JPS5941230B2 - Contour extraction method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an outline extraction method that allows the outline of a planar figure drawn on a drawing to be extracted at high speed using only one-dimensional operations.

Extracting the outline of a figure is one of the very important elements in Kazukazu processing.

The conventional contour extraction method first selects a certain point on the contour, traces the contour from that point in a fixed direction, and completes the extraction when it has gone around the contour. This type of conventional method requires random access to memory, is difficult to implement in hardware, and requires a significant amount of effort when multiple contours are present on the screen. The present invention is based on the above-mentioned idea, and an object of the present invention is to provide a contour extraction method that enables high-speed extraction of the contour of a planar figure drawn on a drawing using only one-dimensional operations. Therefore, the contour extraction method of the present invention is a contour extraction method that extracts contour lines by processing edge figures in a unified manner.
a window circuit that sequentially scans the edge figure; a start point detection circuit that detects a start point from a pattern within the window circuit; and an edge point detection circuit that detects an edge point other than the start point from a pattern within the window circuit. , a domain table, and a write circuit, and the write circuit is configured to write coordinates of a left point and a right point connected to the start point in the domain table when the start point is detected. When an area is set and the edge point is detected, the coordinates of the edge point are written to either the left point or the right point of the area set for the start point connected to the edge point, and If the same coordinates exist in two different areas in the domain table, the areas are combined into one area and the coordinate values other than the matched coordinates are placed in the left point column and right point column. It is characterized by performing writing processing. Hereinafter, the present invention will be explained with reference to the drawings. FIG. 1 is a block diagram of one practical example of the present invention, FIG. 2 is a flowchart showing the operation of the contour extraction method of FIG. 1, FIG. 3 is a diagram showing edge figure extraction, and FIG. Each mouth has a window.
Figure 5 is a diagram showing the reject point window, Figure 6 is a diagram showing the domain table, Figure 7 is a diagram explaining the priority determination logic, and Figure 8 is a diagram showing the pattern.
The figure shows an example of contour extraction, and FIG. 9 shows an example of changing the domain table. In Figure 1, 1 is n-1
Row buffer, 2 is n row buffer, 3 is n+1 row buffer, 4 is 3x3 window, 5 is reject point.
6 is a start point table, 7 is a priority determination table, 8 is an address calculation circuit, 9 is a table rewriting circuit, 10 is a domain table, and 11 is a column address controller.

In the present invention, edges are extracted from a binary or multivalued planar figure, and a contour is extracted based on the edge figure.

Figure 3 shows edge figure extraction, in which only edges in four directions are extracted.
Edge extraction is not performed for edge-width depressions. Note that the edge figure is a series of △ marks. Further, in the following description of the embodiment, processing after extracting the edge figure (contour points indicated by Δ marks) shown in FIG. 3 will be explained. n
Partial image information of the n-1th row of the edge figure is written into the −1 row buffer 1, partial image information of the nth row is written into the n-row buffer 2, and n+1 row buffer 3 is written with the partial image information of the n-th row of the edge figure.
One line of partial image information is written. The contents of the n-1 row buffer 1, the n-row buffer 2, and the n+1 row buffer 3 are sequentially shifted into the 3.times.3 window 4. 3×3
The black and white pattern within window 4 is converted into address information, and this address information is sent to reject point table 5, start point table 6 and priority determination table 7.

Reject point table (RejectPOin)
tTable) 5 is used to ignore the center point when a pattern as shown in FIG. 4A appears. That is, the reject point table 5 holds in advance a pattern as shown in FIG.
If the black and white pattern in the window 4 is the same as the pattern shown in FIG. 4A, a process such as converting black at the center of the 3×3 window to white is performed, for example. The reject point table 5 is for ignoring a recess having a width of two meshes as shown in FIG. That is, FIG. 5 is an explanatory diagram of the process of ignoring a concavity with a width of 2 meshes when it exists in an edge figure.
If the black and white pattern in the x3 window is the same as the pattern in FIG. 4A when the window is scanned, a process is performed to reject the center point of the window. Therefore, as mentioned earlier, the horizontal dent of one mesh is ignored when extracting edges, but when combined with this,
In the present invention, depressions up to two mesh widths are ignored. Start point table 6 is set when a pattern like the one shown in Figure 4 appears in the 3x3 window.
The center point is taken as the starting point. That is, the start point table 6 holds in advance the pattern shown at the beginning of Figure 4, and if it is the same as the black and white pattern in the input 3x3 window, the start point table 6 will start from the center point of the 3x3 window.・Point. In the priority order determination table 7, when an edge point exists at the center of the 3×3 window 4, the mesh points 1, 2, and 3 shown in FIG.
, 4 are black or not, and if there are a plurality of them, one of them is selected according to the priority order. That is, FIG. 7 is an explanatory diagram of the process of determining the next point to which the center point of the 3X3 window can be moved when the edge point at the center of the 3X3 window is connected to the start point. If the edge point at the center of the 3x3 window is connected to the left side of the start point, one black point is selected according to the priority order for the left point shown in Figure 7. If the edge point at the center of the three windows is connected to the right side of the start point, one black point is selected according to the priority order for write points shown in FIG. 7C.

In addition, in Figure 7 (opening and c), the numbers within the writing frame indicate the priority order. For example, if the edge points at the center of a 3×3 window are connected to the left of the start point, and mesh points 2 and 4 are black points, the black point of mesh point 2 is selected. FIG. 7B explains the left point priority order in more detail, and FIG. 7E explains the right point priority order in more detail. 3
There is an edge point at the center of the ×3 window, and a mesh point is 1.
, 2, 3, and 4, the central edge point becomes the integration point.

The integrated point will be explained in detail later. The address calculation circuit 8 calculates the coordinate values of the selected black point in the priority order determination table 7. The table rewriting circuit 9 rewrites the domain table 10. The domain table 10 has a plurality of areas as shown in FIG. 6, and each area is provided with a column for writing an area identification number, left point coordinates, right point coordinates, and area end flag. ing. That is, FIG. 6 shows that when a start point is detected, domain table 1 is
FIG. 2 is an explanatory diagram of an area for writing the coordinates of a left point and a right point connected to the start point, which are set in 0 with coordinate information (area identification number) of the start point. When the start point table 6 detects a start point, the start point flag is turned on, thereby preparing a new area for that start point. The column address controller 11 specifies columns of data read from the n-1 row buffer 1, the n-row buffer 2, and the n+1 row buffer 3. Next, the operation of the apparatus shown in FIG. 1 will be explained with reference to FIG. 2.

The apparatus shown in FIG. 1 performs contour extraction by scanning an edge figure as shown in FIG. 3 from top to bottom using scanning lines running from left to right. The outline extraction process in the n-th row will be described below. One buffer 1, 2, and 3 are prepared, and data of n-1 row, n row, and n+1 row are input into each buffer.

2. Extract data in order from the nth row and first column, and check whether the nth row and mth column is an edge point.

If it is not an edge point, proceed to column m+1. If 3n rows m and 1 tears are edge points, a 3x3 window 4
Look at it to determine whether it is a reject point or a starting point.

4 In the case of Reject Point, proceed to m + 1 and 1 tears, and in the case of Start Point, perform the following step 5.

Rewrite the domain table 10.

To rewrite table 10, it is checked whether an edge point exists in the shaded area in FIG. For the determined edge point, the current point N,
Calculate the coordinate value from m. In the case of a starting point, a new column is provided at the end of the domain table 10,
Write the coordinate values of the left point and right point here. If it is determined otherwise in step 54, check which point in the domain table 10 the current point N,m matches or not, and if it is a left point, use the priority determination logic for the left, and select the right point. If it is a point, use the right point priority determination logic to determine the next left point or right point, obtain the coordinate values of the newly determined left point or right port, and store them in the domain table 10. Rewrite the coordinate values of the left point or right point with new coordinate values.

◇ If no edge point exists in the shaded area in Figure 7A, the current points N and m are called integration points.

An integration point is a point where contours tracked separately meet. The line with this integration point is
When the processing of one line is completed, the following processing (71a) and (71b) are performed. If there is no integration point, move to the next line. (71a) Compare the left point and right point in the domain table 10 and search for the same point.

If the same points are found, the two areas are combined into one, and the domain table 10 is rewritten with the same points as new left points and right points. Note that when the left point and right point are completed within the same area, it means that the tracking of that area has been completed, so a tracking end flag is set. (71b)
If there is no left point and right point that have the same coordinate values, compare the left point and right point, and if the same points are found, combine the two areas into one and create a new area. Write the coordinates of the upper left point in the left point column, and write the coordinates of the lower left point in the right point column.

When processing for all 8 lines is completed, processing for the screen is completed.

FIG. 8 shows an example of contour extraction, and FIG. 9 shows an example of changing the table.

That is, FIG. 8 is a diagram showing the result of performing the above-described contour extraction process on the edge figure shown in FIG. , arrows indicate the direction of processing. Figure 9 5
8 shows the contents of the domain table 10 before integration and the contents of the domain table 10 after integration at the time when processing up to the fifth line indicated by symbol 8 in FIG. 8 is completed, and FIG. 8 indicates the contents of the domain table 10 before integration and the contents of the domain table 10 after integration at the time when the processing up to the 13th line, which is similarly indicated by the symbol in FIG. 8, is completed, FIG. 9 shows the domain table 10 before integration when the processing up to the 16th line indicated by symbol 6 in FIG.
, and the contents of the domain table 10 after integration. In addition, the start point, left point, and right point in the process from the start to the end of contour tracing
The coordinates of the points and the integrated points are sequentially recorded in another file (not shown). As is clear from the above description, according to the present invention, contour extraction can be performed only by one-dimensionally manipulating edge figures, and high-speed processing is possible.

[Brief explanation of drawings]

Fig. 1 is a block diagram of one practical example of the present invention, and Fig. 2 is a block diagram of one practical example of the present invention.
Flowchart showing the operation of the contour extraction method in Figure 3. Figure 3 is a diagram showing edge figure extraction. Figure 4 is a diagram showing the wint pattern for A and mouth. Figure 5 is for explaining the reset point window. Figure 6 shows the domain
FIG. 7 is a diagram illustrating the priority determination logic, FIG. 8 is a diagram illustrating an example of contour extraction, and FIG. 9 is a diagram illustrating an example of changing the domain table. 1...n-1 line buffer, 2...n line buffer, 3...n+1 line buffer, 4...
...3x3 window, 5...Reject...
Point table, 6...Start point table, 7...Priority determination table,
8...Address calculation circuit, 9...Table rewriting circuit, 10...Domain table,
11...Column address controller.

Claims (1)

[Claims] 1. A contour extraction method that extracts a contour line by processing an edge figure in a unified manner, which includes a window circuit that sequentially scans the edge figure, and a start point detection circuit that detects a start point from a pattern in the window circuit. , an edge point detection circuit for detecting an edge point other than the start point from a pattern in the window circuit, a domain table, and a write circuit; Set an area in the domain table to write the coordinates of the left point and right point connected to the start point, and when the edge point is detected, write the coordinates of the edge point to the start point connected to the edge point. Write processing is performed to either the left point or the right point of the area set for the area, and if the same coordinates exist in two different areas in the domain table, the area is merged into one area. A contour extraction method is characterized in that a process is performed in which coordinate values other than the matched coordinate values are written in a left point column and a right point column.