JPS6074092A - Extracting circuit of contour shape information - Google Patents

Abstract

PURPOSE:To speed up the extraction of contour shape information on a binary- code picture as an object of pattern recognition by storing every picture element of the contour shape information together with the number of a picture area, and rearranging storage contents by a rearrangement processing and generating chain codes which make one round of the contour. CONSTITUTION:A binary signal obtained by making a raster scan on picture elements of the binary image to be measured is inputted to the one-line memory 311 of a containment code generating circuit 31 and the lower line of a 2X2 mask circuit 312, and the signal is delayed by one line in the memory 311 and inputted to the upper line to constitute a mask pattern as to four picture elements, generating a code. A circuit 32 inputs the code from the circuit 31 and supplies it to a circuit 33, and the area number of every area of the object is determined to output the number of the frame of a mask pattern to a buffer circuit 341. A circuit 342 inputs the code from the circuit 32 to select an area number of the circuit 341 according to a specific rule, and outputs it to a memory 343. Codes in the memory 343 are converted into chain codes by the rearrangement processor 344.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical field to which the invention pertains] This invention relates to an image information processing circuit. In particular, the present invention relates to a circuit that compresses contour shape information of an object in a binary image into one-dimensional information, which is suitable as one of the means for pattern recognition.

[Description of prior art]

The contour tracking method shown in FIG. 1 is often used to measure the contour shape of an object, which is important in image measurement. This method traces the outline of the object 12 in the binary image clockwise 13 or counterclockwise, goes around the outline of the object, and creates an 8-direction chain code or up, down, left, and right directions as shown in Figure 2. Using four-directional codes, tracking directions are sequentially identified, and two-dimensional information about the outline of the object is compressed into one-dimensional information. There is a method that relies on hardware to perform this contour tracing at high speed, but the circuit becomes complicated and there are few examples of such conventional circuits. Further, when the direction judgment associated with the contour tracking method is performed using CPIJ or the like, there is a drawback that real-time processing is difficult and it is not suitable for high-speed processing.

[Purpose of the invention]

SUMMARY OF THE INVENTION An object of the present invention is to provide a circuit that measures a contour shape using an alternative method to the conventional contour tracing method and eliminates the disadvantages of the contour tracing method associated with high-speed image processing.

[Key points of the invention]

Instead of directly generating a chain code using the contour tracking method, the present invention first generates a continuum code by raster scanning, and processes this code to generate a chain code corresponding to the contour shape of two images. It has an input terminal into which two images plane-scanned according to the horizontal synchronization and vertical spacing 1111 are input, a circuit that numbers each pixel represented by the signal of this terminal for each image area, and an input terminal to which the input terminal is input. a circuit that initiates a continuance coat for four adjacent pixels on the screen represented by the signal; and a memory that stores, for each pixel, the number of the pixel@area and the continuation code of that pixel as a set. circuit, and a processor connected to the memory circuit for rearranging the addresses of the memory contents of the memory circuit. It is characterized by rearranging the addresses so that they are equivalent to the code.

[Explanation based on examples]

The present invention will be explained based on the drawings.

FIG. 3 is a block configuration diagram showing the configuration of the embodiment circuit.

First, the configuration of the circuit of this embodiment will be explained based on FIG.

This circuit consists of an image signal Kugata terminal 30 that inputs binary signals of "1" and rOJ obtained by raster scanning a binary image to be measured, a continuation code generation circuit 31, and a control circuit.・Logic circuit 32
The area number bone includes a circuit 33, a chain code generation circuit 34, and a timing generation circuit 35, and further includes:
Containment code generation circuit 31, one-line memory 311 providing a delay of one horizontal scanning period, and 2×2
The chain code generation circuit 34 also includes a number buffer circuit 341, a number selector circuit 342, a data memory 343, a rearrangement processor 344, and a continuation code counter 345.

Next, the connections of the circuit of this embodiment will be explained based on FIG.

The terminal 30 is connected to an image signal source (not shown), and this image signal is sent to the continuation code generation circuit 3.
1.

The output of the continuation code generation circuit 31 is connected to the control logic circuit 32.

The continuation code output of the control logic circuit 32 is connected to the input of the region number circuit 33, and the control signal output of the control logic circuit 32 is connected to the input of the number selector circuit 342 of the chain code generation circuit 34. is also connected. Further, the continuation code output of the control logic circuit 32 is sent to the data memory 3 of the chain code generation circuit 34.
43, and the timing signal output of the control logic circuit 32 is connected to the input of a continuous code counter 345 of the chain code generation circuit 34. Further, the output of the area number circuit 33 is connected to the input of a chain code generation circuit 340 and a number buffer circuit 341.

The output of the number buffer circuit 341 is sent to the number select circuit 34.
2 input. The area number signal output from number select circuit 342 is connected to the input of data memory 343, and the output of the continuation court counter is connected to be provided to data memory 343 as an address. The output of data memory 343 is connected to the input of reordering processor 344.

The output of the timing generation circuit 35 is sent to the data memory 343 of the chain code ordering circuit and the number buffer circuit 3.
41 input.

Next, prior to explaining the operation of the circuit of the present invention, the main points of the drawings from FIG. 3 to FIG. 13 and the tables from Tables 1 to 4 used for this explanation will be explained.

FIG. 3 is a block configuration diagram showing the configuration of the circuit of this embodiment as described above.

Figure 4 shows the names of the four frames of the mask pattern composed of four adjacent pixels that are input to the 2x2 mask circuit in Figure 3. One set is called the upper line of the mask pattern, and one set of frame C and frame d is called the lower line of the mask pattern.

FIG. 5 shows a state in which the area of the screen 50 is divided into three parts by an outline 52 of the object on the screen 50 and a line 53 that cuts the object into two areas. 2
Region numbers "1" and "2" are assigned to each part, and the region number rOJ is assigned to the background region part.

FIG. 6 is a diagram showing a mask pattern that follows the outline of the object shown in FIG. 5. Among the four pixels of the mask pattern, the shaded pixels are obtained by raster scanning. This signal indicates that the pixel belongs to the object region in FIG. 5, and
A blank pixel corresponds to a "0" value signal, indicating that the pixel belongs to the background region. Note that the black dots on the object outline in this figure indicate the common points of the four pixels of the mask pattern.

FIG. 7 is a diagram showing a generation circuit mask pattern on a line defined by the arrowed line C drawn in FIG. However, the background mask patterns on the left and right sides of the object are omitted.

FIG. 8 shows the continuation coat on the outline of the object shown in FIG. 5, classified into a lower left code and an upper right +1111 code.

Here, the lower left side code refers to number 1.8, which is the left side code, of the continuation code Cc shown in FIG.
．． 9.11 and 13 containment codes and the number 3 containment code, and the upper right code is the right code number 2.4.6.7 and I.
This is a general term for the containment chords numbered 4 and 12.

However, the containment codes numbered 5 and 10 are treated as a combination of the containment codes numbered 1 and 4 and the containment codes numbered 2 and 8, respectively. The containment code on the part 70 indicated by a short vertical line on the outline of the object in the figure belongs to the lower left code of area number "1" in Table 3, and is marked with a broken line along the outline. 6) The containment code on part 8I belongs to the upper right side code of "For area number",
Containment codes on areas 73 and 84 marked with circles on the outline belong to the upper right code of area number "2", and containment codes on area 82 marked with an x on the outline belong to the upper right code of area number "2". belongs to the lower left court of area number "2", the containment code on the part 85 indicated by the black circle on the outline belongs to the lower left court of area number "2'", and the contour The containment coat on the portion 86 marked with a solid line (G) along the line belongs to the upper right code of area number "2'". In addition, in Figure G, the reference numeral 80
The partial surfaces indicated by 0 and 850 respectively correspond to the area "
1'' and ``2'', and the portions surrounded by circles labeled 825 and 875 respectively represent a U-shaped curved location and an N-shaped curved location.

9 to 12 are logic flow diagrams showing the process of sequentially converting codes from Table 2 to Table 4, and illustrate the operation of reordering processor 344. First, it shows the process in which the contents of the data memory 343 are converted into a chain code that goes around the outline of the image object.

Figure 9 shows the summary and order of the processes in Figures 10 and 11. In the process in Figure 10, the conversions shown in Tables 2 to 3 are performed, and in the process in Figure 11, the conversions shown in Tables 2 to 3 are performed. Conversions from Table 3 to Table 4 are performed.

FIG. 12 shows the correspondence between the mask pattern, the continuation coat 'Cc, and the clockwise chain code Cd, and a direction code is also added to the chain code.

Figure 13 shows the mask pattern on the left end of the curved point and the mask pattern on the right 61
This shows the combination with the M mask pattern and is called a recessed point bear. There are three types of curved points: upwardly convex, downwardly convex, upwardly concave (hereinafter referred to as U-shape), and downwardly concave (hereinafter referred to as ``1-shape''). There is.

The names of the curve points are not affected by the number of mask patterns corresponding to continuance code numbers 3 and 12 that are interposed between the curve point pairs.

Table 1 shows the judgment criteria of the number selector circuit of the chain code generation circuit 34, and selects the frame temporarily stored in the number buffer circuit 341 corresponding to the continuation code from the control logic circuit. The area number within the selected frame is given to the data memory 343 as a reference for the selection.

Table 2 shows the area numbers and continuation codes stored in data memory 343 based on the mask pattern generated along the contour of the object in FIG.

However, the X mark shown in the area number row indicates a case where the area number is not selected by the number selector circuit 342.

Table 3 is a table created by combining the area numbers shown in FIG. 8, the lower left coat, and the upper right code.
Figure 10 shows the code processed according to the method shown in the logic flow diagram of Figure 10;

Table 4 shows the data classified as shown in Table 3.
1 shows a state converted into a chain code that goes around the outline of an image area through the processing shown in the logic flowchart of FIG. 1. The continuation code is also shown to clarify the relationship between the tree table and the third table, and also to indicate the compressed state of the data.

Table 1 Next, the operation of the circuit of the present invention will be explained. In FIG. 3, the binary signals of "1" and "0" obtained by raster scanning the pixels of the binary image to be measured are stored in the one-line memory 31 of the continuation code generation circuit 3I.
1 and the lower line of the 2×2 mask circuit 312 shown in FIG. The one-line memory 311 transfers the signal delayed by one line to the 2×2 mask circuit 312.
Enter it on the top line. In this way, a mask pattern for four adjacent pixels on the binary image is constructed, and a continuation code is output from the continuation code generation circuit 31. The continuance codes are generated sequentially from left to right as shown in FIG.

The control logic circuit 32 includes the contimen 1-
- Input a continuation code from the code generation circuit 31 and apply it to the area numbered circuit 33 and the data memory 343 of the chain code generation circuit 34. The area numbering circuit 33 is a circuit that determines the area number for each area of the object based on the input continuation code.
The area numbers of frame a and frame a of the mask pattern are output to the chain code generation circuit 340 and the number buffer circuit 341. The number buffer circuit 341 temporarily stores the area numbers in frame a and frame 3 while distinguishing between them. The number selector circuit 342 of the chain code Q raw circuit 34 inputs the continuation code from the control logic circuit and selects the number in the frame a of the sofa circuit 341 according to the rules shown in Table 1.
Alternatively, the area number of the frame is selected and outputted to the data memory 343 of the chain code generation circuit 34.

On the other hand, the continuation code counter 345 of the chain code generation circuit 34 is amplified by a signal output from the control logic circuit 32 each time a mask pattern on the contour is detected.

At this time, since the continuation codes numbered 0 and 15 do not represent points on the contour, they are stored in the data memory 343.
To prevent these codes from being stored in Contime 1
- The address signal of the code counter is controlled. Addressed by the output signal of the continuation code counter 345, the data memory 343 stores the continuation code which is the output of the control logic circuit 32.
The mask signal which is the output of the code and number selector circuit 342
The frame a of the pattern or the area number of the frame is stored.

The state of this memory is shown in Table 2. To explain with an example, the mask pattern at 05 on the outline of the object shown in FIG. 6 corresponds to the continuation code I2,
As shown in Table 1, the selection between frame a and frame a is not made. Further, the mask pattern at fl on the outline corresponds to containment code 9, and as shown in Table 2, frame a is selected and the area of this frame a is "2". After scanning one screen, the area number and continuation code stored in the data memory 343 are transferred to the reordering processor 34.
4.

Next, the operation of the reordering processor 344 will be explained.

First, as shown in Figure 9, the rearranging process can be roughly divided into the first process, in which the continuance codes are classified into the lower left code and the upper right code by area number, and the first process, in which the contiment codes are sorted clockwise. and a second process in which the codes are rearranged into chain codes and converted into chain codes.

First, the first process will be explained in detail based on FIG. Separate into side code.

This classification follows the method described below. That is, when no region number is selected, when a pair of concave and concave points that exhibit an upwardly convex shape appears, it is determined that those continuance coats belong to a new region number. If a curved point bear with an upwardly convex shape does not appear, the continuation
If the code is number 8, subsequent containment
The codes are read in order, and the area number to which the first continuation code number 13 belongs is determined as the area number to which it belongs. If the continuance code is number 4 or number 12, this
The codes are checked in order, and the area number to which the first continuation code number 14 belongs is determined as the area number to which it belongs. Also, when a U-shaped curved point pair appears, write the continuation codes numbered 14 and 13 corresponding to the mask patterns at both ends for later integration. When a curved point pair of a shape appears, it is determined that the number of curved point pairs has increased by one, and the first left (111 code and The last right-hand code is considered as a pair, and the four-sided point pairs that appear between these are judged to belong to another area number.In the example shown in Figure 8, the part 8 indicated by the black circle on the circle
The continuation code above 5 belongs to the lower left code of area number 12', and the continuation code above 86, which is marked with a solid line along the contour line, belongs to area number 2'.
It is judged that it belongs to the upper right side code.

In the case of the example shown in FIG. 8, the continuation code on the part 85 indicated by the black circle above the +
The continuance court above 86, which belongs to the lower left code of `` and is marked with a solid line along the demarcation line, has area number ``
It is determined that it belongs to the right-hand code of "2'".

In this way, the continuation code in Table 2 is
Based on the above judgment, the items are classified into the table shown in Table 3.

Next, to explain the second process based on FIG. 11, the codes on the right side of the table with the smallest area number are chained in order until a continuation code indicating a U-shape curvature point and an N-shape curvature point appears.・Convert to code. Next, when a continuation code indicating a bending point of the U shape appears, the left side code of the bending point bear of this U shape is reversely rearranged and converted into a chain code. After converting the opponent's right court into a chain code, continue converting the right side code of the table with the smallest area number into a chain code in order.

Next, if a continuation code indicating a curved point of the 1-shape appears, the right-hand code of the curved point pair of this own shape is sequentially converted to a chain code, and then the curved point pair of this n-shape bear is After rearranging the opponent's left-hand code and converting it into a chain code, continue converting the right-hand code of the table with the smallest area number into a chain code in order until the conversion of the right-hand code is completed. After the process is finished, the left-hand code is rearranged in reverse order and converted into a chain code.

As a result of such rearrangement, the containment code on the table shown in Table 3 is converted into the chain code on the table shown in Table 4.

Next, the containment code is stored in the data memory 343.
When storing information on the lower left side or the upper right side, it is also possible to simultaneously store information on the lower left side or the upper right side, thereby reducing the load on the rearrangement processor 344 and speeding up the rearrangement process.

[Usage form]

(1) Binary image reproduction and contour point coordinate calculation By adding the XY coordinate data memory 32 shown in FIG. 14 to the circuit shown in FIG. 3, image reproduction and coordinate calculation can be performed. That is, this XY coordinate data memory is a circuit that stores the coordinates of a point on the contour indicating the continuation code number 8 when a convex curved point pair above the control logic circuit appears. The point indicated by these coordinates indicates the starting point for creating a chain code that goes around the outline of the object, and the original binary image can be reproduced using the coordinates of this point and the chain code.

It is also possible to calculate the coordinates of each contour point.

(2) Expression of contour shape using polar coordinate system This will be explained based on FIGS. 15 and 16.

Figure 15 shows the outline 130 of the object and a fixed point 131 on the object.
For example, with the coordinates of the center of gravity (Xc, Yc), the coordinates of the starting point 132 (, Xo, Yo), and the fixed point 131 as the origin,
A line connecting a fixed point 131 and a starting point 132 is used as a reference line for polar coordinates, and this shows a state in which a vector 135 moves along the outline of an object without showing polar coordinates.

The horizontal axis in FIG. 16 indicates the sequence number of the contour point or the reference line and the deviation angle 134 of the vector 1♀5, and the vertical axis indicates the length of the hector 135, that is, the point on the contour 130 from the center of gravity 135. The distance p3II is shown. The figure is drawn with the starting point 132 as the origin.

Contour shape information can also be expressed using this method, but if a chain code is used at this time, the distance R4 from a fixed point to a point on the contour can be easily calculated.

That is, the coordinates of a point on any contour are (Xl.

Yi), and if the distance between the starting point and the fixed point is Ro, then RO'
−(Xo In other words, a change of "+1" in the X coordinate is the chain number l, 2.8.
Corresponding to the occurrence of the code, therefore °ζRi 2 − (
l ma , -1) 2 +2 (Xi -
1-1) 2 +2 (-Xi-Xc)+I A change of "+1" in the Y coordinate corresponds to the appearance of the chain code number 3.4.12, thus R42-(RH
-x ) 2 +2 (Yi -Yc) +1 Also, a change of 1-1'' in the Y coordinate corresponds to the appearance of the chain code numbered 7.8.6, thus Ri 2- (RH-1) 2 +2 (-Yi -Yc) +1 allows the distance between the point on the contour and the fixed point to be set.

(3) Calculation of area and circumference Calculation of the area of the object is based on the coordinates of the contour point determined by the method (1) above, and it is determined whether the contour point belongs to the right-hand code or the left-hand code. This can be done by going around the contour by adding when it belongs to the right-hand code and subtracting when it belongs to the left-hand code. In this case, the coordinates may be coordinates with the starting point as the origin.

In addition, the circumference can be calculated using a series of mask patterns obtained by raster scanning the - line.
If the code is number 1.2.4.7.8.11.13.14, it will be measured as "JV2" and if the continuation code is number 3.6.9.12, it will be measured as "1". It can be carried out.

〔Effect of the invention〕

Since the present invention can generate a chain code corresponding to the contour shape of a binary image by raster scanning, real-time processing can be performed with simpler hardware than contour shape measurement using the conventional contour tracking method. .

Also, since the binary image is converted into one-dimensional data, data compression is performed, and therefore it is convenient for signal transmission and storage.

Furthermore, as shape information of the object, we can use
Information on the unevenness of the outer periphery can also be obtained, which is useful for identifying the outer shape during pattern recognition.

[Brief explanation of drawings]

FIG. 1 is a block diagram illustrating the contour tracking method. FIG. 2 is a diagram explaining a chain code. FIG. 3 is a block configuration diagram showing the configuration of a circuit according to an embodiment of the present invention. FIG. 4 is an explanatory diagram of frame names of a 2×2 mask pattern. FIG. 5 is an explanatory diagram showing area numbers of images. FIG. 6 is a diagram showing the state of generation of contour continuance codes. FIG. 7 is a diagram showing how continuance codes are generated by a specific raster scan. FIG. 8 is an explanatory diagram showing area numbers of contours. 9-11 are logic flow diagrams illustrating the operation of the reordering processor. FIG. 12 is a comparison diagram of signals. FIG. 13 is an explanatory diagram of curved point pairs. FIG. 14 is a circle showing the connection between XYli standard data and human output. FIG. 15 and FIG. 16 are explanatory diagrams of contour shape expression using the polar coordinate system. 11...Scanning line, 12.130...Object outline, 3
0... Image signal input terminal, 31... Containment code generation circuit, 32... Control logic circuit, 33... Area numbered circuits, 34... Chain code generation circuit, 35 ...timing generation circuit,
50... Screen, 52... Outline of object, 53...
Object area cutting line, 90...Coordinate data memory, 91
...X coordinate signal input terminal, 92...Y coordinate signal input terminal, 131...Fixed point, 132...Starting point, 133...
...Arbitrary point on area number, 134...Deflection angle, 135
... Extreme IJE I vote vector, 311... One line memory, 312... 2x2 mask circuit, 341...
... Number buffer circuit, 342 ... Number select circuit, 343 ... Data memory, 344 ... Reorder processor, 345 ... Continuation code counter. Patent Applicant Yokogawa Hokushin Electric Co., Ltd. Agent Patent Attorney Naotaka Ide 20 No. 3 Figure M 4 '2 Foil 5 Atsushi Zu 7 Kabuto 8 Figure yl 9 Figure M2O Exit No. 1111fi ¥EJ12 Kasumi Foil Figure 13

Claims (1)

[Claims] (1) An input terminal into which a binary image plane-scanned according to horizontal synchronization and vertical synchronization is input, and a circuit that numbers each pixel represented by the signal of this terminal for each pixel area; Adjacent 4 of the screen represented by the signal
a circuit that generates a containment code for each pixel; a memory circuit that stores the pixel area number and the containment code of that pixel as a set for each pixel; The processor rearranges the addresses of the memory contents of the circuit so that the readout order of the memory contents of the memory circuit is equivalent to a chain code that goes around the outline of the image area. A contour shape information extraction circuit characterized by: