JPS61128384A - Picture processor - Google Patents

Abstract

PURPOSE:To distribute an input density picture into an objective substance separating areas and to executed picture processing at a high speed by regarding plural picture elements as one element, counting up the average density and dispersion of each element and deciding an area including the element. CONSTITUTION:The density value of an input density image quantized by 4 bits of one picture element outputted from a picture memory 13 is processed by an average/ dispersion calculating circuit 14 by regarding MXN picture elements, e.g. 4X4, as one element and the average density and dispersion are two-dimensionally stored in picture memories 15, 16 respectively. The contents of these memories 15, 16 are processed by a threshold circuit 17, and when the fraction of these read elements is the threshold or less and the average density difference between the element and its adjacent element is the threshold or less, the element is decided as the objective substance constitutional element and '1' is written in a picture memory 18 two- dimensionally. In other cases, the element is decided as a border area and '0' is written. Then, the coupled property of two picture elements is decided by a coupling circuit 19. The decided result is written in a picture memory 20. Thus, the picture can be distributed into the areas for separating the objective substance and the picture processing can be attained at a high speed with the small volume of calculation.

Description

[Detailed description of the invention] (Industrial application field) The present invention relates to an image processing device, and more particularly, to an image processing device.

The present invention relates to an image processing device capable of dividing a grayscale image into regions for cutting out a target object and performing high-speed image processing.

(Prior Art) A conventional image processing device is shown in FIG. 7. 1-1 is an image processing device that predicts the density value using 4 bits per pixel, and 1-2 is an image storage device that stores the processing image. In memory, the image stored in 1-2 (NXM) pixels.

1. Assign the average density value (1-3) within the element to each element as one element (here, N = M = 12). The first abstract image consisting only of rough density information is stored in the image memory 1.
(1-4) Next, calculate the difference in average density between elements based on the image stored in (1-4) (1-5
), here the average concentration of the current element is D IIJ e '
The average concentration value of the neighborhood is D I, j-t*Dl-t, IyD
l, j hle D When I*z+1 (see Figure 8), set the value to an appropriate value (174 of full scale, or 4N
Is it normalized using 2)?

1@aX (D+* I -DI-J* I -m)
E, normalize the value of m+-1,◆1 with an appropriate value (3″N″/2) and express the initial value of the main contour strength as the probability Xe P +r ″′,
It is stored in the image memory 2 (1-6) as a second abstract image. However, if it takes a value of 1 or more, it is treated as 1.

Based on the second abstract image, main contour elements are extracted using the relaxation method (1-7). Each of the second abstract images (iI
J) From the element values, the initial image space is P”)=(P+'i'l 1≦i≦n, 1≦j≦m;i
, j integer), and perform repeated operations using the following equation. That is.

qIJ(k)=p,,(kl (1-Δ,,(kl)
, where the physical quantity ΔI J (kl is related to the connectivity of each element of the second image, and the value p + of each image element is
+ + ') is updated by its own contour strength Pth (k) and connectivity Δth (k).

This repeated operation is discontinued when the number of elements with a value of 0.9 or more or 0.1 or less reaches 0.95 or more of the total number of elements. -6). The second image is an image composed of major contour elements.

Next, perform region segmentation based on the second image (1-8)
. The main contour elements are spatially integrated by 8 connections and divided by circumscribed rectangles. The representative points are stored in the image table (1-9).

Referring to the image table (L-9), label each small area (1-10). Labeling is performed as follows. Calculate two physical quantities based on the image. That is, (1) It was determined from the difference in average density on both sides of the main contour element. Average value of change values within the region (ΔD); (2) Ratio (DN) of the number of main contour elements occupying the region. “Photo” when ΔD≧0.7 on ΔD#DN plane, ΔD<0.7 and D
When H < 0, 4, label "graph", and when ΔD < 0.7 and DH > 0.4, label "character", and as a result, obtain the image table (1-9) shown in Figure 9. .
- This conventional technique has a drawback that a character string and each character string are not described in an organized manner starting from an area determined to be 9 characters.

Therefore, the present applicant has proposed Japanese Patent Application Laid-Open No. 59-43468 (Japanese Patent Application No. 153992-1982) to solve such drawbacks.
proposed an image processing device that performs the following processing.

That is, this image processing device further performs orthogonal transformation on the area determined to be a character, calculates a rough pitch, and extracts a character string.Then, the size of the extracted character string,
``Title'' and ``Body'' based on the number of lines of positional similar pitches
It is classified into

(Problems to be Solved by the Invention) However, in the conventional technology as described above, the input grayscale image is classified into three types: a photo area, a graph area, and a text area, and the text area is further divided into "title" and "body text". Although it is possible to divide the inside of a photographic area into two types, it is impossible to divide the interior of a photographic area or to divide a gray-scale image from a black and white camera into areas in order to extract the target object. Furthermore, in the prior art, main contour elements are extracted using a relaxation method, which has the drawback of requiring a large amount of time due to repeated calculations and slowing down the processing speed.

The present invention was made in order to eliminate these drawbacks of the conventional technology, and its purpose is to be able to divide a gray scale image into regions for cutting out a target object, and to use raster scanning in image processing. An object of the present invention is to provide a high-speed image processing device that can be suitably used.

(Means for Solving the Problems) In order to solve the problems of the prior art, the present invention is comprised of an accumulation means, an average concentration/variance calculation means, a classification means, and a connection means.

The storage means converts the density of each pixel of the input image into an electrical signal and stores it. The average density/variance calculation means calculates the average density of an element composed of a plurality of pixels stored in the storage means, and also calculates the variance of the density values of a plurality of pixels in the element. The classification means classifies each element into a region component, which is a region of the target object, and a boundary region, based on the magnitude of the variance value calculated by the average concentration/variance calculation means and the magnitude of the difference in average concentration between adjacent elements. . The connecting means connects the elements classified into area constituent elements by the classifying means and assigns a unique number to the connected area.

(Function) According to the present invention, since the image processing apparatus is configured as described above, each technical means functions as follows.

The storage 1 means quantizes and stores the grayscale information of the input original image for each pixel. For example, the average density/variance calculation means calculates the average density within an element using (NXM) pixels of the input image as one element, calculates the variance of the density values of the (NXM) pixels within the element, and uses these calculation results. Output to classification means.

The classification means determines whether each element has a uniform concentration distribution based on the magnitude of the variance of each element, and determines whether adjacent elements have connectivity based on the magnitude of the difference in average concentration between the elements. It is determined whether or not each element is present, and each element is classified into area constituent elements and boundary elements. Then, the region constituent elements and the classified elements are connected by the connecting means, a unique number is assigned to the connected region, and the target object is cut out. In this way, the input grayscale image can be divided into regions for cutting out the target object, and the problems of the prior art described above can be solved.

(Example).

FIG. 1 is a block diagram showing an embodiment of the present invention. In the figure, 12 is an image input device, 13 is an image memory A,
14 is an average density/dispersion calculation circuit, 15 is an image memory 8.
16 is an image memory C117 is a threshold circuit;

18 is an image memory D, 19 is a connection circuit, and 20 is an image memory D.
It is MoIJE.

The image input device 12 is the conventional image input device shown in FIG.

Similar to 1-1, the density value of the input grayscale image is quantized by 100s4 bits. The image memory A 13, like the conventional image memory 1-2 shown in FIG. 7, is an image memory that stores an input image two-dimensionally as shown in FIG. tIs4. The average density/variance calculation circuit 14 calculates the average and variance of the density values of the (NXM) pixels between each element, with (NXM) pixels of the image stored in the image memory A 13 as one element (here, N=M=4). Calculate.

The output of the average density/variance calculation circuit 14 is stored in the image memory B1.
The average density calculated by the average density/variance calculation circuit 14 is stored two-dimensionally in the image memory B 15 as shown in FIG. The variance calculated by the same circuit 14 is also stored two-dimensionally.

The threshold circuit 17 includes an image memory B 15 and an image memory C 16.
・Read out the information stored in, for example, from the upper left element by raster scanning. Then, the read variance value of the current element is smaller than a preset threshold θV, and the difference between the average density of the current element and the average density of the element immediately above the current element is a preset threshold θ. , and 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 θO, "1" is written to the corresponding position in the image memory 018, and the other 0″ if
Write in the corresponding position. In other words, image memory D18
The element labeled "1" in , has a uniform density value distribution, is a component of the target object (hereinafter referred to as region), and is labeled "0" in the right image memory D 18. The identified elements are near the boundaries of the target object. The concatenation circuit 19 integrates the elements labeled u1〃 on the image memory D 1a by four concatenations, assigns a unique number (1 or more) to one integrated block, and writes it to the image memory E 20. Also, the number O is assigned to the element labeled 0'' on the 1-image memory D18.

Write to image memory E2G.

Next, the configuration of the average concentration/variance calculation circuit 14 will be explained in detail. FIG. 2 is a block diagram showing an example of the configuration of the average concentration/variance calculation circuit 14. In the figure, 21 is an average density/variance calculation unit, 22 and 23 are Regis registers, 24 is an adder, 25 is a divider, 26 is a multiplier, 27 is an adder, 28,
29, 30, and 31 are multipliers, 32 is an adder, and 33 is a subtracter.

The average density of (NXM) pixels within each element or calculated according to equation (1).

X==: SatΣXil (i=1~N, j=1~M)
=(1)NM However, XIJ is the density value of each pixel.

To explain the operation of the average density/dispersion calculation circuit 14, the average density/dispersion control unit 21 first supplies a clear signal to the register 22 and the register 23, clears the contents of these registers, and then calculates the average density/dispersion. Distributed control unit 2
1 supplies an address to the image memory A 13 to read the density value of each pixel, and an adder 24 adds the contents of the register 22 and the read density value of each pixel, and stores the result in the register 22. .

If this operation is repeated NXM times (here N=M=4), the register 22 will hold the result of addition of the density values of (NXM) pixels in the image memory A13. The average density is therefore obtained by dividing the contents of register 22 by (N,XM) using divider 25.

On the other hand, the variance σ'' (=V) is calculated using equation (2).

However, XII represents the density value of each pixel, and i represents the average density.

σ2=Σ(XH−)E)” (i=1 to N, j=1 to M) (2) When formula (2) is expanded, formula (3) is obtained.

σ2;Σ(x+j)"+NM'i"-2xΣ)CIJ(
i=1 to N, j=1 to M) ... (3) When calculating variance, average concentration/variance calculation control.

Each time the unit 21 reads out the density value XH of each pixel from the pixel memory A 13, the multiplier 26 performs square calculations x, , *, and the calculation results are stored in the register 23 by the adder 2.
The sum is added to the contents of , and the sum is set in the register 23. Therefore, like register 22, register 23 is Σ(x
lj)''.The second term N M M'' in equation (3) is obtained by squaring the output (average density) i of the divider 25 using the multiplier 28, and then holding the result i3 and NM by the multiplier 29.

The third term 21ΣXB in equation (3) is first
It can be obtained by calculating the xth number and then doubling the result using multiplication @31. Finally, the adder 32 adds the first term and the second term of equation (3).
The subtracter 33 adds the terms, and from the result, the third
One variance σ3 is calculated by subtracting the terms.

Every time (NXM) pixels are read out from the image memory A 13, the average density/variance calculation control unit 21 writes the average density i of the (NXM) pixels to the image memory B 15,
N) Write the pixel variance σ8 into the image memory C16. By applying this process to the entire image memory A 13, (
PXQ) Obtain the average concentration value and variance value of the element. However, when the size of image memory A13 is (RXS) pixels, P
= R/N.

Q=S/M. FIG. 4 is a diagram for explaining the relationship at this time.

Next, the configuration of the threshold circuit 17 will be explained in detail. FIG. 3 is a block diagram showing an example of the configuration of the threshold circuit 17.

In the figure, 34 is a threshold circuit control section, and 35 is a comparator.

36 is a shift register, 37 is a register, and 38 is a subtracter.

39 is an absolute value circuit, 40 is a comparator, 41 is a subtracter, 42
4 is an absolute value circuit, 43 is a comparator, and 44 is a shift register.

45 is a register, 46.47 is an AND gate, 48 is O
R gate, 49 is AND gate, 50 is OR gate, 5
I is a selection circuit.

The threshold circuit control unit 34 simultaneously reads the information stored two-dimensionally in the image memory B 15 and the image memory C 16 by raster scanning, starting from the upper left element, for example, and reads out the information stored in the image memory D in accordance with the conditions shown below each time each element is read. 1g is written in the corresponding position. First, the output of the image memory C16 (the variance value V is compared with a preset threshold θ9 by the comparator 35. The comparison @35 is 1
If the variance value V is larger than the threshold θ, (a state in which the concentration value within the element is not uniform, that is, an element considered to be near the boundary of the target object), "1" is output; otherwise (the concentration value within the element is In a state where the density value is uniform (that is, an element considered to be a region), "0" is output.

On the other hand, the output of image memory B 15 (average density D +,
,; see FIG. 8; however, DI,1 is the same amount as obtained in (1)) is stored in the -row (Q element) shift register 36, and is stored in the -element register 37.
is memorized. In other words, the intermediate value of the shift register 36 is the average density (DI-□,''; FIG. 8) of the element immediately above the element currently being read out from the image memory C16, and the output of the register 37 is the s D + l, which is the average density of the element immediately to the left (Dl, , -1; Fig. 8)
After the difference between J and Dl-1,J is calculated by the subtraction circuit 38, it is converted into an absolute value by the absolute value circuit 39. The comparator 40 is
Compare the preset threshold 0 and the output of the absolute value circuit 39, and if the output of the absolute value circuit 39 is larger than the threshold θ0 (an element where the average density changes significantly, that is, near the boundary of the target object) possible elements) outputs “1”, otherwise outputs “0”.Similarly, DI
. When the output of the absolute value circuit 42 is larger than the threshold value θ0, the comparator 43 outputs '1'.
In other cases, On is output.

Here, the shift register 44 stores threshold processing results for one row (Q elements). The register 45 stores the threshold value processing result for one element. In other words, shift register 44
The output of is the threshold processing result of the element immediately above the current element (T
i-1, I; Fig. 8).

The output of the register 45 is the threshold processing result of the element immediately to the left of the current element (TI, -1; Fig. 5).The threshold processing result is either "1" or "0";" represents the area, and '0" represents the boundary.

The purpose of this embodiment is to extract the area. Comparator 40
and the output of the shift register 44 are connected to an AND gate 46.
is input. AND gate 46 has two inputs both “
1” (i.e. T i-t, 1 is “1”
and 1D+, J-Dl-t, jl>0), outputs 1'', and otherwise outputs 0''. The relationship between the current element and the element immediately to the left is also the same; the output of the comparator 43 and the output of the register 45 are input to the AND gate 47, and the AND gate 47 has two inputs both "1".
If so, output “1”, otherwise “0”
Output. The OR gate 48 inputs the output of the AND gate 46 and the output of the AND gate 47, and if either one is "1"
”, output “1”, otherwise output “O”
” is output.

However, in raster scanning, when processing the first row, the element immediately above it does not actually exist, and the first row
When processing a column, the element immediately to the left also does not exist,
When processing the first row or column, the threshold circuit control unit 34 sets the average density usable signal to "0",
The output of the OR gate 48, which is set to "1" at other times, is ANDed with this average concentration enable signal by an AND gate 49.

The output of the comparator 35 and the output of the AND gate 49 are ORed by an OR gate 50. In other words, the OR gate 50 is
If the variance value V is larger than the threshold θV or if the difference in average density between adjacent elements is large (that is, if the current element is about to be determined as a boundary), 'i' is output; otherwise, 'i' is output. Outputs “O”.

Here, in the threshold processing results, '1' represents the area, and '0' represents the area.
represents the boundary. Therefore, the selection circuit 51 is OR
If the output of the gate 50 is ``1'', it outputs 0'' and the OR
When the output of the gate 50 is “0”, it outputs “1”.The threshold circuit control unit 34 stores the threshold processing result in the image memory D1.
Write it on 8.

Next, the connection circuit 19 will be explained.

As shown in FIG.
4 (random access memory). The program of the microprocessor 52 is written in the ROM 53, reads out one image memory D18, assigns a unique number of 1 or more to the 4 connected blocks of "1", and sets the position of "1" on the one image memory D18. Image memory E compatible with
Write the assigned number in the position on 2G,
In addition, 4-connection means DI, j element and 4-connection in Fig. 8.
Neighborhood (Dl-z, J element, Dll)-1 element, Dl, 1
．． It is said that both the DIIJ element and one element in the 4 neighborhood are "1n" in relation to the j element and one element of D I, j, element), and in this case, these two elements are 4
A block that is said to be connected, and therefore has 4 connections, is a set of ``1'', and the element of ``1'' is 4.
It means that they are connected by a neighborhood relationship.

Since one unique number is assigned to each of the four connected "1" blocks, the image memory E 20 is a number ranging from 0 to the number of "1" blocks existing on the image memory D 18. is held.

However, 0 corresponds to "0" on the image memory D18.

(Effect of the invention) According to the present invention, as described above in detail.

Since multiple pixels of the input image are considered as one element, the average density within the element is calculated, and the variance of the density values of multiple pixels within the element is calculated. It is now possible to determine whether the image is distorted or not, and to determine the connectivity of adjacent elements, thereby making it possible to divide the grayscale image into regions for cutting out the target object. Furthermore, the use of raster scanning has the advantage that processing can be performed at high speed.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of an image processing device according to the present invention, FIG. 2 is a block diagram of an average density/variance calculation circuit, FIG. 3 is a block diagram of a threshold circuit, and FIG. 4 is a block diagram showing pixels and elements. , FIG. 5 is an explanatory diagram of adjacent elements of the threshold processing result, FIG. 6 is a block diagram of a connection circuit, FIG. 7 is a block diagram of a conventional image processing device, and FIG. An explanatory diagram of average density values of neighboring pixels, FIG. 9 is a diagram showing an image table. In the figure, 12... Image input device, 13... Image memory A, I4... Average density/variance calculation circuit, 15... Image memory B, 16... Image memory C1 17... Threshold circuit, 18... Image memory D, 19... Connection circuit. 20... Image memory E. 21... Average concentration/dispersion calculation control section. 22.23...Register. 24.2F, 32... Adder, 25... Divider, 26.28.29, 30.31... Multiplier, 33...
・Subtractor. 34... Threshold circuit control unit, 35.40, 43... Comparator. 36.44...Shift register, 37.45...Register. 38.41...Subtractor. 39.42... Absolute value circuit. 46.47.49...AND gate. 48.50...OR gate, 51...selection circuit, 52...microprocessor. 53...ROM, 54...RAM.

Claims (1)

[Claims] A storage means that converts the density of each pixel of the original image into an electrical signal and stores it; and an average density/density controller that calculates the average density and variance of the density values of an element composed of a plurality of pixels stored in the storage means. a variance calculation means; a classification means for classifying the elements into area constituent elements and boundary elements based on the variance value calculated by the calculation means and the difference in the average concentration value between adjacent elements; and the classification means 1. An image processing device comprising: connecting means for connecting elements classified into area constituent elements by the above method, and assigning a unique number to the connected area.