JPS62118483A - Picture processing method - Google Patents

Abstract

PURPOSE:To carry out a proper picture processing according to the type of a picture by comparing a multivalued picture at every an aperture with the multivalued picture formed based on the ratio of a white area and a black area at every aperture and selecting one aperture at every setting position. CONSTITUTION:Picture data is stored in line memories L1-L9 by a select circuit 1 at every 1 line, 8 lines of the stored data are selected in a select circuit 2 and inputted to a halftone estimation circuit 3. The circuit outputs the estimation value at every aperture and the decision result of the aperture, subsequently, a selecting circuit 4 receives these signals, selects an optimum aperture at every one picture element, and outputs the halftone estimation value based on the aperture. Then, restored picture element data is inputted to a selecting circuit 5 and the decision result of the aperture from the circuit 3 is inputted to a selecting circuit 6 as aperture information. To the circuit 5, original picture element data from the line memories L1-L9 and a selection signal 1 are inputted and to the circuit 6, a selection signal 2 is inputted, when the aperture is small, a line drawing is obtained, and when the aperture is large, a gradation picture is obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION (Industrial Application Field) The present invention relates to an image processing method, and more particularly, to an image processing method using an aperture method. When performing i-image processing, an image processing method is used in which the type of image is identified and determined based on the size of the aperture.

(Prior Art) In the field of image processing, Li2 converts an original image into an electrical signal using a photoelectric conversion element such as a COD, and performs various image processing on the converted image signal.

It is designed to be reproduced as an image on an output device such as a printer. As a method of image processing, the God's method has conventionally been used. For example, there is a method in which various types of images are first identified and then image processing is performed according to the type of image.

(Problems to be Solved by the Invention) Since the conventional method distinguishes the image type for each block of a certain size, it is difficult to identify the image type in an area smaller than the block size. There is. Therefore, when the purpose is printing-level image processing with high resolution during imaging and recording, this is not so much of a problem, but when the target is office work, etc., there is a problem that the resolution is insufficient.

The present invention has been made in view of these points, and
The purpose is to realize an image processing method that can perform appropriate image processing depending on the type of image.

(Means for Solving the Problems) The present invention, which solves the above-mentioned problems, sets a plurality of types of apertures at the same setting position in a multivalued image consisting of a white area and a black area, moves them, and The multi-value image for each frontage is compared with the multi-value image created from the estimated value obtained based on the ratio of the white area to the black area for each opening, and one image [1] is selected for each set position. 1J is characterized in that images are classified into different types based on the size of the frontage and image processing is performed.

(Operation) In the present invention, it is determined whether the 1-1 pixel belongs to the halftone area or the line/character image area at the 1st position of 1 pixel.

(Embodiments) Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a flowchart showing one embodiment of the present invention. The flowchart will be explained below along with flowcharts 1 to 1.

(1) Step (2) A plurality of types of scanning apertures are set for each pixel in a multivalued image consisting of a white area and a black area.

First, as one of the systematic dither methods, a case of a binary image using a 4×4 Bayer type matrix as a threshold matrix will be described as an example.

FIG. 2 is a diagram showing an example of a matrix for explaining the present invention. (A) is the original halftone image converted to digital data, (B) is a 4×4 pay V-shaped dither threshold matrix, (C) is a black and white binary image (dither image) using a threshold value of 7 trix (B) This is a dithered image (2Inn image) of the converted original image (A). here,
The Bayer type threshold value matrix has a dither pattern in which dots are scattered and covered, as shown in Figure (b).

Figure 3 shows - of multiple types of openings (unit areas) used in the present invention.
An example is shown in a dimensional drawing. A has a size of 2 rows x 2 columns, and B has a size of 2
C indicates an opening having a size of 4 rows by 2 columns, and D indicates an opening having a size of 4 rows by 4 columns. Here, the black circles shown in each aperture A to D are the centers of movement when moving the dithered image in FIG. 2(c). Incidentally, the frontage shown in Figure 3 was fixed. until the second
In Figure (c), move the di IJ-Greg 1 soil to 1, count the number of self-pixels or the number of black pixels (here, the number of white pixels) in [1], and count the number of self-pixels or the number of black pixels (here, the number of white pixels) of the halftone image 1f[ 11 If you spell 1, an estimated halftone image as shown in Figure 4 (A) to (D) will be obtained.Here, (A) is based on Figure 3, and (B) is based on Figure 3B. 1: J: ru, (C) and (D) are halftone images shown in FIG. 3C, respectively.

A method for obtaining the halftone image shown in FIG. 4(d) will be explained.

Now, as shown in FIG. 5, the frontage defined in (d) is the initial position of the dithered image, and the center of movement is at the second row and second column.

(hereinafter expressed as (2, 2)). in this case,
J in the figure: It is desirable that each pixel contained within the aperture be completely contained. In other words, some pixels are missing (
'J including J: It is desirable to prevent this from happening. Note that the black values are shown with diagonal lines here for ease of viewing. Next, the number of white pixels in the area surrounded by this opening is counted and the value is taken as the 111 constant value of the intermediate tone. If you count the number of white pixels within the frontage in the size shown in the figure, it is 7. Therefore, the estimated value of the first row and first column (1, 1>) of the estimated halftone image is 7.

Next, the aperture is moved by one pixel (in this case, one column), and the number of white pixels within the aperture is counted in the same way as described above, resulting in 7. Perform the same operation for accompanying. Then, when the first row is finished, move the opening by one row to the next row,
The intermediate density estimation operation starts from the lower right of the pixel whose center is <3.2). Then, the aperture is moved to the last column of the last row to obtain the halftone image estimated value, and the halftone image estimation operation is completed, and the halftone image shown in FIG. 4(d) is obtained.

Next, a method for obtaining an estimated halftone image using the aperture B shown in FIG. 4(b) will be described. In this case, it is necessary to align the center of movement with the largest frontage, so opening B
The movement start position of is as shown in FIG. The number of white pixels in this state is 2, and in order to match the area to the second diagram, it is necessary to set the white l1Iii prime number in the aperture to 248, so the number of white pixels is 2X2=/I. In this case, the gain of the aperture B is said to be 2. Similarly, when calculating the gain of each frontage in Fig. 2, △ is 4, C
is 2. If such a calculation is performed every time the aperture B is moved by one pixel, an estimated halftone image shown in FIG. 11(b) can be obtained. If we think about Figures 4(a) and 4(c) in the same way, then the explanation will be omitted.

As described above, a halftone image can be estimated relatively well using the <' method as well. The data in Figure 4 is obtained in this way: H J. Since the halftone image is estimated from the dithered image (FIG. 161(C)) with a small amount of information, the original halftone image cannot be completely restored as shown in FIG. 4(D). However, except when the pixel level of the original halftone image changes rapidly, a halftone image that is very close to the original halftone image can be obtained.Especially when there is no density change within the aperture, the estimated halftone The image values perfectly match the original halftone image values.

By the way, human vision has the highest gradation discrimination ability in low spatial frequency regions (regions with few pixel level changes), and has the highest gradation discrimination ability in high spatial frequency regions (regions with many pixel level changes).
It has a characteristic that it has only a low pixel level gradation discrimination ability. Therefore, if high gradation is expressed using a large aperture in the low spatial frequency region, and an image with high resolution is reproduced using a small aperture in the high spatial frequency region, the halftone image estimate shown in Figure 4 can be obtained. It is possible to estimate a halftone image even better than the above.

Next, the method of the present invention will be explained. This method assumes that the digital binary images are already stored in a storage means such as a memory, and that the digital binary images are
A plurality of types of apertures are set, a digital binary image is subjected to predetermined arithmetic processing, and one optimal one is selected from the plurality of apertures for each pixel. Then, the number of white pixels (or the number of black pixels) within the selected aperture is counted to obtain an estimated value of the halftone image. The predetermined arithmetic processing includes selecting a large aperture in a low spatial frequency region (region with few pixel level changes) and a small aperture 1-1 in a high spatial frequency region (region with many pixel level changes); An algorithm is used.

The basic idea of the present invention is c: 11. As long as no density change is observed within the aperture, a large scanning width is selected. Therefore, the order of aperture selection is shown in FIG. Take the order of D → 0 → B → Δ like this.

(2) Step ■ First, select the most popular one.

(3) Step (2) Obtain an estimated value based on the ratio of white areas to black areas within the selected aperture, and re-multilevel this 'H1 constant value using dithering 1 to RIX corresponding to the size of the aperture.

Here, the case of a binary image shown in FIG. 8 will be explained as an example.

Here, first, D is considered as the frontage (step (1)
)). Then, when the selected width is superimposed on the initial position of FIG. 2(c) (see FIG. 5), it becomes as shown in FIG. 8(a). The number of white pixels within this aperture is counted as 7. Assuming that this number of white pixels, 7, is the average pixel level, each pixel is compensated for by 7 in J: sea urchin shown in (b). When the average pixel level image shown in (b) is binarized using the threshold value matrix shown in (c), a binary image (re-binary image) as shown in (d) is obtained (q).

(4) Step■ Check whether the original multi-level image and the re-multilevel image match.

To explain FIG. 8, the original binary image (A) shown in (A) and the re-binary image (D) will be compared. As is clear from the figure, patterns (a) and (d) do not match. The fact that patterns (a) and (d) are not the same pattern means that
This means that there has been a change in pixel level. Therefore, in this case, the opening is inappropriate. Process (1
), no opening was selected, so proceed to step ■.

(5) Step ■ Select the next opening (Step (2)). Here, opening C is selected next in the order of frontage selection in FIG. Select ITI G in Figure 2 (c)
When superimposed on the initial position H, the result is as shown in (e). If we count the number of white pixels within this frontage, it is 4. Assuming that 8, which is the number of white pixels multiplied by a gain of 2, is the average pixel level, each pixel is filled in with 8 as shown in (v).

The average pixel level image shown in (1) is set to threshold 7.
1 to 2 risks (consisting of the threshold matrix in Figure 2 (b) and the second and third columns, that is, the threshold matrix within the frontage)
When converted into a value, a binary image as shown in (d) is obtained.

Here, when comparing the original binary image (e) and the re-binary image (chi), they are not the same pattern.

In other words, there is a mismatch. The fact that the patterns (E) and (J) are not the same pattern means that there is no change in pixel level. Therefore, in this case as well, the opening C is found to be inappropriate. In step (2), opening C is selected and
Tlj IJ3 step 0).

Do ■.

The next selected opening is B (step (3)).

Then, when the selected aperture B is placed at the initial position in Figure 2 (C), it becomes as shown in (S).The number of white pixels in this aperture is counted as 2.The gain is multiplied by the number of white pixels. Assuming that 4 is the average pixel level, each pixel is padded with 4 as shown in (J).When the average pixel level image shown in (J) is binarized using the threshold matrix shown in (R), (E) A re-binarized image as shown in is obtained.Here, comparing the original binary image (S) and the re-binarized image (O),
Same pattern. That is, both patterns match. (
The fact that patterns (a) and (d) are the same pattern means that there is no change in pixel level within this width. Therefore, in this case, the opening B is appropriate. If they do not match, perform the same operation on the last opening (
Here, we will repeat up to A). That is, the aperture A is ultimately the aperture to be found.

When the frontage B is selected in this way, the number of white pixels within the frontage B is 2 as described above, 1=. Since the gain of the aperture B is 2, the estimated image value to be obtained is 2X2=4. That is, the pixel level shown in (J) directly serves as the image estimate value.

The above operations are performed on the dithered image (binary image) shown in Figure 2 (c).
When this is performed for each pixel, an estimated halftone image as shown in FIG. 9 is obtained. Incidentally, to explain which frontage was used for each halftone image estimation, taking the case of the first row as an example, the middle sttr constant image image 1.1> is 8. (1,2)
is B, (1,3) is A, (1,4> is r3.(1,5>
is [3, (1, 6> is D, and (1, 7> is D).

The estimated halftone image shown in Fig. 9 estimates a halftone image using a large aperture in areas where there are few changes in pixel level, and uses a small aperture in areas where there are many changes in pixel level. Therefore, it is in line with human visual characteristics. Therefore, the estimated halftone image is as shown in Figure 1.
The image is extremely close to the original halftone image shown in b).

The case where a halftone image is estimated from a dithered image has been described above.

(6) Step ■ The image type is determined based on the matched apertures.

As described above, an aperture is selected for each pixel, but in the present invention, based on the size of the aperture, it is determined whether the pixel of interest belongs to the halftone image area or the line/character image area.

The present invention employs an algorithm in which a large aperture is selected in a low spatial frequency region and a small aperture is selected in a high spatial frequency region. Therefore, if a small aperture is selected, this pixel is assumed to belong to the line/character image area, and if a large aperture is selected, this pixel is assumed to belong to the halftone image area. As described above, images can be identified on a pixel-by-pixel basis.

Normally, the size of D in FIG. 7 is about 4×4, so when opening A is selected, it is determined to be a line/character image, and when openings B to O are selected, it is determined to be a gradation plane.

However, this judgment also depends on the resolution at the time of image capture, and the higher the resolution, the larger the size can be used as a criterion for determining whether it is a line or character image. When D is about 8×8, 4
The frontage of ×1 person size becomes the boundary.

(7) Step ■ Image processing is performed based on the size of the frontage determined in the steps up to the above.

The optimum size aperture and image type are determined for each pixel in the sequence up to step (2).1. : Therefore, the IC can perform appropriate image processing depending on the image type.

FIG. 10 is a block diagram illustrating an embodiment of an image data processing circuit that implements the method of the present invention.

Image data is stored in the first select circuit 1 for each line (scanning line) in the reline memory 1-1-'-19. As shown in the figure, since the circuit is capable of determining a frontage of a maximum size of 8×8, a line memory of 9 lines 11 to L9 is prepared.

The reason why the number of lines is 9 instead of 8 is that one of the lines is used as a buffer memory during data processing to enable continuous operation. As these line memories 1 to L9, static RAMs are used when the processing speed is fast, and dynamic RAMs are used when the processing speed is slow.

Of the data stored in line memories L1 to L9, 8
The lines are selected from 1 to 1 by the second selection circuit 2 and input to the halftone estimation circuit 3. At this time, if the input data is a binary image, it is input as is, but if it is a multivalued image (for example, 6.8 bit / pel), the image data is binarized before input. Ru. The halftone estimation circuit 3 outputs the estimated value for each aperture and the aperture determination result. The subsequent selection circuit 4 receives these signals, selects the optimal aperture for each pixel, and generates a halftone estimated value (
(restored pixel data).

The restored pixel data enters the first selection circuit 5, and the aperture determination result from the halftone estimation circuit 3 enters the second selection circuit 6 as aperture information. The first selection circuit 5 has a line memory 1-
Original pixel data and selection signal 1 from 1 to L8 are input,
A selection signal 2 is input to the second selection circuit 6 . Here, when the opening is small, a line drawing is used, and when the opening is large, a gradation drawing is used. If only the line drawing part is passed through and a small opening [1 is selected, Goo 1 to 7 become A, and the image data is output. Note that the numerical values shown in the figure indicate the number of bits of the signal line.

FIG. 11 is a diagram showing a specific example of the configuration of the halftone estimation circuit 3. The binary data output from the second select circuit 2 enters a shift register 11 composed of latches 1-Δ1 to 1-Δ8. The shift register 11 is capable of estimating aperture halftones of a maximum size of 8×8. The circuit shown in the figure shows an aperture determination circuit.

The 4-bit binary data selected by the second select circuit 2 is transferred to a shift register 11 consisting of latches LA+ to LAa, from right to left in the diagram according to timing signals from a timing generation circuit (not shown). Schiff) to be done. In addition, the ○ mark shown on the data line in the figure indicates one piece of image data (
binary data). In the case of ID, the width is 4 rows x 4 columns, so it is only necessary to count the number of white pixels in the shift register 11 each time it is shifted, but if you use this method, It takes time and the circuit becomes complicated. Therefore, the present invention utilizes the property that binary data is shifted from the right side of the figure to the left, and that only the data in one column at the end (in this case, the contents of latch L A 7) is replaced, to increase the number of white pixels. simplified counting.

I will explain in detail. When the data is shifted by one column, new binary data is latched into latch LA3.

The number of white pixels for one column is counted by a counter 12. Furthermore, the data for one column that has overflowed from the shift register 11 due to this shift operation is latched into the latch LA7. This latched number of white pixels for one column is determined by the counter 13.
is counted.

On the other hand, since the latch 14 holds the number of white pixels in the aperture before shifting, the subtracter 15 subtracts the number of white pixels for one column that protrudes from this number of white pixels, and the reduced white pixels are If the adder 16 adds the number of white pixels to compensate for the number of white pixels for one column newly input, the number of white pixels within the aperture after shift 1 can be determined. The determined number of white pixels is newly latched into the latch 14. The output of the latch 14 is multiplied by the gain (x4 in this case) by a multiplier 17, outputted as a halftone image estimated value, and sent to the subsequent selection circuit 4.

Above, we have explained the operation of the constant circuit for the halftone image @11 of the aperture, but the same applies to the half-tone image @11 shown in FIG. The data extraction position is changed, the number of white pixels is counted, and a constant value of the halftone image 111 is output.

Next, we will explain the operation of the pattern comparison circuit between the original 211 WiiIj image and the re-binary image.
8 stores patterns according to the type of frontage,
Receives the white pixel number count and position information as an address, and outputs the pattern (re-binary image) stored at the corresponding address.

In this way, the density pattern (rIi' binary image) output from the density pattern ROMI (3) is output to the judgment circuit 1.
9, it is compared to see if it is the same pattern as the binary image output from the shift register 11. If the pattern is the same, the level is set to, for example, "1", and if different, the level is set to, for example, rt.
The OIT level is output from the determination circuit 19.

Although the opening pattern comparison circuit has been described above, the operation is exactly the same for other openings, except that the number of dots to be compared is different.

FIG. 12 is a block diagram showing another embodiment of the image data processing circuit. Components that are the same as those in FIG. 10 are designated by the same reference numerals. In the figure, digitized data (usually 6.8 bits/pe1) obtained by capturing an image is input and stored in line memories ]-1 to -9. Thereafter, eight of the nine line memories are selected by the select circuit 2'. After selection, the same selection circuit 2' binarizes the image using a single threshold value.

The threshold value at this time is preferably set to a slightly higher density side than the background density of the image (or vice versa in the case of an inverted surface). If taken in this way, the intermediate density part of the image will always be completely black, and in the present invention, since the selection is made from the one with the largest aperture,
A large scanning aperture will be selected for the halftone portion.

Therefore, depending on the size of the scanning aperture, it is determined whether each pixel belongs to a halftone or a line/character image area. By this scanning, for example, the size of the aperture is changed from Δ to G (Δ: 2X2.B: 4
X2. C; 2X4. D; /IX4. , E; 8
X4°F; 4X8. G; 8×8), for example, D to G are selected for a halftone image.

At this time, the open [1 determination result from the halftone estimation circuit 3 enters the threshold address selection circuit 21, and an address is output from the threshold address selection circuit 21 to select the optimal threshold data for binarizing the halftone image. and enters the threshold pattern ROM 22. At this time, a halftone image threshold matrix is selected from the threshold pattern ROM 22 and stored in line memory in the binarization circuit 23! A comparison is made with the pixel data (original binary image) in ~L3 or the restored image. Then, the binarization circuit 2
3 outputs binary data. It goes without saying that when opening A-C is selected, the threshold matrix for lines and character images is selected.

FIG. 13 is a diagram showing an example of the threshold value 71-risk used here. Both (a) and (b) are threshold matrices for halftone images, where (a) is a distributed type and (b) is a concentrated type. Both (c) and (d) are threshold matrices for lines and character strokes.

Note that not only the method using such a systematic dither method but also a density pattern may be used. Furthermore, a binarization method such as the minimum average error method may be applied to the halftone image.

FIG. 14 is a processing circuit showing another embodiment of the image data processing circuit. The original image data or restored image data output from the selection circuit 5 is commonly input to the line drawing encoder 31 and the gradation image encoder 32.

Which of these encoders 31 and 32 is selected is determined by the code selection circuit 33 that receives the frontage determination result. Here, the encoding method for line drawings is the usual R, L, M
, H, M, and R methods are applicable. In addition, as encoding methods for gradation images, pit interleaving method, interlaced inversion method,
Differential block method.

There are adaptive predictive coding methods, etc., and it is sufficient to select an appropriate method.

In this way, r, t +:j outputted from either one of the line drawing cylinder pJ converter 31 and the gradation drawing encoder 32
The conversion signals are output from groups 1 to 24.

FIG. 15 shows that both (9 (processing section) I!/
FIG. In the figure, /10 is a scanner (image input device), and the original fn unit /11. light source 42
, optical system 43. CCD44 and Δ/D conversion.

It is composed of a processing circuit 45 that performs shading correction and the like. 51 is a first image processing unit that performs enlargement/reduction rotation, etc., 52 is a second image processing unit that performs aperture determination, and 53
54 is a memory for storing image data, etc., 54 is a printer for outputting images, 55 is a floppy disk, 56 is a magnetic tape, and DB is a bus for data transfer.

In the above description, the case of processing a binary image was mainly used as an example, but the present invention 1.1 is not limited to this, and the processing of a ternary image /Il+C1? It can be applied in exactly the same way to the multivalued image of '7. In addition, as an application of the present invention, 1) Selection of threshold value or processing method in (-) threshold processing of 21 inch/3.1 white 1 h of halftone image; Adaptive encoding method that uses different encoding methods for lines and character images ■A method of automatically extracting or trimming halftone images or parts of line and character images can be considered. Furthermore, by combining it with the OCR function, a very wide range of applications is expected.

(Effects of the Invention) As described above in detail, according to the present invention, the type of image is identified based on the size of the frontage, so appropriate image processing is performed according to the type of image. This has great practical effects.

[Brief explanation of drawings]

FIG. 1 is a flowchart 1-- showing an embodiment of the method of the present invention.
1. FIG. 2 is a diagram showing dither image creation, FIG. 3 is a diagram showing types of frontages, and FIG. 4 is a diagram showing examples of halftone images for each frontage.
Figures 5 and 6 are diagrams showing the initial position of the frontage, Figure 7 is a diagram showing the order of frontage selection, Figure 1 23-8 is an explanatory diagram of opening determination, and Figure 9 is estimation based on R appropriate opening. 1 shows an example of a halftone image, FIG. 10 is a block diagram showing an embodiment of the image data processing circuit, FIG. 11 is a diagram showing a specific example of the structure of the halftone circuit, FIGS. FIG. 11 is a block diagram showing another embodiment of the image data processing circuit, FIG. 13 is a diagram showing an example of a dark value matrix, and FIG. 15 is a diagram showing an example of the structure of the image processing apparatus. 1.2.2'... Redirect circuit 3... Halftone estimation circuit /1~G... Selection circuit 7
．． 34...Gate 11...Shift register 1
2.13 River counter 14... Latch 15... Subtractor 16... Adder 17... Multiplier 18... Density pattern ROM 19... Judgment circuit 21... Threshold address selection circuit 22 . . . Threshold pattern ROM 23 . . . Binarization circuit 31 . . . Line drawing encoder 32 .
52... Image processing circuit 53... Memory 54... Printer 55
... Floppy disk 56 ... Magnetic tape 11-L9 ... Line memory L A l - L A s ... Latch patent applicant Roku Konishi Photo Industry Co., Ltd. Representative Patent attorney Fuji Ijima Jigai 1 Figure 2 (A) (B) (C) (
A) (8) CC) (D) Figure 4 (a) (mouth)
Sekiguchi A Kuniguchi B (c) (d) Opening C
Dark mouth D

Claims (1)

[Claims] Multiple types of apertures are set and moved at the same setting position in a multivalued image consisting of a white area and a black area, and based on the multivalued image for each aperture and the ratio of the white area to the black area for each aperture. A multi-valued image created from the estimated values obtained is compared, one aperture is selected for each set position, the type of image is identified based on the size of the aperture, and image processing is performed. An image processing method characterized by: