JPH01245366A - Labeling processing system - Google Patents

Abstract

PURPOSE:To prevent continued binary images from being separated to attain labeling by executing the integrating processing of inter-label connecting information stored in a table between lines of raster scanning. CONSTITUTION:An image labeled by primary labeling processing is stored in an image memory 21 and delayed for one image. After passing the final line of the image through a primary labeling part 10, the label integration of the line is executed and the integrated information is copied on a table LRAMB 22 by a copying circuit 23. At the time of executing the primary labeling processing of the succeeding input image, the image in the memory 21 is extruded like a pipeline. At that time, the value of the image already labeled by the primary labeling is used as an address to access the LRAMB 22 and data are outputted from the LRAMB 22. Thereby, labeling can be executed without generating the separation of the binary images.

Description

Detailed Description of the Invention [Objective of the Invention] (Industrial Application Field) The present invention is directed to a plurality of arbitrary shaped Regarding a labeling processing method that assigns a number (label) to a connected region in order to identify the connectivity of a pattern (object), shape, area, etc., it is particularly applicable to continuous binary images at video rate. The present invention relates to a labeling processing method that allows labeling to be performed without causing separation.

(Prior Art) As a conventional labeling processing method, there is one in which connected patterns are recognized while raster scanning continuous images sent at a video rate of 30 images per second. In this method, for example, two consecutive rows of 2 pixels by 3 pixels horizontally are cut out in raster scanning form from the original image as a logical mask, and processing such as labeling and connectivity recognition is performed on the cut pixels. That is, the label value of the pixel is determined and assigned based on the label value of the pixel that is adjacent to (or connected to) the pixel of interest among the extracted pixels and has already been assigned a label. This series of steps is sometimes called primary labeling.

Furthermore, in this case, when pixels with two different label values are connected (also called merging of labels), that information is written in the inter-label connection information table,
After the first raster scan of the entire image is completed, the inter-label connection information in the inter-label connection information table is integrated. After that, the temporary label image given by the primary labeling process is raster-scanned again, and the pixels on the connected pattern are processed to have the same label value based on the integrated table described above to obtain the final label image. . This procedure is sometimes called secondary labeling.

A specific example of the conventional labeling processing method using the above primary labeling and secondary labeling is disclosed in Japanese Patent Application Laid-Open No. 62-46381.
This will be explained with reference to FIG. 8 shown in the publication. That is, as shown in FIG. 8, even if the object is the same, different label values are attached to the original image as separate objects.

This is because it is later determined that the objects are the same object at the point where the separated label values join (joint point), so the relationship between the separated labels is stored in the label relationship table LRT at this point.

The process of creating the label relationship table LRT while attaching such temporary labels is called temporary labeling (primary labeling). Following this temporary labeling, the label relationship table LRT is analyzed by software to create a label conversion table LTT for attaching the final label. This is called table conversion. Next, this label conversion table L
Convert the temporary label to a final label based on TT. This is called relabeling (secondary labeling). Here, in order to obtain the final label image, the temporary label obtained by the temporary labeling process is temporarily stored in the image memory to obtain the temporary label image, and then this temporary label image is used to create a table in the label conversion table LTT. The final label image is obtained by reference.

In the conventional method described above, if the pattern is relatively simple as shown in Figure 8, it is possible to accurately assign a label value to each pattern without causing separation, but if it is a complex connected pattern, it is possible to give a label value to each pattern. may occur. Hereinafter, examples in which separation occurs will be described with reference to FIGS. 9 to 13. That is, as shown in FIG. 9, there is one connected pattern FAT in the image FM, and an example will be described in which this is labeled with the logical mask MA. FIG. 10 is an enlarged view of FIG. 9, and the final label image should originally have the label value of the connected pattern FAT shown in FIG. 11 as 1, for example, and the surrounding label values as 0. be.

Here, as shown in FIG. 12, in which rows 0 to 12 are defined for convenience, the connection pattern FA from rows 1 to 12 is
Label values 1 to 8 are assigned to each area pixel of T by recognition of connectivity in the logical mask area, and this label relationship is the 13th
Written in the label relationship table LRT shown in Figure (a),
A temporary label will be created. Then, by performing the integration process, the label conversion table LTT shown in FIG. 13(b) is completed. The final label image based on this label conversion table LTT becomes as shown in FIG. 14, and as shown in the figure, label values 1 and 3 are separated even though they should originally be connected.

This kind of separation is due to recognition of merged label pairs on one label relationship table, so as a measure to prevent this from occurring, it is necessary to memorize the label relationship table created after temporary label processing. In addition, this can be realized by storing all merged label pairs (for example, 2-1) corresponding to positions (addresses) on the image memory.

However, if such a method is adopted, the result is the same as the existence of two label relationship tables, and the amount of comparison calculations for integration increases significantly. only performs comparison operations. ) and the processing was complicated, making it impractical.

As a method other than the above, there is a method that focuses on the fact that separation is a result of recognizing confluent label pairs on a single label relationship table, and expresses label relationships using a plurality of tables. However, like the above-mentioned method, this method also had the problem of significantly increasing the amount of comparison calculations for integration and complicating the processing, and was not practical.

(Problems to be Solved by the Invention) As mentioned above, in the conventional technology, the amount of comparison calculation for label integration increases significantly and the processing becomes complex, so that the processing time is required for both methods. Therefore, it is not easy to label continuous binary images at video rate.

SUMMARY OF THE INVENTION An object of the present invention is to provide a labeling processing method that can label continuous binary images at a video rate without causing separation.

[Structure of the Invention] (Means for Solving the Problems) The present invention is characterized by taking the following measures in order to solve the above problems and achieve the objects. That is, in the present invention, for a pixel column obtained by raster scanning a binary image, a logical mask spanning two consecutive rows is used to label the pixel of interest in the local area, and to calculate the label value. , is characterized in that it allocates it to the image memory and writes it in a table, identifies the connectivity with surrounding pixels, and updates the label value written in the table based on the recognition, in this case, The link amount information of labels held in the table may be integrated and processed between rows of raster scanning.

(Function) According to this configuration, the label value is assigned to the image memory by labeling each pixel of interest by raster scanning of the logical mask, and the label value is updated by writing to the table and connectivity identification, and each row Since the subsequent labels are integrated at once, it is possible to label a binary image having an arbitrary-shaped pattern without causing separation. In addition, the process of allocating to the image memory, writing to the table, and updating the label value by identifying connectivity is simply data write process and judgment process for the number of pixels, so it is difficult to compare it with conventional comparison operations. It is simple and requires no processing time.

Therefore, by allocating and processing the label values held in the table in the image memory between the rows of raster scanning, it is possible to label consecutive binary images at the video rate without causing separation. It is something that can be done.

(Example) An example of the labeling processing method according to the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the hardware configuration of a processor that can execute the method of this embodiment, FIG. 2 is a diagram showing a logical mask, and FIG. 3 is a diagram showing a labeling algorithm in this embodiment. FIG. 4 is a diagram showing an inter-label connection information integration table and its integration algorithm, and FIGS. 5 to 7 are diagrams showing specific examples of labeling for connection patterns.

In the processor shown in FIG. 1, connectivity determination and labeling operations are performed while raster scanning the image memory FM using the logical mask MA shown in FIG. In reality, the processor shown in Figure 1 uses hardware-based vibration line processing for continuous images sent at video rate.
Primary labeling process by the next labeling unit 10, that is, 2
A logical mask (local mask) covering two consecutive rows for the pixel column obtained by raster scanning the value image.
A label is attached to the pixel of interest P in the local area by MA, and the label value is assigned to the image memory 21 and written to the table 13, and the connectivity with surrounding pixels is identified and based on the recognition. The process of updating the label value written in the table 13, and the subsequent secondary labeling process by the secondary labeling unit 20, that is, the table LRA is updated between the lines during the blanking period of raster scanning.
This process is performed in two steps, including the process of integrating the link information between the label values held in M^13.

Also, x, y, z in the logical mask MA in FIG.

U is a variable representing the label value at the point (pixel) that they are located, and P is the value of the binary pattern at that point (1 or 0)
Let be a variable that represents Further, the term "concatenation" will be used hereinafter to mean 8-concatenation. In FIG. 3, a table indicating connection information is represented by a one-dimensional array LRAM (.).

The table LRAMA13 is implemented using a RAM (random access memory) or the like. Also, label (number)
Let the counter value be LC, V-wax (X, U), Z'-LRAM (Z)
The symbol "-" represents assignment to a variable.

First, the primary labeling section 10 will be explained.

That is, the processing in this primary labeling unit 10 is 8
Connectivity judgment and labeling are performed according to the algorithm shown in FIG. 3 by dividing the cases accordingly.

The algorithm shown in FIG. 3 attaches a label counter value LC that is incremented in the order of appearance to the pixel of interest P according to steps (1) to (8), and writes the label value to the table LRAMA13 as its connection information. It looks like this. In this case, if it is determined that the pixel of interest P and surrounding pixels adjacent to it have connectivity, the label value of the pixel of interest P is transferred to the connected pixels (which have already been assigned label values). Processing is performed to set the label value of ).

The above algorithm uses pixels X, Y, Z,
Labeling of P and U is performed by the logic circuit 12.

In addition, for each row scan, the link information table LRAMA13
Integration of label information is performed. That is, this label integration processing algorithm is based on the label integration circuit 15.
This is done by executing steps St to S4 in FIG. 4(b) on the table LRAM^13 shown in FIG. 4(a). Here, the maximum value of the label value is N, and k is the address value to the table LRAMA13. In the integration process performed by this algorithm, label integration is performed by using the label value of the created table L RAM A13 as an address, incrementing the address by 1, and sequentially resubstituting the address as the label value. It becomes like this.

Next, the secondary labeling section 20 will be explained.

That is, the images labeled by the above-described primary labeling process are stored in the image memory 21 and delayed by one image. After the last line of the image passes through the primary labeling unit 10, label integration is performed on that line, and at the same time, the integrated information is stored in the table LRAMs by the copy circuit 23.
Copied to 22. Thereafter, when the next input image is subjected to primary labeling processing, the image in the image memory 21 is pushed out in a pipeline manner. At this time, the table LRAMe22 is accessed using the value of the image already labeled in the primary labeling as an address, and the data is output. That is, the table LRAMF122 becomes a kind of density conversion table, and through this conversion, each connected component becomes the same label value.

The above specific examples are shown in FIGS. 5 to 7. That is,
The connected pattern FAT is labeled line by line from Figure 5 to Figure 6, and in the 12th line, integration processing is performed by determining connectivity, so the final label image is output as shown in Figure 7. It turns out.

As described above, according to this embodiment, label values are assigned to the image memory 21 by labeling each pixel of interest by raster scanning of the logical mask MA, and label values are assigned to the image memory 21 by writing to the table LRAM and identifying connectivity. Since the updating is performed all at once, the connectivity is not classified on the table LRAMA13 as in the conventional method, so that it is possible to label the binary image without causing separation. In addition, allocation processing to the image memory 21, table L
Writing to RAM A 13 and updating label values by identifying connectivity is simply data writing and judgment processing for the number of pixels, so it is simpler and requires less processing time than conventional comparison operations. It is not necessary. Therefore, it is possible to label continuous binary images at the video rate without causing separation.

Furthermore, by using the row spacing of raster scanning, that is, the blanking period, to integrate the link information between label values held in the table LRAM, it is possible to process the - layer in a short time.

The present invention is not limited to the above embodiments, but can be implemented with various modifications without departing from the gist of the present invention.

[Effects of the Invention] As described above, according to the present invention, one table is used to perform label value assignment, writing, and label value update processing through connectivity identification, so that at a video rate, Furthermore, it is possible to provide a labeling processing method that can label continuous binary images without causing separation.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the 71-ware configuration of a processor that can execute an embodiment of the labeling processing method according to the present invention, FIG. 2 is a diagram showing a logical mask, and FIG. Figure 4 is a diagram showing the labeling algorithm, Figure 4 is a diagram showing the inter-label linkage information integration table and its integration algorithm, Figures 5 to 7 are diagrams showing specific examples of labeling for connection patterns, and Figure 8 is the conventional Figures 9 to 17 showing examples of bones are diagrams showing problems with the conventional method. 10... Primary labeling section, 11... Register group,
12... Labeled logic circuit, 13... Linked information table (LRAMA), 14... Line buffer, 15... Label integration circuit, 20... Secondary labeling unit, 21... Image Memory, 22... Linkage information table (LRAMB), 23... Copy circuit. Applicant's representative Patent attorney Takehiko Suzue Figure 3 Figure 4 Figure 8 MA Figure 9 (a) Figure 13 (b)

Claims (2)

[Claims] (1) For a pixel column obtained by raster scanning a binary image, a logical mask spanning two consecutive rows is used to label the pixel of interest in the local area, and the label value is stored in the image memory. A labeling processing method, characterized in that the label value written in the table is updated based on the recognition by identifying connectivity with surrounding pixels and writing in a table. (2) The labeling processing method according to claim 1, wherein the inter-label connection information held in the table is integrated between the rows of raster scanning.