JPH0454263B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (1) Technical field of the invention The present invention is a binary image that efficiently removes noise from a binary image in which an image to be extracted is expressed in state 1 and other images are expressed in state 0. This invention relates to an image noise removal method.

(2) Background of the technology Generally, in order to extract the outline of an object as an image, the object is photographed with a television camera (hereinafter referred to as a TV camera), and this photographed image is converted into a multilevel digital grayscale image using an AD converter. After conversion, this digital grayscale image is subjected to differentiation processing, smoothing processing, threshold processing, and thinning processing to obtain the outline of the object (generally, it often constitutes a closed loop) expressed in state 1. What is necessary is to obtain a value image. In this case, the above-mentioned binary image is designed to have only the image that constitutes the outline in state 1 by differential processing, threshold processing, etc., but the above-mentioned binary image is A situation arises in which the image in state 1 remains as noise in other parts of the image.

(3) Prior art and problems Therefore, as a conventional method for removing noise from such a binary image, there is a method using, for example, a logical filter. When an isolated point consisting of one pixel on a binary image is in state 1, the isolated point is removed as noise, so whisker-like noise or noise of several pixels cannot be removed. This causes a problem.

(4) Purpose of the Invention The present invention has been made from the above viewpoint, and its purpose is to provide a binary image that can reliably remove various noises that occur on a binary image. The purpose of this invention is to provide a noise removal method.

(5) Structure of the Invention The basic structure of the present invention focuses on the fact that the outline of an object generally forms a closed loop, and includes a polar point extracting means for extracting a polar point by comparing the gray levels of adjacent pixels. A method for removing noise from a binary image, which removes noise from a loop binary image obtained by extracting a rough contour of a multi-level grayscale image, comprising a binarization means, the image to be extracted is in a state 1,
A binary image in which the other images are expressed as state 0 is subdivided into pixel units arranged in a matrix, a window is constructed from an arbitrary pixel of interest and eight peripheral pixels surrounding this pixel of interest, and the pixel of interest is When is in state 1, the condition that the pixel of interest is not a pixel on the image to be extracted is determined in advance from the arrangement of surrounding pixels in the window, and when this condition is met, the pixel of interest is treated as noise. The present invention provides a method for removing noise from a binary image, in which noise is removed by detecting the noise and changing it to an image in state 0.

(6) Embodiments of the invention Hereinafter, the present invention will be described in detail based on embodiments shown in the accompanying drawings.

First, before explaining the noise removal method according to the present invention, a binary image extraction method will be briefly explained.
An object 2 placed on the top is photographed using, for example, a TV camera 3, and this photographed image is converted into a multi-level digital grayscale image using an AD converter 4, and then this digital grayscale image is converted into a digital grayscale image of the object using a binary image extraction circuit 5. The image is converted into a binary image expressed in state 1, which indicates the outline. In this case, the binary image extraction circuit 5 performs differential processing on the digital gradation image signal from the AD converter 4 to emphasize the gradation level of the digital gradation image, and then performs threshold processing to reduce the outline of the object to a certain width. After that, the outline of the object 2 is extracted in state 1 by thinning processing. An example of an analog representation of the binary image obtained in this manner is shown in FIG. In FIG. 3, the extracted binary image Z
is generally a closed-loop image _Z1 in state 1 showing the outline of object 2 and an image that does not form a closed loop in state 0.
It consists of Z ₀ .

Next, the basic principle of the noise removal method according to the present invention will be explained.

First, as shown in FIG. 3, the binary image Z is subdivided into g pixels arranged in a matrix, that is, in the horizontal and vertical directions, and a window W is created that moves on the binary image Z every pixel g. Set. For example, as shown in FIG. 4, this window W is composed of a pixel of interest E and eight surrounding pixels A, B, C, D, F, G, H, and I.

Next, using the window W, the pixel of interest E
Examine the pattern when is a pixel on the closed-loop image _Z1 . In this case, in order to simplify the explanation, if the closed-loop image Z ₁ has a corner, it is assumed that the pixel corresponding to the corner is treated as not a pixel on the closed-loop image Z ₁ . . First, since the closed-loop image Z ₁ is continuous, the closed-loop image Z ₁ always crosses the window W, and at least two of the surrounding pixels within the window W are pixels on the closed-loop image Z ₁ . It can be said that it must be. Considering the above preconditions, it can be said that the two peripheral pixels on the closed loop image _Z1 must be in a non-adjacent positional relationship. From these facts, in order for the pixel of interest E to be a pixel on the closed-loop image _Z1 , when the pixel of interest E is in state 1, at least two surrounding pixels that are not adjacent to each other must be in state 1. Become. 16 items that meet this condition
There are several cases, and each case is represented by a window W as shown in FIG.

Thereafter, a pattern in which the basic pattern does not occur within the window W shown in FIG. 5 is examined and set as a noise removal condition for removing noise of several pixels or whisker-like noise. In this case, the first noise removal condition is as shown in FIGS. 4 and 6 a.
When the pixel of interest E is in state 1, at least the surrounding pixels along two adjacent sides of window W are in state 0, that is, the upper l ₁ of window W
And the surrounding pixels A, B, C, D, G along the left side _l2 are in state 0, the surrounding pixels A, D, G, H, I along the left side _l2 and bottom side _l3 of window W are in state 0, window W Peripheral pixels G, H, along the lower side _l3 and right side _l4 ,
For example, I, F, and C are in the state 0, and the surrounding pixels I, F, C, B, and A along the right side _l4 and the upper side _l1 of the window W are in the state 0. As shown in FIGS. 4 and 6b, the second removal condition is that when the pixel of interest is in state 1, at least the peripheral pixels along one side of the window W and the peripheral pixels adjacent thereto are The state is 0, and the peripheral pixels located at the center of the opposite side of the window W are in the state 1. For example, peripheral pixels A, B, C along the upper side _l1 of the window Wa and peripheral pixels D, F adjacent thereto. is in state 0, and the peripheral pixel H located at the center of the lower side _l3 of window W is in state 1. Note that the pixels marked with an x in FIGS. 6a and 6b indicate that they may be in either state 1 or state 0.

Therefore, the window W moving on the binary image Z
If the pattern within matches either of the first and noise removal conditions above, even if the pixel of interest E is in state 1, the pixel of interest E is in the closed loop image Z ₁
It will be detected as noise, not as an upper pixel, and this noise will be effectively removed.
Specifically, FIG. 10 shows a case where the leftmost window pattern in FIG. 6a under the first noise removal condition is removed as noise. Since a is an isolated point,
It can be considered as noise. Since b to d are 0 and end points in 7 consecutive directions in 8 connections (an expression indicating connections with surrounding 8 pixels), e and f are also 0 and end points in 6 consecutive directions in 8 connections. , all of which can be considered noise.

g is a case where three pixels form a corner, and the pixel located at the apex of the corner (in the figure, the pixel located at the center) is connected to the remaining two pixels under the condition of 8-connection, so Since the closed loop is not interrupted even if the pixel is removed, it is considered as noise. Furthermore, h is a case in which four pixels form a cluster. This kind of pixel cluster is
This generally occurs when the binarization threshold is set relatively low, and under these pixel conditions, if noise removal is finished while leaving the area, processing such as contour extraction and shape recognition will be difficult. This creates ambiguity and makes it impossible to make correct recognition. Further, even if this cluster forms part of a closed loop, the loop will not be interrupted even if the pixel of interest (the pixel located in the center in the figure) is removed, so the pixel is considered to be noise. Note that the other window patterns in FIG. 6a are simply rotated, so they can be considered in the same way.

In addition, the pattern removed by the leftmost window pattern in Figure 6b under the second noise removal condition is
Shown in Figure 1. Since a is 0 and an end point in 7 consecutive directions among 8 connections, and b and c are 0 and end points in 6 consecutive directions among 8 connections, both can be considered as noise. Furthermore, in d, the three pixels on the bottom side are connected, so even if the pixel of interest is deleted, the closed loop will not be interrupted, so it can be considered as noise. Note that the other window patterns in FIG. 6b are simply rotated, so they can be considered in the same way.

As described above, when there are continuous pixels around the pixel of interest, pixels other than the outline are regarded as noise without leaving a uniform area under the conditions shown in Figure 5 with 8-connection. It can be removed by
For example, as shown in FIG. 3, if there is a noise N in which four or fewer pixels are clustered together, it is removed under the first noise removal condition, and a whisker-like noise N′ extending from the closed-loop image _Z1 is removed. If it exists, it will be removed by the first and second noise removal conditions.

An example of a circuit for implementing the above-described noise removal method is shown below.

For example, as shown in FIG. 2, this noise removal circuit 6 stores and stores data signals necessary for noise removal processing among the digital image data signals from the binary image extraction circuit 5, and also performs horizontal scanning of the TV camera 3. It consists of a data shift circuit 6a that shifts the data signal in synchronization with the signal, and a data processing circuit 6b that performs calculations for noise removal processing based on the data signal stored in the data shift circuit 6a.

The data shift circuit 6a is shown in FIGS.
As shown in the figure, assuming that the number of pixels arranged horizontally on the captured image screen of the TV camera 3, in other words, the number of pixels arranged horizontally on the binary image, is 2N+
It stores binary data corresponding to three pixels, and consists of 1-bit registers Re ₁ to Re ₉ and N
- Consists of shift registers S ₁ and S ₂ that store 3-bit data, and registers Re ₁ to
The data stored in Re ₉ corresponds to pixels I to A shown in window W in FIG. That is, as shown in FIG. 8, when the captured image screen of the TV camera 3 is sequentially scanned from top to bottom with horizontal scanning lines h ₁ , h ₂ , h ₃ . . . , the scanning lines
At the stage where the scanning by h ₁ and h ₂ is completed and the scanning of the first three pixels by scanning line h ₃ is completed, the data shift circuit 6a receives the data from the binary image extraction circuit 5.
2N+3 pixel data will be stored, and the registers Re ₁ to Re ₉ will contain pixels g ₃₃ (pixel located in the third row and third column on the captured image screen, the same applies hereinafter), g ₃₂ , _g31 , _g23 ... _g11 are stored, and each data is used as data for pixels I to A of window W. When the scanning of pixel g ₃₄ by scanning line h ₃ is completed, the data corresponding to pixel g ₃₄ is newly stored in register Re ₁ , and the data in each register is shifted as a whole. , in this staircase, registers Re ₁ to Re ₉ each correspond to a pixel
The data corresponding to g ₃₄ , g ₃₃ , g ₃₂ , g ₂₄ , . . . g ₁₂ will be stored, and the window W will be moved one row to the right. Furthermore, as scanning by scanning line _h3 progresses, the window W shifts to the right in synchronization with the scanning signal, and when scanning by scanning line _h4 progresses, the window W moves down one line. After that, it moves sequentially from left to right. In this way, the window W sequentially moves as the scanning line m scans, and at the staircase where the scanning line m has finished, the window W moves over the entire range of the photographed image pixel by pixel. As a result, all the data necessary for noise removal processing on the binary image are read, and each data is sequentially sent to the data processing circuit 6b.

Further, the data processing circuit 6b outputs the level of the pixel E of interest as state 0 when the first noise removal condition or the second noise removal condition is satisfied, and otherwise outputs the level of the pixel E of interest as state 1. The data processing circuit 6b sequentially obtains binary image data from which noise has been removed. Then, the data processing circuit 6b
Specifically, is configured to satisfy the following logical formulas (a) to (d).

E^ ₁ = {(A∪B∪C∪D∪G)∩(A∪D∪G∪H∪I)∩(G∪H∪I
∪F∪C) ∩(A∪B∪C∪F∪I)}∩E={(A∪B∪C∪A∪D∪G)∩(
A∪D∪G∪G∪H∪I) ∩(G∪H∪I∪C∪F∪I)∩(A∪B∪C∪C∪F∪I)｝∩E = {(X ₁ ∪X ₂ )∩(X ₂ ∪X ₃ )∩(X ₃ ∪X ₄ )∩(X ₄ ∪X ₁ )}∩E
…(a) E^ ₂ = [{(A∪B∪C∪D∪F)∪｝∩｛(A∪D∪G∪B∪H)
∪｝∩{(D∪F∪G∪H∪I) ∪｝∩{(C∪F∪I∪B∪H)∪｝]∩E＝[{(X ₁ ∪
D∪F)∪｝∩(X ₂ ∪B∪H) ∪}{(X ₃ ∪D∪F)∪｝∩{(X ₄ ∪B∪H)∪｝〕∩
E…(b) E^=E^ ₁ ∩E^ ₂ …(c) X ₁ = A∪B∪C, X ₂ = A∪D∪G, X ₃ = G∪
H∪I _,
Indicates the inverted data of F and H, E^ ₁ and E^ ₂ are index values indicating whether the first noise removal condition and the second noise removal condition are satisfied, respectively, and E^ is the level of attention E that is noise. This is an index value indicating whether or not.

FIG. 9 shows a specific configuration of the data processing circuit 6b that satisfies this logical formula. In Figure 9, OR _1
OR14 to OR14 are OR gates, _AN1 to _AN3 are AND gates, and _NT1 to _NT4 are NOT gates.

Note that the specific means for implementing the noise removal method according to the present invention are not limited to those shown in the above embodiments, and of course the settings can be changed as appropriate.

(7) Effects of the Invention As explained above, according to the method for removing noise from a binary image according to the present invention, a binary image is created in which the image to be extracted is expressed as state 1 and the other images are expressed as state 0. is subdivided into pixel units arranged in a matrix, a window is constructed from a given pixel of interest and eight surrounding pixels surrounding this pixel of interest, and when the pixel of interest is in state 1, the pixel of interest is the extraction target. The condition that the pixel is not on the image is properly determined from the arrangement of surrounding pixels in the window, and when this condition is met, the pixel of interest is detected as noise and it is changed to an image with state 0. Therefore, it is possible to reliably remove various noises such as noise of several pixels that occurs on a binary image and whisker-like noise that extends from the figure to be extracted.
Binary images can be made accurate. In particular, the present invention is extremely effective in removing noise when images to be extracted form a closed loop.

[Brief explanation of the drawing]

Fig. 1 is an explanatory diagram showing the state of an object photographed by a TV camera, Fig. 2 is an explanatory diagram showing the basic principle for extracting a binary image, and Fig. 3 is a closed loop binary image obtained from a binary image extraction circuit. An explanatory diagram showing the state of a value image, Fig. 4 is an explanatory diagram showing a window used to remove noise, and Fig. 5 is a basic pattern within the window obtained when the pixel of interest is a pixel on a closed-loop image. Explanatory diagram showing, Figure 6 a, b
is an explanatory diagram showing the basic pattern within the window obtained when the pixel of interest is not a pixel on a closed-loop image, FIG. 7 is a circuit diagram showing a specific example of a data shift circuit, and FIG. 8 is an illustration of the operating principle of the data shift circuit. FIG. 9 is a circuit diagram showing a specific example of a data processing circuit, and FIGS. 10 and 11 are diagrams showing examples of arrays of pixels containing noise to be removed in the present invention. Z... Binary image, Z ₁ ... Closed loop image, E...
Pixel of interest, A, B, C, D, F, H, I... Surrounding pixels, W... Window, l ₁ , l ₂ , l ₃ , l ₄ ... Edge of window, 6... Noise removal circuit, 6a...data shift circuit, 6b...data processing circuit.

Claims (1)

[Claims] 1 A noise removal method for binary images that simultaneously removes isolated points, whisker-like noises, and noises consolidated into several pixels other than the rough contours of a binary image from which the rough contour of an object has been extracted. 1. Subdivide the binary image in which the other images are expressed in state 0 into pixel units arranged in a matrix, configure a window with an arbitrary pixel of interest and eight peripheral pixels surrounding this pixel of interest, and perform the above steps. When the pixel of interest is in state 1, at least peripheral pixels along two adjacent sides of the window are in state 0, and at least peripheral pixels along one side of the window and adjacent peripheral pixels are in state 0. 0 and surrounding pixels located at the center of the opposite side of the window are in state 1, detecting the pixel of interest as noise and changing it to an image in state 0. A method for removing noise from a binary image, characterized by removing noise.