JP2924744B2 - Image region extraction method and device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for extracting a required target area in an image, a technique for creating a mask for the target area, and a technique for cutting out the target area in various image processing apparatuses. A mask is a data used to cut out a partial area in one image and combine it with another image or combine it with a document, and specify which pixels of the original image are required in pixel units Means

[0002]

2. Description of the Related Art Conventionally, this type of image area extracting apparatus uses a lightness, hue, saturation, or RG of an area to be extracted.
A process of designating the B value and extracting only pixels within a certain threshold range from the designated value has been performed.

An embodiment of a conventional image region extracting apparatus will be described with reference to FIGS.

An embodiment of the conventional apparatus includes an image input device 201 such as a scanner, a monitor 202 for displaying an image, and a position specifying means 20 for an operator to specify a reference point in a target area.
4. Reference pixel value storage buffer 206 for storing pixel values of reference points, input image memory 203, mask image memory 20
7, threshold value processing means 205 for comparing the RGB value and the pixel value of the reference point, and an output device 208 for outputting the generated mask data.

[0005] Image data input from the image input device 201 is held in an input image memory 203 and simultaneously displayed on a monitor 202. The operator wants to mask while viewing the image displayed on the monitor 202 (target area)
One of the points is designated as a reference point using the position designation means 204 (step A101). Examples of the position specifying unit 204 include a mouse, a tablet, and a keyboard.
Next, the pixel value (R0, G0, B0) of the reference point is extracted from the input image memory 203, and is stored in the reference pixel value storage buffer 20.
6 (step A102).

The image scanning position is reset to an initial value (step A103), and the image is scanned from one end to the other by using the data of the input image memory 203.
At 05, for each pixel value (r, g, b), (R
(0, G0, B0). For a pixel determined to be extracted, TRUE is set in the mask image memory 207 at the corresponding position,
Otherwise, set FALSE (steps A104, A105, A106, A107, A108,
A109). After scanning to the end of the image, the contents of the mask image memory 207 are output to the output device 208, and the process ends. Generally, the contents of the mask image memory 207 are
E and FALSE. In the following description, TRUE indicates that the pixel is extracted, and FALSE indicates that the pixel is not extracted.

[0007] As a determination method of the threshold value processing means 205, assuming that Th is a certain threshold value, (R0-Th <r <R
0 + Th), (G0−Th <g <G0 + Th) and (B0−Th <b <B0 + Th), TRUE, and FALSE otherwise. Further, a method of performing threshold processing after converting into hue, saturation, and brightness values instead of RGB values, or (R0−r) ² + (G0
−g) ² + (B0−b) ² <Th.

As an example of a conventional apparatus, Japanese Patent Laid-Open Publication No.
29487 and JP-A-5-28263.

[0009]

In a conventional image region extracting apparatus, if a pixel having the same color exists in a region other than the region to be extracted, that portion is similarly removed, so that the target region intended by the operator can be accurately determined. There was a problem that it could not be extracted. Further, when the region to be extracted is composed of a plurality of color pixel patterns, there is a problem that the region cannot be extracted.

FIG. 12 shows partial areas 101 and 10
2, a partial area 103 and a partial area 104. The partial area 101, the partial area 102, and the partial area 103 have substantially the same pixel value in each area. The pixel values of the partial area 101 and other areas are significantly different. Partial area 10
Although the position of 2 and the partial area 103 are apart from each other, the pixel values in the area are assumed to be substantially equal. Also, the partial area 10
4 is composed of a pattern of pixels of two colors, a first color pixel 105 and a second color pixel 106.

In a conventional image area extracting apparatus, an area such as the partial area 101 having a pixel value different from that of another area can be extracted. However, partial area 1
02 is to be extracted, the partial area 1 having the same pixel value
03 is also extracted at the same time. In addition, there is a problem that, in most cases, only one of two color pixels can be extracted from an area configured with a pattern such as the partial area 104. Of course, if masking is performed using both the color of the first color pixel 105 and the second color pixel 106, extraction can be performed to some extent. However, not only is the burden on the operator, but also the extraction accuracy is extremely poor. Was.

[0012]

In order to solve the above-mentioned problems, an image area extracting apparatus according to the present invention comprises an input image memory for holding input image data, and a monitor for displaying image data and area extraction results. Position specifying means for specifying a target region to be extracted, a reference point feature storage buffer for holding a pixel feature of a reference point for region extraction, a reference point position storage buffer for holding the position of the reference point for region extraction, Neighboring pixel searching means for searching for an address of a pixel adjacent to a certain pixel;
A determined pixel storage buffer for storing a pixel to be masked, a pixel characteristic amount comparison unit for comparing characteristic amounts of two pixels, a pixel characteristic amount extraction unit for calculating a characteristic amount of a certain pixel, and holding a region extraction result A neighboring pixel searching means, wherein the mask image memory is
Initialize with FALSE, and execute the
Read the position of the determined pixel from the
Direction (here, the direction from left to right is defined as + X direction)
To find the boundary pixels of the target area, and
Stored as a can pixel, and stored on the mask image memory
Set TRUE at the position of the scan pixel, and
Elementary + Y direction (here, the direction from bottom to top is + Y
Direction) (hereinafter referred to as pixel P1).
And the pixel feature amount of the pixel P1 and the pixel feature amount of the reference pixel
Is less than or equal to a predetermined threshold (hereinafter referred to as maskable).
And the pixel adjacent to the pixel P1 in the −X direction is
If masking is not possible, the pixel P1 is replaced with the determined pixel
Stored in the storage buffer, adjacent to the scan pixel in the -Y direction
For a pixel (hereinafter, referred to as a pixel P2), the pixel P2
Maskable and adjacent to the pixel P2 in the -X direction
When the pixel is not maskable, the pixel P2 is determined.
+ X of the scan pixel
Determines whether adjacent pixels in the direction can be masked, and
If determined to be possible, the position of the scan pixel is set to + X
Move one pixel in the direction
The same processing is repeated for the determined pixel position storage buffer.
Check if there are any judged pixels that have not been read out yet.
In addition, if there is a determined pixel that has not been read yet,
The same processing is repeated by reading the determined pixel.
It is characterized by returning .

Further, the image region extracting method according to the present invention includes a step of designating a reference pixel in a target region to be extracted, and one or more different pixel characteristic amount extracting steps of extracting a characteristic amount of an arbitrary pixel. Memory for storing mask images (hereinafter
(Referred to as a mask image memory below) with FALSE
And a buffer for storing the determined pixel position (hereinafter referred to as a buffer).
Below, it is called the determined pixel position storage buffer).
Reading the elementary position (step B21);
From the determined pixel in the -X direction (here, the direction from left to right
Direction to + X direction) to find the boundary pixel of the target area.
(Step B22) and scanning the boundary pixels.
Storing as a pixel (step B23);
TRUE at the position of the scan pixel on the mask image memory
Setting step (step B24) and scan image
Elementary + Y direction (here, the direction from bottom to top is + Y
Direction) (hereinafter referred to as pixel P1).
And the pixel feature amount of the pixel P1 and the pixel feature amount of the reference pixel
Is less than or equal to a predetermined threshold (hereinafter referred to as maskable).
And the pixel adjacent to the pixel P1 in the −X direction is
If masking is not possible, the pixel P1 is determined as the determined pixel position
Storing in the information storage buffer (step B2
5, Step B26) and the neighbor of the scan pixel in the −Y direction.
For a contact pixel (hereinafter, referred to as a pixel P2), the pixel P2
Can be masked, and is adjacent to the pixel P2 in the -X direction.
If the contact pixel cannot be masked, the pixel P2 has been determined.
Storing in the pixel position information storage buffer (step
B27, step B28) and + X direction of the scan pixel
Determining whether the neighboring pixel in the direction is maskable
(Step B29) and masking is possible in Step B29
If determined, move the position of the scan pixel in the + X direction.
Move one pixel and move from step B24 to step B29.
(Step B30) and the determined image
Determined pixels that have not yet been read into the elementary position storage buffer
Check if there is any
If there is a prime, Step B21 to Step B30
(Step B31) and extracting a connected region having a uniform pixel feature amount including the reference pixel.

The principle of the image region extracting method and apparatus of the present invention will be described.

FIGS. 4, 5, 7 and 8 are flowcharts showing the principle of the present invention. 9 and 10 are enlarged views of the input image data. The image data is digitized by a scanner or the like, and each pixel is represented as a small rectangle surrounded by a grid in the figure.

The processing flow of the present invention will be described with reference to FIG. First, the operator specifies a pixel in the target area to be extracted (step A1). The pixel specified by the operator is set as a reference pixel. Then, the position of the reference pixel is registered as a first determined pixel position (step A).
2, A3), and at the same time, the pixel feature amount of the reference pixel is calculated (step A4). Next, a connected area extraction process is performed to extract a target area (step A5). In the connected region extraction processing, a connected region having a uniform pixel feature amount including the reference pixel is extracted. The connected region is a region formed by a set of adjacent pixels having substantially the same pixel feature amount. Further, expansion and contraction processing is performed on the mask image data to correct the shape of the area (step A6). Finally, the mask image data is output (step A7).

The processing flow of the present invention will be described with reference to FIG. First, the operator specifies a pixel in the target area to be extracted (step A21). The pixel specified by the operator is set as a reference pixel. Then, the position of the reference pixel is registered as a first determined pixel position (step A).
22, A23), and calculate the pixel feature amount of the reference pixel (step A24). Next, a connected area extraction process is performed to extract a target area (step A25). In the connected region extraction processing, a connected region having a uniform pixel feature amount including the reference pixel is extracted. The connected region is a region formed by a set of adjacent pixels having substantially the same pixel feature amount. Then, the obtained mask data is examined to determine whether or not the mask has been normally created (step A25).

One example of the mask creation determination result is to check whether n or more pixels of the mask TRUE have been extracted. n is usually 2, but may be a larger number. In this case, when n or more TRUEs are present, the mask creation is considered to be successful, and when less than n, the creation is considered to have failed. There is also a method of determining in advance how many times the feature amount should be changed before processing. Also, when all the pixel feature extraction means in the apparatus have been executed,
It shall be considered successful and proceed to the next step.

If the mask is not created normally, the pixel feature to be used is changed (step A27), and the pixel feature of the reference pixel is calculated again (step A23).
Repeat from Mask creation result determination processing (step A2)
If it is determined in 6) that the process has succeeded, the process proceeds to the next step. Next, expansion and contraction processing is performed on the mask image data to correct the contour of the area (step A28). Finally, the mask image data is output (step A29).

The output data in the present invention is TRUE,
FALSE two types of mask data, TRUE indicates that the extraction area is the extraction area at the position of the input image, and FALSE indicates that the extraction area is not extracted. However, data to be output is not limited to this, and pixel values are set to TRUE and FALSE to 0 and 25.
For example, there is a method of expressing the value with a value of 5, mapping the pixel value of the input image to the position of TRUE and outputting the result.

Here, the pixel feature value in the present invention will be described. The pixel feature amount of the pixel X is V (X), and V (X) may be a single value or a plurality of values representing the characteristics of the pixel. As an example of the pixel feature amount: The pixel value is used as it is. For a gray image, gray scale is used, and for a color image, RGB values are used as they are. HSV values (hue, saturation, lightness values). Several calculation methods for obtaining these have been conventionally proposed, and can be calculated from RGB values according to, for example, Ref. [1]. XYZ values, L ^* a ^* b ^* values, L ^* u ^* v ^*
Values can also be used. These are generally widely known in the standards set by the CIE (International Commission on Illumination), and the calculation method is described in detail in reference [3] and the like. A value obtained by applying a local filter to the vicinity of a pixel. A value obtained by applying a local filter after performing color conversion such as HSV and XYZ. and so on.

A method of applying a local filter to the vicinity of a pixel will be described with reference to FIG. Local filter 7
0 indicates a scene where the local filter is applied at the target pixel position 71. i represents the position in the x direction of the image, and j represents the position in the y direction. In the figure, the size of the local filter is 3
Although it is × 3, it is generally represented by M × N, and its size is arbitrary. The pixel value at position i, j is
Assuming that (i, j), when a 3 × 3 local filter is applied, the output I ′ (i, j) is I ′ (i, j) = I (i−1, j−1) * a (1,1) + I (i, j-1) * a (1,2) + I (i + 1, j-1) * a (1,3) + I (i-1, j) * a (2,1) + I (i, j) * a (2,2) + I (i + 1, j) * a (2,3) + I (i-1, j + 1) * a (3,1) + I (i, j + 1) * a ( 3,2) + I (i + 1, j + 1) * a (3,3) (1) If there are three pixel values like RGB,
By applying a local filter to each of R, G, and B, three characteristic amounts of R ', G', and B 'are obtained.

One of the effective local filters in the present invention is a low-frequency filter. The numerical value of the low-frequency filter 72 shown in the figure is an example, and the low-frequency filter described here is a filter that acts to filter only low-spatial-frequency components after Fourier-transforming an image. Collectively. Such a low-frequency filtering filter can actually be represented as a local filter that performs weighted averaging of pixels around the pixel of interest. As an example of these, the low-frequency filter 72 is a local filter that takes a uniform average of peripheral pixels.

The low-frequency filtering filter includes a first color pixel 105 and a second color pixel 10 such as a target area 104 in FIG.
This is effective for extracting an area in which 6 is repeated as a fine pattern. Since an area having such a pattern appears as a mixture of both colors when displayed on a monitor, such an area is often used to display an arbitrary color on a monitor having a limited number of colors, such as 16 colors. An image is created and used. For example, white and black patterns can represent gray halftones, and red and blue patterns can represent purple.

The first color pixel 105 is red and the second color pixel 106
Is blue, and when displayed on a monitor,
It looks almost purple. In order to reproduce this effect, a low frequency filter is used. By using a low-frequency filtering component as a feature value of a certain pixel of interest in the target area 104, a weighted average value of red and blue is obtained, which is almost the same as a purple pixel value. In other words, by using the low-frequency filtering filter, even in a region configured by a fine pattern, a mask creating operation can be performed with a color as seen by a human.

Other examples of effective local filters include a high-frequency filter 73 and an edge strength extraction filter 74. In addition, the high frequency filter 73 in the figure,
The numerical values shown in the edge strength filter 74 are just examples, and several filters are described in literature [2] and the like. The high frequency filter and the edge strength extraction filter are effective for extracting regions having the same color but different textures.

Next, a connected area extraction process (step A5,
The contents of A25) will be described.

FIG. 7 shows a reference technique of the connected area extraction processing.
There is. First, all output mask image data is FAL
Initialize to SE (step B1). One determined pixel position is read (step B2). Note that the location information item once read must be deleted from the registration. TRUE is set at the position of the determined pixel in the mask image (step B3). Next, a pixel near the determined pixel is acquired as a mask candidate pixel (step B4).
For each mask candidate pixel, the pixel feature amount is compared with the reference pixel feature amount, and if masking is possible, the pixel position is registered as a determined pixel position (steps B5 and B).
6, B7, B8). Then, the processing from step B2 is continued until there are no more determined pixels.

An example of a method of determining whether masking is possible will be described below. Assuming that N feature amounts are used, N reference pixel feature amounts are defined as Vi = (V1, V2, V3... V
n), the characteristic amount of the target pixel is expressed as Xi = (X1, X2, X3
.. Xn). If the allowable range of each feature is Ti = (T1, T2, T3... Tn),
It can be determined based on the following formula.・ If all i from 1 to N, | Xi−
If Vi | <Ti holds, the pixel is maskable (TRUE); otherwise, FALSE If the feature value is an RGB pixel value, the respective values of R, G, and B of the target pixel are set to the reference pixel. From the value, TRUE if the value falls within a certain range, FALSE otherwise.

The actual processing will be described with reference to FIG. The reference pixel 53 (pixel indicated by A in the figure) is
It is registered in advance as the first determined pixel position 54. After setting TRUE to the first determined pixel position 54 in the mask image data, four neighboring pixels, a, b, c, and d, are obtained as mask candidate pixels, and the respective pixel feature amounts V (a), V (b), V (c), V (d)
Is compared with the reference pixel feature amount V (A). If the feature value is close to the reference pixel value, the pixel is a maskable pixel and is registered as a determined pixel. In this example, a, b, c, d
All are registered as determined pixels.

Next, it is assumed that the pixel read from the determined pixel buffer is c. Since c is located at the second determined pixel position 55, TRUE is set to the second determined pixel position 55 in the mask image memory. The mask candidate pixels near the pixel c are A, e, f, and g, but A is not determined because A is already registered in the mask image. e,
Comparing the feature values of f and g, e and f become TRUE and can be masked.
ALSE. As described above, by expanding the search pixel from the vicinity of the reference pixel, a target region including the reference pixel can be extracted.

The connected region extraction processing (step A) of the present invention
5, A25) is shown in FIG. First, all output mask image data is initialized to FALSE (step B20). One determined pixel position is read (step B21). The input image is searched in the −X direction from the determined pixels, and a pixel serving as a boundary of the target area is found (step B).
22) The boundary pixel is set as a scan pixel p (step B23). Next, TRUE is set at the position of the scan pixel in the mask image (step B24). Next, it is determined whether the adjacent pixel P1 in the + Y direction of the scan pixel can be masked (step B25). If maskable, p1 has been determined only when the adjacent pixel in the −X direction of p1 cannot be masked. Register as a pixel (step B
26). Next, an adjacent pixel p2 in the −Y direction of the scan pixel
Is determined whether or not is maskable (step B27). If masking is possible, p2 is registered as a determined pixel only when an adjacent pixel in the −X direction of p2 cannot be masked (step B28).

Next, it is determined whether a pixel adjacent to the scan pixel in the + X direction can be masked (step B29). If masking is possible, the scan pixel is incremented in the + X direction (step B30), and the process returns to step B24. If masking is not possible, the processing from step B21 is repeated for the next determined pixel. This is continued until there are no more determined pixels.

The actual processing will be described with reference to FIG. The reference pixel 63 (pixel indicated by A in the figure)
It is registered as a first determined pixel position 64 in advance. When the search is performed in the -X direction from the first determined pixel position 64, the boundary pixel p is first found. Therefore, let p be a scan pixel. After setting the mask data at the position of the scan pixel to TRUE, it is checked whether the p1 pixel above it can be masked. However, since p1 is outside the target area, the process proceeds to the next step, and it is checked whether the p2 pixel below the scan pixel can be masked. Since p2 is within the target area, the next pixel in the -X direction of p2 is examined. Then
The adjacent pixel in the -X direction of p2 is FALSE outside the target area.
Therefore, the condition is met and registered as a determined pixel.

Since the adjacent pixel q in the + X direction of the scan pixel p can also be masked, the upper and lower pixels are similarly examined, and it is found that q1 is registered as a determined pixel. When the processing proceeds as described above, another boundary pixel r is eventually reached, and the processing for one line is completed. Thus, FIG.
In the mask determination processing method described above, one determined image is stored for each line, and it can be seen that the processing can be performed at a very high speed.

Next, expansion and contraction processing (steps A6 and A2)
8) will be described.

The expansion process is a process for increasing the area by one pixel, and the contraction process is a process for reducing the area by one pixel. This processing is also described in reference [4].

In the present invention, the mask image is TRUE
And FALSE pixels, the input mask image can be expressed as f (i, j) and the output mask image can be expressed as g (i, j) as follows.・ Expansion: f (i, j) or its vicinity 4 (around 8)
Is TRUE, g (i, j) = TRU
E. Otherwise, g (i, j) = FALSE Shrinkage: f (i, j) or its four neighbors (eight neighbors)
Is FALSE, g (i, j) = FALSE
SE. Other than that, g (i, j) = TRUE Here, 4 neighbors indicate four adjacent pixels on the upper, lower, left and right sides of the pixel, and 8 neighbors are 8 pixels in the upper, lower, left, and right directions of the pixel.
One adjacent pixel. In the present invention, expansion and contraction are performed N times, and fine irregularities and holes in the region are removed.

In the above description of the operation, it has been described that the operator specifies one pixel as the reference pixel. However, when the operator specifies a plurality of pixels, the above operation is repeated a plurality of times. The results can be integrated. The manner in which the masks are integrated will be described with reference to FIG. In the input image example 110, the operator tried to extract the target area 116, but initially could only extract the A area 111. Therefore, the A region 111 is stored, and then the B region 112 is extracted and integrated with the A region 111. In this way, if the E region 115 is integrated at any time, the entire target region 116 can be extracted.

An example of such a process will be described with reference to FIG. When the area extraction for one reference pixel is completed, the data in the mask image memory is not output but transferred to the second mask image memory (step C3). Then, the area extraction result is displayed on the monitor (step C4), and the next instruction from the operator is waited. When the next reference point is designated by the operator, the area extraction processing is performed again, and the result is integrated with the previous result held in the second mask image memory (step C2). If there is an end instruction from the operator, the mask image data is output and the processing ends.

[0041]

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing an image area extracting apparatus according to an embodiment of the present invention. FIG. 1 is a block diagram showing an embodiment of an image region extracting apparatus according to the present invention. FIG. 2 is a block diagram showing another embodiment of the image region extracting apparatus according to the present invention. FIG. 3 is a block diagram showing an embodiment of a pixel feature extracting means in the image region extracting apparatus of the present invention.

The first embodiment of the present invention comprises an image input device 1 such as a scanner or a camera, a monitor 2 for displaying an image,
An input image memory 4 for holding image data, a position designating means 3 for designating a target area, a reference point feature quantity storage buffer 6 for storing feature quantities of reference points, and a reference point for storing locations of reference points The pixel feature amount is compared with the position storage buffer 7, the neighboring pixel search means 8 for calculating the address of the neighboring pixel, the determined pixel position storage buffer 9, the mask image memory 5, and the data output device 10 such as a hard disk. Pixel feature quantity comparison means 11 and pixel feature quantity extraction means 12
And a CPU 15 for controlling each module.

The second embodiment of the present invention is different from the first embodiment in that it has a pixel characteristic amount extracting means 12 comprising a pixel accumulation extracting means 40.

The third embodiment of the present invention is different from the second embodiment in that the color conversion means 4
And a pixel characteristic amount extraction unit 12 having a configuration obtained by adding the pixel characteristic amount extraction unit 12.

The fourth embodiment of the present invention is a pixel feature extraction unit 12 having a configuration in which a low-frequency filtering unit 41 is added to the pixel feature extraction unit of the second and third embodiments.
Having.

In the fifth embodiment of the present invention, N (N is 2 or more) N pixels are used in place of the pixel feature extracting means 12 of the first embodiment.
And a plurality of pixel characteristic amount extracting means 21 composed of different pixel characteristic amount extracting means, and a mask creation result determining means 22.

According to the sixth embodiment of the present invention, the pixel value extracting means 40 is replaced with the pixel feature extracting means 12 of the first embodiment.
A plurality of pixel feature quantity extraction means 21 comprising a first pixel feature quantity extraction means 23 having a pixel value extraction means 23, a second pixel feature quantity extraction means 24 having a pixel value extraction means 40 and a low-frequency filtering means 41, And a result determination unit 22.

The seventh embodiment of the present invention is the first embodiment,
In addition to the components of the second embodiment, the third embodiment, the fourth embodiment, the fifth embodiment, and the sixth embodiment, a mask region expanding unit 13 and a mask region contracting unit 14 are provided. .

The eighth embodiment of the present invention is the first embodiment,
In addition to the components of the second embodiment, the third embodiment, the fourth embodiment, the fifth embodiment, the sixth embodiment, and the seventh embodiment, the second mask image memory 26 , And a mask area integrating means 27.

[0050] and FIG. 4, the first invention with reference to FIG. 8, second, third, the operation of the fourth embodiment.

Image data is input from the image input device 1 to the input image memory 4 and displayed on the monitor 2. The image input device 1 includes not only imaging devices such as a scanner and a camera, but also
It includes a device for acquiring images from a database connected via a network. The operator designates a point in the target area to be cut out from the image by using the position designation means 3 such as a mouse or a keyboard (step A).
1).

The position information (x) of the designated point (reference pixel)
The coordinates (y-coordinate) are first stored in the reference point position storage buffer 7 (step A2). Then, the reference point position information is also stored in the determined pixel position storage buffer 9 (step A3).

Next, the pixel feature amount of the reference pixel is extracted by the pixel feature amount extraction means 12, and the output value is stored in the reference point feature amount storage buffer 6 (step A4). The method of realizing the pixel feature amount extracting means 12 has already been described in the section of operation. Examples include those that output RGB values of the pixel, those that output hue, saturation, and lightness values, and those that apply a local filter to neighboring pixels.

An example of the pixel characteristic amount extracting means 12 will be described with reference to FIG. In FIG. 3, all inputs are pixel addresses (pixel positions) and outputs are pixel feature values. The pixel feature amount extracting unit 30 is provided with a pixel value extracting unit 40.
This is an example of the simplest configuration including only a pixel address. When a pixel address is input, a pixel value is returned.

The pixel feature quantity extracting means 31 comprises a pixel value extracting means 40 and a low frequency filtering means 41. The low-frequency filtering unit 41 causes a local filter such as the low-frequency filtering filter 72 of FIG. 11 to act near the input pixel.

The pixel feature quantity extracting means 32 includes a pixel value extracting means 40 and an edge strength extracting means 42. The edge strength extraction means 42 is provided with an edge strength extraction filter 74 shown in FIG.
Is applied to the vicinity of the input pixel.

The pixel characteristic amount extracting means 33 comprises a pixel value extracting means 40 and a high frequency filtering means 43. The high frequency filtering means 43 causes a local filter such as the high frequency filtering filter 73 of FIG. 11 to act near the input pixel.

The pixel characteristic amount extracting means 34 comprises a pixel value extracting means 40 and a color converting means 44. The color conversion means 44
RGB values are converted into HSV (hue (H), saturation (S), lightness (V)) values, XYZ values, L ^* a ^* b ^* values, L ^*
This is a module for converting to u ^* v ^* values. A method for converting to an HSV value is known from literature [1] and the like. Further, a method of converting the values into XYZ, L ^* a ^* b ^* , and L ^* u ^* v ^* values is also known from literature [3].

The pixel characteristic amount extracting means 35 includes a pixel value extracting means 40, a color converting means 44, and a low frequency filtering means 41, and outputs a value subjected to color conversion after the low frequency filtering operation as a characteristic amount. .

[0060]

[0061]

[0062]

[0063]

[0064]

First, the mask image memory 5 is initialized by FALSE (step B20). Next, one determined pixel position is read from the determined pixel position storage buffer 9 (step B21). At this time, the read position information item must be deleted from the determined pixel position storage buffer 9. A search is performed in the −X direction from the determination pixel position obtained by using the neighboring pixel search unit 8 and the pixel feature amount extraction unit 12,
A boundary pixel of the target area is found, and that pixel is set as a scan pixel (steps B22 and B23).

TRUE is set at the position of the scan pixel in the mask image memory 5 (step B24). The address of the adjacent pixel p1 in the + Y direction of the scan pixel is calculated by using the neighboring pixel searching means 8, and the pixel characteristic amount extracting means 12 is used.
Is used to extract the pixel feature amount. The feature value at that point and the contents of the reference point feature value storage buffer 6 are compared by the image feature value comparison means 11 to determine whether masking is possible (step B6). If it is determined that masking is possible, it is determined whether the adjacent pixel in the −X direction of the adjacent pixel is maskable. If the adjacent pixel in the −X direction is not maskable, p1 is input to the determined pixel position storage buffer 9 I do. The same processing is performed for the adjacent pixel p2 in the −Y direction of the scan pixel. These processes are continued until the determined pixel position storage buffer 9 becomes empty.

[0067] and FIG. 5, a fifth of the present invention with reference to FIG. 8, the operation of the sixth embodiment will be described.

This embodiment is the same as the first embodiment in the former part, and uses the first feature amount extracting means 23 as the pixel feature amount extracting means. Once the mask image data has been created, the contents of the mask image memory are checked using the mask creation result determination means 22. The mask creation result determination unit 22 determines whether or not the region extraction has been completed with a good result, or whether or not the region extraction has been performed using all the pixel feature amounts specified in advance. If the mask creation result determination unit 22 determines that the process has failed, the first pixel feature amount extraction unit 23 is replaced with the second pixel feature amount extraction unit 24, and the process is repeated from step A23 again. If the mask creation result determination means 22 determines success, the data held in the mask image memory 5 is saved in the data output device 10 such as a hard disk, and the process ends.

When the image data of the next mask is created, if the mask creation result judging means 22 judges that the image data has failed, the mask making result judging means 22 judges that the third pixel feature quantity extracting means is used. Repeat until it is done.

As an example of the mask creation result judging means 22, there is a method in which it is checked whether or not the number of TRUE pixels is n or more. If all of the N pixel feature amount extraction units implemented in the embodiment have been used, all the determinations are regarded as successful, and the process is forcibly terminated.

The operation of the seventh embodiment of the present invention will be described with reference to FIGS.

In this embodiment, in the first embodiment and the second embodiment, immediately before data is output, the data is held in the mask image memory 5 using the mask area expanding means 13 and the mask area reducing means 14. The expansion / contraction processing is performed on the area data. As an example of the processing, the mask data is first subjected to the expansion processing m times in the region of TRUE, and after removing the irregularities and small holes, the contraction processing is performed m times.
(M is an arbitrary integer of 1 or more).

The operation of the eighth embodiment of the present invention will be described with reference to FIG. In this embodiment, before outputting the mask image data in the above embodiment, the data is once held in the second mask image memory 26 (step C3). Further, the data of the mask image memory 5 is displayed on the monitor 2 as a region extraction result, and the next instruction from the operator is waited (steps C4 and C5). If the operator instructs to perform region extraction again, the process returns to the region extraction process C1 to continue the process.

Then, the new area extraction result in the mask image memory 5 is integrated with the data held in the second image memory 26 by the mask area integration processing 27 (step C2). As an example of the mask area integration process 27, for each pixel, if either is TRUE, the TRU
What is necessary is just to perform the process which sets it as E. If TRUE = 1,
If FALSE = 0, this processing is equivalent to taking the OR of both.

The integrated data is held in the mask image memory 5 and simultaneously transferred to the second mask image memory 26 again. Then, the result is displayed on the monitor 2 again,
Wait for operator's instruction. When the end instruction is given by the operator, the contents of the mask image data 5 are output to the data output device 10 and the operation ends (step C7).

[0076]

As described above, the image region extracting apparatus according to the present invention is capable of extracting only a connected region including a reference point specified by an operator (a region formed by a set of adjacent pixels having substantially the same pixel feature amount). There is an effect that it can be extracted. Thereby, even if the target area 102 in FIG. 12 is specified, the problem that the target area 103 is erroneously extracted does not occur.

Further, since the present apparatus processes only the connected area, there is an effect that the processing speed is improved as compared with the conventional apparatus.

Further, an apparatus using a low-frequency filtering filter for the pixel characteristic amount has an effect that an area constituted by a pattern such as the target area 104 in FIG. 12 can be extracted.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of an image region extracting apparatus according to the present invention.

FIG. 2 is a block diagram showing an embodiment of an image region extracting apparatus according to the present invention.

FIG. 3 is a block diagram showing an embodiment of a pixel characteristic amount extracting means in the image region extracting apparatus of the present invention.

FIG. 4 is a flowchart of an image region extraction method according to the present invention.

FIG. 5 is a second flowchart of an image region extraction method according to the present invention.

FIG. 6 is a third flowchart of an image region extraction method according to the present invention.

FIG. 7 illustrates a connected region extraction process in the image region extraction method.
Flow chart of related technology .

FIG. 8 is a flowchart of a connected region extraction process in the image region extraction method of the present invention.

FIG. 9 is an explanatory diagram of a technique related to a connected region extraction process.

FIG. 10 is an explanatory diagram of a connected region extraction process .

FIG. 11 is an explanatory diagram of a local filter.

FIG. 12 is an example of an input image.

FIG. 13 shows an example of an input image.

FIG. 14 is a block diagram showing an embodiment of a conventional image region extracting apparatus.

FIG. 15 is a flowchart of a conventional image region extraction method.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 image input device 2 monitor 3 position designation means 4 input image memory 5 mask image memory 6 reference point feature amount storage buffer 7 reference point position storage buffer 8 neighboring pixel search means 9 determined pixel storage buffer 10 data output device 11 pixel feature amount Comparison means 12 Pixel feature quantity extraction means 13 Mask area expansion means 14 Mask area reduction means 15 CPU 21 Multiple pixel feature quantity extraction means 22 Mask creation result determination means 23 First pixel feature quantity extraction means 24 Second pixel feature quantity Extraction means 25 N-th pixel feature quantity extraction means 26 Second mask image memory 27 Mask area integration means 30 Pixel feature quantity extraction means 31 Pixel feature quantity extraction means 32 Pixel feature quantity extraction means 33 Pixel feature quantity extraction means 34 Pixel features Amount extraction means 35 Pixel feature extraction means 40 Pixel value extraction means 41 Low Wave filtering means 42 Edge strength extracting means 43 High-frequency filtering means 44 Color conversion means 50 Enlarged image 51 Target area 52 Target area outline 53 Reference pixel 54 First determined pixel position 55 Second determined pixel position 60 Enlarged image 61 Target area 62 Outline of target area 63 Reference pixel 64 First determined pixel position 65 Second determined pixel position 66 Third determined pixel position 70 Local filter 71 Target pixel position 72 Low frequency filter 73 High frequency filter 74 Edge intensity extraction filter 100 Input image example 101 Target area 102 Target area 103 Target area 104 Target area 105 1st color pixel 106 2nd color pixel 110 Input image example 111 A area 112 B area 113 C area 114 D area 115 E area 116 Target area 201 Image input device 202 Monitor 203 thresholding means 204 position specifying means 205 the input image memory 206 reference pixel location buffer 207 mask image memory 208 output device

Claims (10)

(57) [Claims] 1. A step of designating a reference pixel in a target area to be extracted, one or more different pixel feature extraction steps of extracting feature of an arbitrary pixel, and a memory for storing a mask image (hereinafter referred to as a mask). Image memory
) With FALSE, and a buffer for storing the determined pixel position (hereinafter, determined pixel
Read the position of the determined pixel from the
(Step B21), and from the determined pixel in the −X direction (here, from left to right).
Direction is defined as + X direction)
Determining (step B22) and storing the boundary pixels as scan pixels (step B22).
B23) and TRUE at the position of the scan pixel on the mask image memory.
Setting step (step B24) and the + Y direction of the scan pixel (here, from the bottom to the top)
The direction is defined as + Y direction) and adjacent pixels (hereinafter referred to as pixel P1).
), The pixel feature amount of the pixel P1 and the reference pixel
The difference from the pixel feature is equal to or less than a predetermined threshold
And is adjacent to the pixel P1 in the -X direction.
If the contact pixel is not maskable, the pixel P1 has been determined
Step of storing in pixel position storage buffer (step B
25, step B26) and a pixel adjacent to the scan pixel in the −Y direction (hereinafter, pixel P2 and
Pixel), the pixel P2 can be masked,
In addition, a pixel adjacent to the pixel P2 in the −X direction cannot be masked.
The pixel P2 is stored in the determined pixel position storage buffer.
Step of storing (Step B27, Step B28)
And whether the adjacent pixel in the + X direction of the scan pixel can be masked.
(Step B29), and if it is determined in Step B29 that masking is possible,
Move the position of the can pixel by one pixel in the + X direction, and
Step from Step B24 to Step B29
(Step B30) and a judgment that the data has not yet been read out to the determined pixel position storage buffer.
Check if there is a fixed pixel and have not read it out yet
If there is a determined pixel, the process proceeds from step B21.
Steps to repeat step B30 (step B31)
A method for extracting an area of an image, comprising: 2. A step of designating a reference pixel in a target area to be extracted, one or more different pixel feature extraction steps of extracting a feature of an arbitrary pixel, and a memory storing a mask image (hereinafter referred to as a mask). Image memory
) With FALSE, and a buffer for storing the determined pixel position (hereinafter, determined pixel
Read the position of the determined pixel from the
(Step B21), and from the determined pixel in the −X direction (here, from left to right).
Direction is defined as + X direction)
Determining (step B22) and storing the boundary pixels as scan pixels (step B22).
B23) and TRUE at the position of the scan pixel on the mask image memory.
Setting step (step B24) and the + Y direction of the scan pixel (here, from the bottom to the top)
The direction is defined as + Y direction) and adjacent pixels (hereinafter referred to as pixel P1).
), The pixel feature amount of the pixel P1 and the reference pixel
The difference from the pixel feature is equal to or less than a predetermined threshold
And is adjacent to the pixel P1 in the -X direction.
If the contact pixel is not maskable, the pixel P1 has been determined
Step of storing in pixel position storage buffer (step B
25, step B26) and a pixel adjacent to the scan pixel in the −Y direction (hereinafter, pixel P2 and
Pixel), the pixel P2 can be masked,
In addition, a pixel adjacent to the pixel P2 in the −X direction cannot be masked.
The pixel P2 is stored in the determined pixel position storage buffer.
Step of storing (Step B27, Step B28)
And whether the adjacent pixel in the + X direction of the scan pixel can be masked.
(Step B29), and if it is determined in Step B29 that masking is possible,
Move the position of the can pixel by one pixel in the + X direction, and
Step from Step B24 to Step B29
(Step B30) and a judgment that the data has not yet been read out to the determined pixel position storage buffer.
Check if there is a fixed pixel and have not read it out yet
If there is a determined pixel, the process proceeds from step B21.
Step to repeat Step B30 (Step B31)
And a step of integrating target regions extracted by a plurality of reference pixels. 3. An input image memory for holding input image data, a monitor for displaying image data and an area extraction result,
There are position specifying means for specifying a target region to be extracted, a reference point feature amount storage buffer for holding a pixel feature amount of a reference point for area extraction, and a reference point position storage buffer for holding a position of the reference point for area extraction. A neighboring pixel searching means for searching for an address of a pixel adjacent to the pixel, a determined pixel storage buffer for storing a pixel to be masked, a pixel feature comparing means for comparing the feature of two pixels, and a feature of a certain pixel And a mask image memory for holding a region extraction result , wherein the neighboring pixel search means initializes the mask image memory with FALSE, and stores a determined pixel from the determined pixel storage buffer. Read the position of
From the determined pixel, in the −X direction (here, from left to right
Direction is defined as + X direction)
Then, the boundary pixel is stored as a scan pixel, and the TRU is stored at the position of the scan pixel on the mask image memory.
E is set and the scan pixel is set in the + Y direction (here, from the bottom to the top).
The direction is defined as + Y direction) and adjacent pixels (hereinafter referred to as pixel P1).
), The pixel feature amount of the pixel P1 and the reference pixel
The difference from the pixel feature is equal to or less than a predetermined threshold
And is adjacent to the pixel P1 in the -X direction.
If the contact pixel is not maskable, the pixel P1 is determined as described above.
The pixel is stored in a fixed pixel storage buffer, and a pixel adjacent to the scan pixel in the −Y direction (hereinafter, pixel P2
Pixel), the pixel P2 can be masked,
In addition, a pixel adjacent to the pixel P2 in the −X direction cannot be masked.
The pixel P2 is stored in the determined pixel position storage buffer.
And whether the adjacent pixel in the + X direction of the scan pixel can be masked.
And if it is determined that masking is possible, the position of the scan pixel
Is moved by one pixel in the + X direction to create a new scan pixel.
The same processing is repeated for the determination that the pixel position has not yet been read out to the determined pixel position storage buffer.
Check if there is a fixed pixel and have not read it out yet
If there is a determined pixel, the determined pixel is also
An image region extracting apparatus, which reads out and repeats the same processing . 4. The apparatus according to claim 3, wherein said pixel characteristic amount extracting means includes a pixel value extracting means for extracting a pixel value.
A region extraction device for the described image. 5. The image area extracting apparatus according to claim 3, wherein said pixel characteristic amount extracting means includes a pixel value extracting means for extracting a pixel value and a color converting means for performing a color conversion. . 6. The image region extracting apparatus according to claim 4, wherein said pixel characteristic amount extracting means further comprises a low frequency filtering means. 7. An image processing apparatus comprising: N (N is 2 or more) different pixel feature amount extraction means; and a mask creation result determination means for determining an area extracted using one type of pixel feature amount. The image region extracting apparatus according to claim 3, wherein: 8. The image processing apparatus according to claim 1, further comprising a first pixel characteristic amount extracting unit having a pixel value extracting unit for extracting a pixel value, and both said pixel value extracting unit and a low frequency filtering unit, in place of said pixel characteristic amount extracting unit. 4. The image region extracting apparatus according to claim 3, further comprising: a second pixel feature amount extracting unit; and a mask creation result determining unit that determines an area extracted using one type of pixel feature amount. . 9. The apparatus according to claim 3, further comprising mask area expanding means for increasing the area by one pixel, and mask area reducing means for reducing the area by one pixel. A region extraction device for the described image. 10. A system according to claim 1, further comprising a second mask image memory for holding a previous region extraction result, and a mask region integrating means for integrating the contents of the mask image memory and the second mask image memory. 9. The image region extracting apparatus according to claim 3, 4, 5, 6, 7, or 8.