JPS59216283A - Picture processing device - Google Patents

Abstract

PURPOSE:To improve a signal-to-noise ratio by extracting the outline of a light and shade picture as a binary picture in real time in synchronization with horizontal scanning signals of a TV camera. CONSTITUTION:A pole extracting section 14 that extracts the pole of oblique direction, a threshold value processing section 15 that removes background noises and a noise removing section 16 that removes whiskerlike noises and isolated noises other than several picture elements are provided. By selecting load coefficients of the first space filter arithmetic section 12 and second space filter arithmetic section 13 properly, random noises on the picture can be reduced and sudden change of brightness can be emphasized. The load coefficient is taken out from a load coefficient table 22 by a CPU21. By extracting the outline of light and shade picture as a binary picture in real time in synchronization with horizontal scanning signals of a TV camera, generation of interrupted part is prevented, background noises are eliminated and whiskerlike and isolated noises are removed.

Description

[Detailed Description of the Invention] [Technical details of the invention] The present invention pertains to one image processing device that can perform smoothing, gradient calculation, 2-g, 2-cyan calculation, extreme point extraction, threshold processing, and The present invention relates to an image processing apparatus that executes a series of processes for extracting the outline of an object as a binary image by performing noise removal processing and the like at high speed in real time in synchronization with the horizontal scanning signal of a TV camera.

[Technology background]

Recently, industrial robots have been used in various fields. For example, in the case of mechanical assembly robots, robots can move objects for a second process in a short period of time until the end of one operation, such as tightening a screw. If you don't recognize what it is, you won't be able to continue the action. For this purpose, it is necessary for the robot's eyes to process the multi-level image signals input from the TV right camera at high speed to quickly recognize what the object is. To do this, we extract parts where the brightness changes suddenly at high speed.

It is necessary to capture the outline of an object.

[Prior art and problems]

For image data obtained using a general KTV camera, one screen is, for example, 256 x 256 (or 128
12B) Since it is a two-dimensional 8-bit gray level of pixels, the amount of information is extremely large. Conventionally, when trying to handle such image data, the image signals from the TV camera are stored in an image memory for one frame, and then the data is transferred to a large computer for processing. Alternatively, techniques have been used in which the slice level of the video signal from the TV right camera is determined by some method, the data is binarized, the data is compressed, the data is stored in a memory, and the data is subjected to logical operation processing.

Therefore, while the former can perform advanced image processing, it has the disadvantage of requiring large memory and high-speed large-scale computers to achieve high-speed processing, resulting in high costs. It is difficult to determine the appropriate slice level for digitization, and since multilevel information is omitted to binary level information, the amount of information is missing, so even if this is processed, it is difficult to Inability to process 7 For example, it has the disadvantage of not being able to process images with complex changes in shading.

Recently, some general-purpose image processing devices are designed to speed up image processing by arranging one basic calculation module for each pixel or multiple pixels and performing parallel calculations. Since a large number of basic arithmetic modules must be arranged in order to process , the processing device itself becomes large and costly, and address control becomes complicated.

Therefore, in order to improve such drawbacks, the present invention provides a device for extracting a contour line drawing of an object from a grayscale image.

As shown in FIG. 1, an object whose two contours are to be extracted is photographed using, for example, a TV camera 51, and this photographed image is converted into a multilevel digital grayscale image using an AD converter 2, and then this digital grayscale image is converted into a binary grayscale image. The applicant has filed application No. 57-104742 for converting the image extraction circuit 3 into a binary image representing the outline of an object.

In this case, the binary image extraction circuit 3 performs 1 multiplication, addition,
A spatial filter calculation section consisting of a division circuit is used, and noise reduction and image enhancement are performed by appropriately selecting the weighting coefficients of this spatial filter calculation section.

The outline of the object is extracted as a binary image by extracting the horizontal and vertical poles from the data.

This method is very effective when the target is a single object, the background has a uniform gray level, and there is a high contrast between the target and the background.

However, with normal image data capture systems, the surface condition of the target object (9 colors, shape, lighting conditions, lens brightness, aperture, etc.) can cause uneven shading levels and make it difficult to obtain high contrast between the target and the background. There may be none. In such a case, a wedge-shaped pattern appears in the background and the S/N ratio decreases. In order to eliminate this background noise, threshold processing is first required. In addition, if the objects overlap and five or more regions intersect, it is not possible to simply extract the horizontal and vertical poles from the data after the smoothing processor Laplacian processing. An interrupted portion occurs as shown by No (the reason will be explained later).

Furthermore, as shown in Fig. 6, whisker-like noise Nl branched from two contour lines and isolated noise Nl of several pixels or less due to external factors.
! etc., cannot be removed and remains.

[Purpose of the invention]

An object of the present invention is to solve the above-mentioned problems by not only extracting extreme points in the horizontal and vertical directions, but also extracting polar points at 45°
Then, by performing polar point extraction processing in the diagonal direction of 165°, sudden changes in brightness on the original image are extracted, the occurrence of discontinuous parts is eliminated, background noise is eliminated by threshold processing, and noise is removed from the binary image. An object of the present invention is to provide an image processing device that can remove whisker-like noise and isolated noise through processing, extract only the outline of an object as a series of connections, and obtain an image with a good S/N ratio.

[Structure of the invention]

In order to achieve this purpose, the image processing device of the present invention has two steps:
In order to process a grayscale image pixel by pixel and extract its outline, a data shift circuit holds and stores a certain length of data, and a multiplier that multiplies the held data by a preset weight coefficient. a spatial filter operation unit that includes a circuit, an addition circuit that adds the results, a factoring circuit that keeps the level increased by multiplication φ addition within an appropriate range, and extracts horizontal and vertical poles. In an image processing device having a pole point extraction unit, a pole point extraction unit extracts pole points in an oblique direction, a threshold processing unit removes background noise, and a noise removal unit removes sink-like noise and isolated noise of several pixels or less. The contour of the grayscale image is calculated in binary 1 in real time in synchronization with the horizontal scanning signal of the TV camera.
It is characterized by being able to extract the iiiI image as well.

[Key points of the invention]

The gist of the present invention is to extract pixels with locally high gray levels from gray scale image data after applying a spatial filter or FFT as candidate points for the outline of a target object.
Pole processing is performed in the vertical direction, 45° diagonal direction, and 135° diagonal direction to extract an uninterrupted contour line even when three or more areas intersect, and threshold processing is performed to reduce the It is assumed that the level background noise has been removed, and that even if the number of contour lines extracted as a binary image is small, it constitutes a closed loop with, for example, 8 connections.9
Only the outline of the object is extracted from the original image by processing such as removing whisker-like noise and isolated noise through logical operations within the local window.

[Embodiments of the invention]

An embodiment of the present invention will be explained with reference to the accompanying drawings starting from FIG.

FIG. 4 shows a simple configuration diagram of the image processing apparatus of the present invention.

Its basic configuration is: first spatial filter calculation section 12, second spatial filter calculation section 16. Pole point extraction section 14゜Threshold processing section 1
5. A noise removal unit 16, which is a first spatial filter calculation unit 1
2 and the second spatial filter calculation section 16 are the changeover switch section 17.
are connected in series or in parallel, and in the case of parallel connection, an absolute value calculation unit 18 is inserted.

Now, when object 4 in Figure 5 (a-1) is viewed from above,
The image shown in FIG. 5(a-2) is obtained, and by passing the image along the line X-Y, the output signal shown in FIG. 5(a-3) is obtained. If this is differentiated, the output shown in FIG. 5 (b-3) can be obtained. If the differential output shifts to slice levels A, B, and C, the 6th
Although the output of Fig. 1a'l, (bl, (cl) is obtained, it is difficult to select the slice level B to obtain the ideal contour of Fig. 6 1bl.
Figure 5 (b-3) obtained by differential processing of the signal of -3)
If more extreme point extraction is performed, an output like that shown in FIG. 6 1bl can be obtained without particularly setting the slice level. Note that FIGS. 5 and 6 will be described in detail later.

The first spatial filter calculating section 12 and the second spatial filter calculating section 16 have the same configuration, and the details thereof will be representatively explained using the first spatial filter calculating section 12.

The spatial filter calculation unit performs the calculation given by the ninth-order equation (1) in a hardware manner.

ym, n= Σ Σ Xm+l+n+j・ft, 1/
H-・--1it1--1 j--1 Here, Xff1+l'l l Ym,n is the value before and after filtering of the gray level value of the pixel in m rows and n columns on the screen, respectively, as shown in FIG. In the figure, r'1.1 is a weighting coefficient, and H is a normalization coefficient. This filling changes the value of Ym,n to the pixel of interest Xl,1. fl and its surrounding 8 pixels Xm+I
toil (i=1,0.

-1, j=1, 0, -1 (excluding 1-j-0), the weighting (ft, +) is determined by calculation.

As shown in FIG. 13, the characteristics of the filter can be varied in various ways by selecting the value of <+'I**.

Therefore, the load coefficient 'td is determined as shown in Figure 7 (b).

When the normalization coefficient H is defined as H-16, HYm+11 is obtained by the following equation (2).

Ym, -16(XlTl-1, n-I X 1 +Xm
, n-t X 2 + Xm+1. n-+X 1+X
In-1, nx 2+Xm, nX 4 +Xrn+1.
nX 2 10Xm-1, n+tX 1+Xm, n+I×
2+Xm+1. n+1×1〕...beat...song...song...
(2) In the state shown in FIG. 4, the first spatial filter calculation section 12 and the second spatial filter calculation section 13 are connected in series by the changeover switch section 17. In this case.

By using the first spatial filter calculation unit 12 in the first stage for a smoothing filter and the second spatial filter calculation unit 13 in the second stage for an image enhancement filter (bypass, Lagrangian), image noise can be reduced in real time. This is to extract the outline of the object.

Now, the hardware configuration for realizing the calculation of the above equation (1) will be explained. As shown in FIG. 8, this spatial filter section includes a data shift circuit 30. Multiplication circuit 31° Addition circuit $32
．． The divider circuit 36 is composed of four parts.

The data shift circuit 3o executes 6×3 spatial filtering processing on image data for one screen, as shown in FIG. 9, as shown in FIG. The j-element data is memorized and stored using a shift register and sent to the 9th stage multiplication circuit as Xm+
l+n+l(+--' +0+1 + J=' +
It has the function of transmitting 9 pixels of 0.1).

In FIG. 9, the case of screen data with 1 row of p pixels and 2q# light level will be described. The A/D converter 11 shown in FIG. 8 converts data displayed by 1 pixel and q bits into
The data is held using a shift register for 3 pixels in the first stage, and 9 pixels in the second and third stages. For the last six pixels of each stage, latches or free lag flops R1~
By cascading R9 or using a serial-in/rallel-out shift register, nine pixels can be transmitted. And the second and third rows are (p-3)
Shift registers 54 and 55 for pixels are used to hold data for pixels other than those for transmission.

This data shift circuit shifts one pixel at a time in response to a scanning signal from the TV camera 210.

As shown in FIG. 10, the multiplication circuit 31 has a weighting coefficient f+,1 (1--1) preset for each of the nine pixels sent from the data shift circuit 30.

0.1, j=-1, 0, 1) to obtain the following equation (3).

Wm+i, n+j=Xm+i, n+j×fi,j ・
・・・・・・・・・・・・・・・・・・・・・(3)(＋
=-1+0*1+j=-1,0,1) For this purpose, as shown in FIG. is the load factor f-
1, -1 are set, and multiplication with pixels x, , -, , n-1 sent from the data shift circuit is performed.

A weighting coefficient f-1,G is set in the second multiplier 37, and the pixel X sent out from the data shift circuit
Multiplication by m-1,n is performed. In this way the first
The multiplication circuit shown in FIG. 0 can perform the multiplication in equation (3) above.

The adder circuit 32 calculates the nine values Wm+I, n+j(1""' 1 +0.1+) calculated by the multiplier circuit 31.
J-1*o*1), it has an addition η function that calculates the following equation (4), which is the sum of these.

(1""' 1+0+1+J-1to, 1) For this purpose, as shown in FIG. 11, first adder 45 to eighth adder 5 are used.

Furthermore, the division circuit 33 adjusts the gray levels Zm and n increased by the multiplication and addition operations in the multiplication circuit 61 and the addition circuit 32 so that they fall within the gray level range of the image memory 19 and the monitor TV shown in FIG. , the normalization coefficient H
Divide by.

This is what we seek. Here, the normalization coefficient H is the actual circuit 3
3 is composed of a first data selector 53 to an eighth data selector 6o, as shown in FIG.
), Y is the final calculation result of the above equation (1). , n can be obtained.

FIG. 12 shows 16-bit gray level data 2.2 by multiplying and adding 8-bit gray pixels as described above. ,,. In the example of reducing the data to 8-bit data Ym, n, input 2 to 9 pits of 16-bit data Z□, 0 to the first data selector 53, and There are 16 pit data Zm,
3 to 10 bits of n are input, and similarly 4 to 11 bits are input to the third data selector 55, 5 to 12 bits are input to the fourth data selector 56, and 6 bits are input to the fifth data selector 57. ~16 bits to the sixth data selector 5
7 to 14 bits to B, 8 to 15 bits to the seventh data selector 59, and 9 to 1 bits to the eighth data selector 6°.
Input 6 bits each. and selector line 8
In response to the 3-bit selection signal transmitted from L, each data selector 56-6o selects and outputs any one of the first to eighth bits among the 8-bit inputs.

For example, when quoting zm and n by 27, each data selector 56 to 6o is controlled so that the first bit is output from each data selector 53 to 60.
The first data selector 56 outputs the first bit of 2 to 9 bits, that is, the 2nd bit of 16 bits, and the second data selector 54 outputs the first bit of 3 to 10 bits. The 6th bit out of 16 bits is output, i\! 6 data selector 5
5 outputs the first bit of bits 4 to 11, that is, the 4th bit of 16 bits, and similarly, the eighth data selector 6o outputs the first bit of bits 9 to 16. That is, the 9th bit out of 16 bits will be output, and as a result, the pot data selectors 53 to 60 will output 2 to 9 pits, which is the value obtained by dividing the 16 bits Zm and n by 2. Become. same&
When dividing K and zm and n by 26, it is better to output the second bit from each data selector 56 to 60, that is, 3 to 10 bits from the 16 bit data.
In the case of quotient by 28, it is sufficient to output the third bit from each data selector 53-60 and output the 4th to 11th bits from the 16-bit data. In this way, the 21''-N to 28-N bits of the 16-bit data are input to the N-th data selector, and the number divided by 2 is 2k.
By setting the value of k in the selector line corresponding to , division can be realized at high speed. in this way,
By using a data selector, the speed of division can be increased.

By appropriately selecting the weighting coefficients of the first spatial filter calculation unit 12 and the second spatial filter calculation unit 16 in FIG. 4, it is possible to reduce random noise on the image and emphasize sudden changes in brightness. This load factor can be taken out by the CPU 21 from the load factor table of the load factor holding section 22, for example.

For example, the state after smoothing Laplacian processing is applied to the original image of Fig. 5 (al) is shown in [bl]. Fig. 5 (a
-1) is the case where the light hits the cylindrical object 4 from diagonally above the right side, (a-2) is the view when it is viewed from directly above, and (a-3) is the case where the light is hitting the cylindrical object 4 from diagonally above the right side. It shows the change in brightness in the x-yi dimension direction of a-2). Also, Fig. 5 (b-
1) is a three-dimensional representation of the gray scale level after smoothing Lagrangian operation, and (b-2) shows isobright lines when viewed from directly above. Furthermore, FIG. 5(b-3) shows a cross section in the X-Y one-dimensional direction of FIG. 5(b-2).

Let us now consider extracting the contour of the object from FIG. 5(b-1) by threshold processing. When five levels of FF are selected as slice levels, A, B, and C, and binarized, the 6th
As shown in Figures (al to (cl), different contour lines are obtained. In Figure 6 (al), the level is too high and the contours are interrupted, and in the same fcl, the level is too low and the contours become thick. Therefore, it is necessary to repair the discontinuity or thin the obtained contour line again.These processes require complicated algorithms and take time, so they are not suitable for real-time processing. (It would be convenient if you could set just the right slice level like bl, but the shading level of the original image varies depending on the surface condition, color, shape of the target object, lighting conditions, lens brightness, aperture, etc.) It is difficult to find the slice level.In the end, the target is extracted as a series of line drawings only by threshold processing and the S/N is
It is difficult to obtain a good image of.

Therefore, in the present invention, contour candidate points are first extracted by extreme point extraction processing. The principle of this polar point extraction operation will be explained with reference to FIG. To simplify the explanation, an example will be described in which the extreme point R is determined when the original image P shown in FIG. 14(a) has the one-dimensional density distribution as shown. Original image P
The left-shifted image obtained by shifting one pixel △ to the left is set as PL, and -Ii! to the right! The right-shifted image shifted by Δ is set as PR. Then, the original image P and the left-shifted image PL are compared, and a portion QL in which the density of the original image is higher or equal (hereinafter referred to as P≧PL) is extracted, and the i-divided original image P and right-shifted image PR are extracted. A portion QB in which the density of the original image is greater than or equal to that of the original image (hereinafter referred to as P≧PR) is extracted. Since the extreme point (maximum point) °R found at this time is considered to be at least at a level greater than or equal to the pixels on both sides, the portion Qt is determined as shown in FIG. 14(E).

The extreme point R can be found by extracting the point R and QR. However, P=PL! Exclude when doing PR.

By the way, when a pole exists in a pattern like the one shown in FIG.
! , , g cannot be extracted, so by scanning this also in the Y direction, the pole sides g, , g4 can be extracted. Therefore, 2 For example, Fig. 14 (
An extraction window W′ having pixels A to ■ as shown in g)
If you use l), E, and F, you can obtain
The poles in the direction are the poles in the Y direction by B, E, and H, the poles in the 45° diagonal direction by C, E, and G, A, E,
According to I, the diagonal poles of 135° can be extracted.

Now, one frame shown in FIG. 14 (H) is sequentially scanned by the scanning line h (1111! * h@...) according to the synchronization signal output from the scanning signal generator 20 in FIG. By comparing and processing as above, horizontal, vertical,
Pole points in the 45° and 135° directions can be sequentially extracted.

The pole extractor 14 shown in FIG. 4 is composed of a data shift circuit 62 and a data comparison circuit 63, as shown in FIG.

The data shift circuit 62, as detailed in FIG.
The data of 2 rows + 3 iijI elements are held sequentially using a shift register, and the pixel A is sent to the data comparison circuit 63 of the 9th stage.
It has a function of sending out nine pixels of .about.I, and is configured similarly to the data shift circuit shown in FIG. The output of the spatial filter 61 consisting of the first spatial filter calculation unit 12 and the second spatial filter calculation unit 13 is stored in registers rl to r9,
The data is input to a data shift circuit 62 composed of shift registers 64 and 65 with a length of N-3 bits, and pixels A to ■ are outputted from each register r and M+re.
It will be input to the data comparison circuit 65.

As shown in FIG. 17, this data comparison circuit 66 includes comparators C-1 to C-8, OR circuits 0R-1 to 0R-9, AND circuits AN-1 to AN-4. NAND circuit NAN-1~N
It is composed of an AN-4, a beak φ register PR, and the like. The comparator C-1 compares pixels F and E in the register r4er5, which are spatially filtered data, and the comparator C-2 similarly compares pixels F and E. Therefore, if pixel B is the pole in the horizontal direction (X direction), E
>D, E)F.

Since D~E~F, 0R-2, 0R-3 and NAN
-1 outputs “1” in both cases, and 0R-i outputs “1”.
'' is output, the pixel E as the horizontal peak value is held in the peak register PR, and this is output as the extreme point.

Also, pixels B and E are compared in the comparator C-5.

Since pixels E and H are compared in comparator C-4.

If pixel E is a vertical pole, it is set to “1” from AN-2.
is output, and since "1" is output from 0R-1, pixel E is held in the peak register PR. Similarly, pixels C, E, and G in a 45° diagonal direction are compared by comparators C-s and C-6, and if pixel E is the pole, AN-5
"1" is output from the comparator C-7 and C-8.
35° diagonal pixel A, Fi.

(2) is compared, and if pixel E is the pole in this 135° diagonal direction, "1" is output from AN-4.

Therefore r[! II element E is horizontal, vertical, 45°.

In the case of an extreme point in any one of the 135° diagonal directions, 0R-1 outputs "1", the pixel E is held in the peak register PR, and is output as an extreme point K. In this way, if the pixel of interest E is an extreme point, it can be extracted.

As shown in FIG. 18, the threshold processing section 15 consists of a data shift circuit and a threshold circuit, and the data shift circuit 62 uses the same circuit as that of the pole extraction processing section. The latter is comparator 6
6, has an AND circuit 67, a register 68, and a comparator 66
The slice level set in advance and the pixel of interest (for example, F
I), and when the pixel of interest has a density greater than the slice level, the comparator 66 outputs "1" and turns on the AND circuit 67. This signal causes the pixel of interest E
By controlling the level of , it is possible to extract only pixels that are at the extreme points in each of the horizontal, vertical, 45°, and 135° directions at a certain level or higher.

In this way, in the present invention, the extremal point extracting section 14 extracts contour candidate points, and the threshold processing section 15 performs threshold processing.

In other words, the contour of the object to be contour extracted is considered to correspond to a portion where the brightness changes suddenly, in principle. Therefore, a Laplacian filter (second-order differential) or a gradient filter (first-order differential) is used.

In the data after applying boundary line extraction), the part that corresponds to the outline of the object has a higher level than other parts.

So, for the data after filtering, horizontally.

Pole processing is performed in the vertical, 45° diagonal, and 135° diagonal directions to extract contour candidate points whose brightness suddenly changes on the original image. If this is all that is done, a lot of background noise due to external factors will be extracted. However, the background noise level is 1
Since they are usually at a lower level than the target contour points, they can be removed by threshold processing. The purpose of the threshold processing in the present invention is not to directly extract contour points from the data after filtering. It is enough to remove noise.

This process can be performed in parallel with the polar point extraction process.

By arranging the extreme point extraction section 14 and the threshold value processing section 15 in parallel, the processing speed can be increased. The result is horizontal.

At the vertical, 45° diagonal and 155° diagonal poles.

In addition, points above a certain slice level can be extracted as contour candidate points as a binary image.

By the way, in the threshold processing in the threshold processing unit 15, the slice level was set so that the sequence of contour candidate points is not interrupted, so some whisker-like noise and isolated noise of several pixels or less remain as shown in FIG. . Therefore, assuming that the contour candidate point sequence extracted as a binary image constitutes a closed loop with at least 8 connections, sink-like noise and isolated noise can be eliminated by logical operations within a 5 x 3 local window. Perform removal. Removal of each of these noises is performed by the noise removal section 16.

Before explaining the detailed structure of the noise removal section 16, its operating principle will be explained.

As shown in Figure 19 - 0 ＜+ 3×5 window W
If the pixel of interest E at the center of is not noise, it is always a closed loop image. Since the closed-loop image is continuous, this closed-loop image necessarily crosses the window W, and it can be said that at least two of the surrounding pixels within the window W must be pixels on the closed-loop image. Taking the above preconditions into account, it can be said that the two peripheral pixels on the closed loop image must have a positional relationship that is not adjacent to each other. from these things,
In order for the pixel of interest E to be a pixel on a closed-loop image, 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.

There are 16 cases that meet this condition, and each case is represented by a window W in FIG. 19. In this case, in order to simplify the explanation, it is assumed that if the closed-loop image has a corner, the pixel corresponding to the corner is treated as not a pixel on the closed-loop image.

Thereafter, a pattern in which the basic pattern does not occur within the window W shown in FIG. 19 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 FIG.
), when the pixel of interest E is in the state 1, as shown in fbl (■ to ■), at least the surrounding pixels along two adjacent sides of the window W are in the state 0, i.e. , peripheral pixels A, B, C, D, and G along the upper side e1 and left side 12 of the window W are in the state 0. Surrounding pixels A and D that appear on the left side 12 and bottom side 13 of window W
,G.

H, I are in state 0. Lower side e3 and right side 14 of window W
The peripheral pixels G, H, I, F, and C along the line are in state 0. Right side I of window W! 4 and the surrounding pixels along the upper edge.

One example is that F, C, B, and A are in state 0. As the second removal condition, Fig. 20 (C) ■～■
As shown in , 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 in the state 0, and the window W
The peripheral pixels located at the center of the opposite sides of are in state 19
For example, peripheral pixels A and B along the upper side 11 of the window W
, C and the surrounding pixels adjacent thereto, F are in state 0, and the surrounding pixel H is located at the center of the lower side 13 of the window W.
For example, the state remains in state 1. Note that the pixels marked with an X in FIG. 20 indicate that they may be in either state 1 or state 0.

Therefore, if the pattern within the window W that moves the binary image matches either of the above-mentioned first and second noise removal conditions, even if the pixel of interest E is in state 1, the pixel of interest E is not a pixel of a closed-loop image, but is detected as noise, and by setting it to "0", this noise can be effectively removed. For example, the third
As shown in the figure, noise N is a cluster of 4 or fewer pixels.
If the noise N1 exists, it will be removed under the first noise removal condition, and if the whisker-like noise N1 from the closed-loop image exists, it will be removed under the second noise removal condition.

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

This noise removing section 16 is as shown in FIG.

It is comprised of a data shift circuit 70 to which the binary image from the spatial filter/threshold processing section 69 is input, and a logic operation circuit 71.

First spatial filter calculation unit 12. Second spatial filter calculation unit 16. Pole extraction unit 14. Threshold processing unit 15 distaff 2, 7
5 is input to the data shift circuit 70 configured by
As a result, iiijgA-I is output to each register R; to J, and input to the logic operation circuit 71.

Here, the logic operation circuit 71 is configured as shown in FIG.
FIG. 20(b), (When the first noise removal condition or the second noise removal condition of C1 is satisfied, the level of the target pixel E is output as the state υ; otherwise, K is the level of the target pixel E. The logic operation circuit 71 outputs the level of 1 as it is in state 1, and the logic operation circuit 71 sequentially obtains binary th image data from which noise has been removed.Specifically, the logic operation circuit 71 It is configured to satisfy the following logical expressions fal to tdl.

E, -((AuBuClJDl, 1G)A (At/D
UGIJHLj I) Convex (GtJHLIILJFUC)
11 (AIJBIJCυFrie))rlB((AIJ
BllCtJAUDUG) (1(ljDUGL3Gυ)
IIJI) convex (ouHuIucuFuI) n (AuBu
CuCυFUI))nE −((XIUX2)n(X2tlX3)n(X3LjX
t)11(X4LIX1))nE・・・・・・・・・
・・・・・・ (a) E2=(((AtJBtJCLI
DUF)UH)II((AtJDtJGUBLJH)I
JF)n((DIJFtJGIJHLII)LJB)n
((CuFLJItJBLIH)tJD))nE=((
(X, tJDLIF) tJH) n((X, tlBtlH
)1. IP)n((XsvDup)up)n((x, u
Bun) IJD)) to E Hit...
...Tap...(blE-E, fih2
...river...mountain fc)Xl-
Au BυC, X2=A'J DLI G, X5=
(J*HOI.

XkeC%JFUI Hit...
... River (dl(al to (dJ formula), A to I indicate the data of pixels A to I, B, D, F
, H is pixels B, D, F.

Indicates the inverted data of H + "1 and E2 are index values that indicate whether the first noise removal condition and the second noise removal condition are satisfied, respectively, and E indicates whether or not the attention level E is noise. This is the index value shown.

FIG. 25 shows a specific configuration of the logical operation circuit 71 that satisfies this logical formula. In Figure 26, 0R-1' to 0
R-14' is an OR circuit, AN-1' to AN-3'
is an AND circuit, and NT-i' to NT-4' are NOT circuits.

For example, if the target pixel E and other pixels are under the conditions shown in FIG. As a result, AN -1' and AN -3' are ro
Therefore, the pixel of interest E is output as "0".

In addition, in the case of the conditions shown in Figure 20 (cl■), 0R-1
', 0R-5' output '0', and NOT circuit NT-1' also outputs '0', so OR circuit 0R-11
' outputs "0". Kono result AN -2', AN
-6' outputs rOJ, so the pixel of interest E is rOJ
This will be output as follows. Similar to the other circuits mentioned above, such control is sequentially performed in synchronization with the horizontal synchronizing signal output from the scanning signal generating section 2o in FIG. 4.

In this way, the logic operation circuit 71 operates as shown in FIG.
l, (condition shown in C1, that is, when there are no pixels on the upper and right sides, the right and lower sides, the lower and left sides, or the left and upper sides of window W (noise removal condition 1).

Also, there are no pixels on the top side, left or right, and there are pixels at the bottom. There are no pixels on the left side, top or bottom, and there are pixels on the right side. There are no pixels on the bottom and left and right sides, and there are pixels on the top. Or if there are no pixels on the right side, top or bottom, but on the left (
Under noise removal condition 2), the pixel of interest can be regarded as noise and removed.

In this way, the contour image of the object from which noise has been removed can be stored in the image memory 19.

The changeover switch unit 17 connects the first spatial filter calculation unit 12 and the second spatial filter calculation unit 13 in series or in parallel.2For example, the operator may manually switch this from the panel, or the switch may be switched automatically. It may also be controlled. The state in FIG. 4 shows a case of series connection.

If the first spatial filter calculation unit 12 and the second spatial filter calculation unit 13 are connected in series, the data is stored in the load coefficient table 22 from the CPU 21 via an internal path, for example. Two of the weight coefficients shown in FIG. 13, for example, the weight coefficient for the smoothing filter is sent to the first spatial filter calculation unit 12, and the weight coefficient for the second-order differential filter is sent to the second spatial filter calculation unit 16. By performing 2p2cyan calculation, it is possible to perform processing that makes the outline of the object appear raised.

In addition, the first spatial filter calculating section 12 and the second spatial filter calculating section 13 are connected in parallel to the absolute value calculating section 18.
B-1) to the @22 empty filter calculation unit 13 with a weighting coefficient of 1.
B-2) is sent.

It can perform gradient calculations and is effective when you want to extract directional straight lines. For example, you can easily extract diagonal lines.

The reason why the present invention can suppress the occurrence of the discontinuous portion No. in FIG. 2 is as follows.

That is, as shown in FIG. 2, when three or more areas intersect at one point, a level higher than the intersecting pixel level may appear in surrounding pixels. For example, in FIG. 24, two pixels or H may be at a higher level than E. In this case, the 9th pixel E should have a higher level than pixels A and I (that is, since pixels A and I are internal pixels)
, such a case can be covered by performing polar point extraction in an oblique direction.

For example, when there is original image data as shown in FIG.
This is subjected to amplasia/conversion processing to obtain data as shown in FIG. This is the conventional horizontal.

If the poles are extracted only in the vertical direction, the circled portions will be extracted. At this time, although 51 in the window W in FIG. 26 is originally a pole, it cannot be extracted as a pole when compared in the horizontal direction, but if the pole is extracted diagonally as in the present invention, + (e), ( 51), this can be extracted as an extreme point by comparing (-*s).

Even where the three regions intersect, the extreme points can be extracted correctly without creating any discontinuities.

In the image processing apparatus of the present invention, by synchronizing the data shift of the data shift circuit with the horizontal scanning signal of the TV camera, the grayscale image processing for one frame can be performed by 2, for example, 16
．． It can be completed in 7ms. In addition, by providing an address generation section after the noise removal section, if the pixel is at a pole in the horizontal, vertical, 450 diagonal, and 135 degree diagonal directions, is at a certain slice level or higher, and does not satisfy noise removal condition 1.2. If only the address information is transferred to the computer side, only the necessary contour parts can be stored in the memory, making it possible to dramatically save memory.

In addition, in the above explanation, the case where a 6×6 spatial filter was used was described, but the size of the filter was changed to 5×
By increasing the number to 5.787, etc., detailed processing can be performed.

In addition, by making this image processing device a one-chip VLSI, it will be possible to recognize two-dimensional shapes using a microcomputer-level computer in fields that require visual recognition, such as object classification and product inspection. become.

〔Effect of the invention〕

According to the present invention, it is not only possible to extract the outline Q of an object at high speed in real time in synchronization with the horizontal scanning signal of the RV camera 2, but also to extract an uninterrupted outline. . Furthermore, it is possible to remove whisker-like noise, isolated noise, and the like.

Since the two spatial filter calculation units can be selectively connected in series or parallel, it is possible to perform processing suitable for the purpose of use.

[Brief explanation of drawings]

FIG. 1 is a schematic diagram of a conventional image processing device, and FIG. 2 is a schematic diagram of a conventional image processing device. Fig. 3 is a diagram explaining the problem, Fig. 4 is a configuration diagram of an embodiment of the present invention, Fig. 5 is an illustration of an object and its contour extraction state, and Fig. 6 is an illustration of output when changing the slice level. , 7th
The figure is an explanatory diagram of the principle of a spatial filter, FIG. 8 is an explanatory diagram of the spatial filter calculation section, FIG. 9 is an explanatory diagram of the data shift circuit, FIG. 10 is an explanatory diagram of the multiplication circuit, and FIG. 11 is an explanatory diagram of the addition circuit.
Fig. 12 is an explanatory diagram of the division circuit, Fig. 13 is an explanatory diagram of the purpose of the spatial filter and its weighting coefficient, Fig. 14 is an explanatory diagram of the basic operation of the pole extractor, Fig. 15 is a configuration diagram of the pole extractor, 16
The figure shows a data shift circuit, FIG. 17 shows a data comparison circuit, FIG. 18 shows a threshold processing section, and FIG. 19 shows an explanation of contour data.
Fig. 20 is an explanatory diagram of noise removal conditions, Fig. 21 is a noise removal section, Fig. 22 is a data shift circuit, Fig. 23 is a logic operation circuit, and Figs. 24, 25, and 26 are illustrations of interrupted parts. It is. In the figure, 10 is a TV camera, 11 is an A/D converter, 1
2 is a first spatial filter calculation section, 13 is a second spatial filter manipulation section, 14 is a pole extraction section, 15 is a threshold processing section, 16 is a noise removal section, 17 is a changeover switch section, 18 is an absolute value calculation section, 19 is a 1iki image memory, 20 is a scanning signal generator,
21 is a CPtJ, 22 is a load coefficient holding section, and 30 is a data shift circuit. 31 is a multiplication circuit, 62 is an addition circuit, and 36 is a division circuit. Patent Applicant Fujitsu Ltd. Representative Patent Attorney Akira Yamatani Eisai 18 f+q flash ■ ■ ■ ■ ■ [Phase] ■ @0 [Phase] ■ ■

Claims (2)

[Claims] (1) In order to process a grayscale image pixel by pixel and extract its outline, a data shift circuit that holds and stores a certain length of data, a multiplication of the held data by a preset weighting coefficient η, A spatial filter calculation unit includes a multiplication circuit that performs the multiplication, an addition circuit that adds the results, and a division circuit that keeps the increased level due to multiplication and addition within an appropriate range; In an image processing device having a pole point extracting unit that extracts pole points,
The TV is equipped with a polar point extraction unit that extracts the polar point in the fF direction, a threshold processing unit that removes background noise, and a noise removal unit that removes whisker-like noise and isolated noise of several pixels or less.
An image processing device characterized in that the contour of an Isodan image is extracted as a binary image in real time in synchronization with a horizontal scanning signal of a camera. (2) providing a first spatial filter calculating means and a second spatial filter calculating means as the spatial filter calculating section;
Claim 1 (1) characterized in that the first spatial filter performance means and the second spatial filter calculation means can be selectively connected in series or in parallel.
) Image processing device described in section 4゜