JPS61251969A - Image processing method - Google Patents

Abstract

PURPOSE:To classify an area to which a factor belongs or an interface and to obtain the excellent classified result on the overall surface of a light and dark image by setting the factor in such a way that one factor and picture elements in the factor adjacent to the former factor can overlap by the half when a means density and its variance in the factor are calculated with plural picture elements of an input image as one factor. CONSTITUTION:An adder 29 adds the contents of an XH address counter 25 and those of an XL address counter 26, which turns out the X address of an image memory A2. The contents of an YH counter 27 and those of an YL counter 28 are added 30 to be the Y address of the memory A2. A mean density/variance control part 24 reads the density values of NXM picture elements out of the memory A2, calculates the mean density -x and the variance sigma<2>, writes them in picture memories E22 and F23, respectively, after which it adds 31 the contents of the counter 25 to an M/2, sets the result to the counter 25, and advances an NX address counter 33 by one. When the counter 33 comes to W-1, the counter 25 is cleared, and an NY address counter 34 is advanced by one. Then the contents of the counter 27 are added 32 to an N/2, and the result is set to the counter 27, thereby obtaining the mean density and the variance.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to an image processing method for dividing a grayscale image into regions.

(Prior Art) An example of an image processing apparatus to which a conventional image processing method is applied is the one described in Japanese Patent Application No. 59-250314. A conventional image processing method will be explained using FIG. 3, which is a block diagram of this image processing apparatus.

The image input device 1 quantizes the density value of the input grayscale image using L bits per pixel (here, L=8). Image memo IJA
2 is a memory for two-dimensionally storing the input image as shown in FIG.

The average density/variance calculation circuit 3 converts (NXM) pixels of the image stored in the image memo 1JA2 into one element (here, N=M=
4) Calculate the average and variance of the density values of (NXM) pixels within each element. The output of the average density/variance calculation circuit 3 is connected to the image memo IJB4 and the image memory C5, and the image memo IJB4 has an average density of 2 as shown in FIG.
On the other hand, the image memo IJC 5 also stores the variance calculated by the average density/variance calculation circuit 3 two-dimensionally. The threshold circuit 6 reads out the information stored in the image memo IJB 4 and the image memory C5, respectively, starting from the upper left element, for example, by raster scanning. Then, the read variance value of the current element is smaller than the preset threshold θ7, and the difference between the average density of the current element and the average density of the element immediately above the current element is set in advance. Threshold θ. 1# if the difference between the average density of the current element and the average density of the element immediately to the left of the current element is smaller than the threshold θゎ.
is written to the corresponding position in the image memory D7. For other operations, write °0" in the corresponding position. In other words, the element labeled °'1"' in image memory D7 has a uniform distribution of density values and corresponds to the object. On the other hand, on the image memory D7, “
Elements labeled '0'' have uneven distribution of density values or the average density varies greatly between adjacent elements, and are near the contour of the object (hereinafter referred to as boundary).

Next, the average concentration/variance calculation circuit 3 will be explained in detail. The average density of (NXM) pixels within each element is calculated according to (1).

However, X + + is the density value of each pixel. On the other hand, the variance σ2 of (NXM) pixels within each element is calculated using equation (2).

Expanding equation (2) yields equation (3).

FIG. 5 is a block diagram showing an example of the configuration of the average concentration/variance calculation circuit 3. As shown in FIG.

There are two types of addresses in the image memory A2: X and y. X
The address counter consists of an XH address counter 9 and an XL address counter 10.

Also, the X address counter is YH address counter 1.
1 and a YL address counter 12. XL
The address counter 10 is a counter that counts from 0 to (M-1), and the YL address counter 12 is a counter that counts from O to (N-1)t. Therefore, XL address counter 10 and YL address counter 1
(NX
M) Pixels can be read out.

By the way, when M is a power of 2 (that is, M =
1, 2, 4. , 8, 16, . . .), the X address of the image memory A2 can be controlled separately by the lower XL address counter 10 and the upper XH address counter 9. In other words, (NXM
) After reading out the pixels, by incrementing the XH address counter 9 by one, it becomes possible to read out the (NXM) pixels of the next element. Therefore, the XH address counter 9 is a counter that counts from O to (Q-1)'1. However, the size of image memory A2 (RXS
) When starting from a pixel, Q=S/M.

Similarly, the YH address counter 11 is from O to (P-1)
This is a counter that counts up to. However, P = R/N
It is.

On the other hand, the XH address counter 9 is incremented by one every time (NXM) pixels of the image memory A2 are read.
This is the X address of image memory B4 and image memory C5. Similarly, YH address counter 11 becomes the X address of image memory B4 and image memory C5.

The average density and variance σ2 of (NXM) pixels in each element from the image memo IJA2 are sent to registers 13 and 14, and adder 1.
5, divider 16, multiplier 17, adder 18, divider 19
, a multiplier 20 and a subtracter 21 (indicated by a broken line in the figure).

Average density/dispersion control unit 8, XH address counter 9,
L address counter 10, YH address counter 11,
By controlling the YL address counter 12, the image memo 1JA
(NXM) pixels for each element are read out from 2, and the average density ma and variance σ2 are calculated by controlling the registers 13 and 14 of the calculating section.

Furthermore, this calculation result is stored in the image memo IJB4 and the image memory C5.

Next, the operation will be explained.

The average density/dispersion control section 8 firstly controls an XH address counter 9, an XL address counter 10 . A clear signal is supplied to the YH address counter 11 and the YL address counter 12 to clear these address counters. next,
The average density/dispersion control section 8 supplies a clear signal to the register 13 and the register 14 to clear the contents of these registers.

The average density/dispersion control unit 8 has four address counters (X
H address counter 9, XL address counter 10, Y
H address counter 11 and YL address counter 1
2), the adder 15 adds the contents of the register 13 and the read density value, and stores the result in the register 13. Next, the XL address counter 10 is incremented by one, the density value of the adjacent pixel is read out, and the same operation is performed. In this way, the XL address counter 10 is advanced one by one, and the addition amount of the density value of each pixel is determined from the force. When the XL address counter 0 reaches (M-1), the average density/dispersion control unit 8 does not advance the XL address counter 0 by one, but clears the XL address counter 0 and increments the YL address counter 2. Advance only one step. In the end, image memo IJA2
When read out (NXM) times, the register 3 holds the result of addition of the density values of (NXM) pixels of the image memo IJA2. Therefore, the average density Ma is divided by JE unit 1
．． 6 by dividing the contents of register 3 by a constant NM.

On the other hand, each time the average density/dispersion control unit 8 reads out the density value XI of each pixel from the image memory A2, the multiplier 17
Perform multiplication calculation XIj2, add the calculation result to the contents of register 4 in adder 18, and add the sum.
Set to 4. Therefore, like register 3, register 4 holds the result of Σ(X, ρ2. The variance σ2 is calculated using equation (3), but the first term of equation (3) is 4 by the constant NM in the divider 19. On the other hand, the second
The term is the output (average density) X of the divider 16 and the multiplier 20
It is obtained by multiplying. Therefore, the variance σ2 is calculated from the first term of equation (3) by the subtracter 21 to the second term of equation (3).
Calculated by subtracting terms.

The average density/dispersion control unit 8 is connected to the image memory A2 (NX
After reading out M) pixels, writing the average density map of (N Perform processing. When the XH address counter 9 reaches (Q-1), the average concentration/dispersion control unit 8 clears the XH address counter 9 and advances the YH address counter 11 by one.

In this manner, the average density/dispersion control unit 8 obtains the average density value and variance value of the (PXQ) element by calculating the average density and variance over the entire surface of the image memo IJA2.

(Problems to be Solved by the Invention) However, the image processing method has the following problems.

If the outline of the object is exactly present between the elements, the variance values of these two elements will both be small values (that is, values smaller than the threshold value θ7). In such a case, a threshold value θ for the difference in average density between two adjacent elements. This element must be classified as a boundary by setting it to an appropriate value. But from Moth, a certain threshold θ. However, this method has the disadvantage that it is not possible to obtain good element classification results (regions or boundaries) for the entire screen.

The present invention makes it possible to classify the object into regions or boundaries even when the outline of the object exists between the elements described above, and to achieve good classification results for the entire grayscale image. The purpose of the present invention is to provide an image processing method that can obtain the following.

(Means for Solving the Problems) In order to solve the above problems, the present invention converts the density of each pixel of an original image into an electrical signal and stores it, and an element composed of a plurality of stored pixels. In an image processing method that performs image processing by calculating the average density and variance of density values of and classifying into area constituent elements and boundary elements based on the calculation results, each element is partially connected to an adjacent element. They are set so that they overlap.

Preferably, each element is set to overlap an adjacent element by half.

(Function) According to the present invention, since the image processing method is configured as described above, it functions as follows. ``Pixels within each element are set to partially (for example, only half) overlap the pixels of adjacent elements, and the average density and i displacement are calculated for each element. Therefore, it becomes impossible for the boundary of the object to exist exactly between the elements,
It becomes possible to classify whether each element belongs to the region of the object or to the boundary. Further, by using a certain threshold value, it is possible to obtain good classification results for the entire screen of grayscale images.

(Example) Figure 2 explains how adjacent elements overlap11)
,.

It is something that Element a and element S are adjacent elements (
-NXM) Only pixels overlap. In conventional image processing methods, elements (a, c,...
), but in the present invention, the average density and variance are calculated for elements that overlap (a, b, c, d, . . . ). Furthermore, the processing in the vertical direction is exactly the same as the processing in the horizontal direction.

FIG. 1 is a block diagram showing an embodiment of the present invention. In this figure, the same reference numerals as in FIG. 5 indicate the same components. The calculation section (within the broken line in the figure) that calculates the average density and variance for each element from the pixels read out from the image memory A2 is the same as in FIG. Therefore, the average concentration/dispersion control section in FIG.

This calculating section is similarly controlled by the average concentration/dispersion control section 24 corresponding to 8, and calculates the average concentration and variance σ2 within each element.

Next, components different from those in FIG. 5 will be explained.

The image memory E 22 corresponds to the image memory B 4 in FIG. 5, and is a memory for storing average density. Also, (12)
゛The image memory F23 corresponds to the image memory C5, and is a memory that stores variance. Image memory E 22 and image memo!
The size of JFv23 is the (VXW) element.

(See Figure 2). However, V = 2R/N, W =
It is 2S/M. This is because pixels in adjacent elements are set to overlap by exactly half.

The XL address counter 26 is a counter that counts from O to (M-1), and the YL address counter 28 is a counter that counts from O to (N-1). Therefore, by controlling the XL address counter 26 and the YL address counter 28, (N
XM) pixels can be read out.

Here, even if M is a power of 2, the X address of the image memory A2 cannot be controlled by the conventional method shown in FIG. In other words, the upper XH address counter 25 and the lower XL address counter 26 are directly stored in the image memory A.
It cannot be an X address of 2. The contents of the XH address counter 25 and the contents of the XH address counter 26 are added by an adder 29, and the result is set as the X address of the image memory A2. Similarly, the contents of the YH address counter 27 and the contents of the YL address counter 28 are added by an adder 30, and the result is set as the X address of the image memory A2. The XH address counter 25 is a counter that counts from O to <s-7>, and has its own output and a constant M/2.
By adding the values in the adder 31, the value of every N/2 (0
, N/2, M.

3B4/2, 2M,...). Similarly YH
The address counter 27 is a counter that counts from 0 to (RT), and by adding its own output and a constant V2 using an adder 32, the address counter 27 calculates a value every N/2 (0, N/
Take 2°N, 3N/2, 2N,...).

NX address counter 33 and NX address counter 34
are address counters of image memory E 22 and image memory F 23.

Average density/dispersion control unit 24 and XH address counter 25
, XH address counter 26, YH address counter 2
7 and the YL address counter 28, the X and Y addresses of the image memory A2 are designated and the stored density values of the pixels are read out, and the registers 13 and 14 of the calculation section are controlled to The average concentration and variance are calculated in the same way as in the conventional example shown in FIG. Further, the NX address counter 33 and the NY address counter 34 are controlled to store the calculation results for each element in the image memo 1JE 22 and the image memo IJF 23.

Next, the operation will be explained. The average density/dispersion control unit 24 first controls the XH address counter 251. A clear signal is supplied to the XL address counter 26, YH address counter 27, and YL address counter 28 to clear these address counters.

The average density/dispersion control unit 24 also supplies a clear signal to the NX address counter 33 and NY address counter 34 to clear these address counters.

The address of image memory A2 is generated as follows. The contents of the XH address counter 25 and the contents of the XL address counter 26 are added by an adder 29;
The output becomes the lX address of image memory A2. Further, the contents of the YH address counter 27 and the contents of the YL address counter 28 are added by an adder 30, and the output of the adder 30 becomes the Y address of the memory A2.

Next, the average concentration/dispersion control section 24 supplies a clear signal to the register 13 and the register 14 to clear the contents of these registers. The average concentration/dispersion control section 24 is XL
The address counter 26 and the YL address counter 28 are controlled in the same manner as before, and from the image memory A2 (NX
M) Read the density value of the pixel and calculate the average density value and variance σ2 of (NXM) pixels. Next, the average density/dispersion control unit 24 records the average density as an image memo! Write in JE22 and take an image memo of the variance σ2! After writing to JF23, XI
(The contents of the address counter 25 and the constant M/2 are added to the adder 3.
The result is set in the XH address counter 25, and the NX address counter 33 is incremented by one. Therefore, the XH address counter 25 is V
2, and the Y)I address counter 27 is 0, so the next element to be processed is exactly half of the previously processed element and the (NXM) pixels of image memory A2.
It means it matches.

By controlling as described above, the average concentration/dispersion control section 24 controls N
When the X address counter 33 inputs (w-i), NX
Clear the address counter 33 and the XH address counter 25, advance the NY address counter by one, add the contents of the YH address counter 27 and a constant N/2 using the adder 3.2, and add the result to the YH address counter. Set it to 27. Therefore, YH address counter 27 is N/2, and XH address counter 25 is O.
Therefore, in the next element to be processed, exactly half of the (NXM) pixels of the image memo IJA2 coincide with the element immediately above it.

In this way, the average density/dispersion control unit 24 calculates the average density and variance over the entire surface of the image memo 1JA2, thereby obtaining the average density value and variance value of the (VXW) element.

(Effects of the Invention) As described above in detail, according to the present invention, when calculating the average density and variance within an element using a plurality of pixels of an input image as one element, the pixels within one element and its elements are Since the elements are set so that the pixels in the adjacent elements overlap by half, it is impossible for the boundary of the object to exist exactly between the elements, and each element It becomes possible to classify whether the image belongs to a region or a boundary, and it is possible to obtain a good classification result for the entire screen of a grayscale image by using a constant threshold value (θD).

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is an explanatory diagram showing the correspondence between pixels and elements of the present invention, and FIG. 3 is a block diagram showing an image processing apparatus to which a conventional image processing method is applied. figure,
Figure 4 is an explanatory diagram showing the correspondence between conventional pixels and elements;
The figure is a block diagram showing the configuration of a conventional average concentration/variance calculation circuit. 1... Image input device, 2... Image memo IJ A, 3
...Average density/dispersion calculation circuit, 4...Image memo IJ
B, 5... Image memo IJ C16... Threshold circuit, 7
...Image memory D. 8... Average density/dispersion control unit, 9... XH address counter, 10... XL address counter, 11...
YH address counter, 12... YL address counter, 13... register, 14... register, 15...
・Adder, 16... Divider, 17... Multiplier, 18...
- Adder, 19... Divider, 20... Multiplier, 21...
...Subtractor, 22...Image memory E123...Image memory F124...Average density/dispersion control section, 25...
XH address counter, 26...XL address counter, 27...YH address counter, 28...YL
Address counter, 29...Adder, 30...Adder, 31...Adder, 32...Adder, 33...N
X address counter, 34...NY address counter.

Claims (2)

[Claims] (1) Convert the density of each pixel of the original image into an electrical signal and accumulate it, calculate the average density and variance of the density value of an element composed of multiple accumulated pixels, and based on the calculation results, 1. An image processing method for performing image processing by classifying an image into constituent elements and boundary elements, characterized in that each element is set to partially overlap an adjacent element. (2) The first aspect of the present invention is characterized in that each element is set to overlap an adjacent element by half.
Image processing method described in section.