 Figure M3 figure yak figure before correction After correction

Claims (1)

[Claims] 1) Imaging means that optically scans character and graphic information drawn on a map or drawing and outputs it as a grayscale signal, and binarization that binarizes the grayscale signal and outputs it as a bit pattern. a thinning means for performing line thinning processing on the binarized bit pattern as one of the image data processing required for subsequent figure recognition processing; data compression means for compressing bit pattern data by linear approximation, the drawing reading device scans the bit pattern again after the thinning process and performs the subsequent figure recognition process. 1. An image processing method for a drawing reading device, characterized in that, if a connection between bits that causes a problem is found during the image processing, the connection is canceled and converted into a bit pattern in which the problem is removed.