JPS62118482A - Picture processing method - Google Patents

Abstract

PURPOSE:To obtain an excellent reproducing picture from a multiplexing picture by restoring an M-coded halftone picture from an N coded picture and applying a prescribed processing to the restored halftone picture to obtain an N'-coded picture. CONSTITUTION:From the N-coded picture data, the M-coded picture data (M>N) is restored, and this M-coded data is made the N'-coded (N>M) data. Namely, an original picture is separated at every block of 2X2 (indicated by a heavy line frame) the number of black picture elements in the respective blocks is counted to obtain density data as shown. Thereby, this density data is obtained by restoring from a binarization picture of 5 levels (5 coded) from 0 to 4.

Description

DETAILED DESCRIPTION OF THE INVENTION (Industrial Application Field) The present invention relates to an image processing method, and more specifically, an image processing method for performing image processing on a restored halftone image. Regarding.

(Prior Art) In the field of image processing, an image input device (referred to as Suki I/Suki) reads a document image and binarizes the read image data. In recent years, with the advancement and development of image processing technology, when recording or displaying images, it is becoming increasingly important to use not only conventional binary image data but also ternary, quaternary, etc. image data. is also becoming possible. For example, three-value or four-value recording in thermal transfer recording,
There are multi-value recording in inkjet recording, multi-value modulation recording using an LD (semiconductor laser) in laser recording, and the like. In these multilevel recordings, in the case of thermal transfer recording, there are methods such as the 3-L method, the method of using a multilevel ink sheet, or the method of changing the pulse width or number of pulses, and in the case of inkjet recording, there are methods such as changing the dot diameter. be.

(Problems to be solved by the invention) Conventional image processing of binary or ternary images (enlargement/reduction, 1l
When performing Pi tone processing (1 image enhancement, etc.), it has been difficult to obtain sufficient processing results because the number of levels (2 or 3) per pixel of these images is small. This is because the procedure is usually to process an image with 6 to 8 bits per pixel, and then to convert the image into 2 or 3° quaternary values.

Therefore, if you check the processing results and perform image processing again if the image is insufficient, this can often happen especially in printing, but it is necessary to memorize the original image. There is. Now, I am uploading a Δ4 size image for 20 dots/
If the solution for mm is 1 x 1 σ, the memory capacity required to store one screen is 210 x 297 x 16 x 16 x 8 = 16 (M bytes), with 8 bits per pixel (256 gradations). Here 210 x 297 is paper nice, 16
×16 is a resolution corresponding to 20 dots/mm, and 8 is the number of bits. Therefore, it is impossible to store the data on a single floppy disk, which is commonly used, and the data must be stored on several disks. There are 4q of colanders. Therefore, magnetic tape or magnetic disks have to be used, but the former has the disadvantage of slow access time, and the latter, although easy to use, cannot store many screens and is not user-friendly. Furthermore, both are expensive if the drive device is included. On the other hand, recently, the capacity of dynamic memory has increased and the price has become cheaper, but this capacity requires about 488 bit memories for 256, which takes up board space and is expensive even though it is cheap. , which is not practical.

The present invention has been made in view of these points, and
The purpose is to obtain excellent reproduced images from multivalued (binary to 8valued) images, especially binary and ternary images, and image processing that can reduce the memory capacity for storing original images. The method lies in cutting the ring.

(Means for Solving the Problems) The present invention, which solves the above problems, restores an N-valued image J: an M-valued halftone image, and performs predetermined processing on the restored halftone image. This method is characterized in that an N'-valued image is obtained.

(Function) The present invention provides a method capable of performing various types of image processing while storing an image in the form of a binary (3-, 4-value) image. Tsu J: From the N-valued image, M (M
≧N (preferably MC·N) A method is provided for restoring a digitized image and performing image processing on the restored image. Therefore, even if an image memory for two images including the work area is required, the total capacity is 210 x 297 x 16 x 16 x 2 = 4 (M bytes) for binarized images, which is 1/4 of the conventional memory capacity.

Therefore, even if dynamic memory is used, the price can be simply reduced to 1/4.

(Example) Hereinafter, an example of the present invention will be described in detail with reference to the drawings. Here, a method for restoring an image will be explained first, and then a method for processing the restored image and a method for converting it into an N' value will be explained. In this example, in order to achieve the above purpose, N
The M value image data (M>N) is restored to IFF double image data, and this M value data is converted into N' value (N'<M) data. Here, generally M>N, N', and N and N
' is normally N']·N due to the above-mentioned restrictions on recording production, but in general they may be equal or may be different in size (that is, there is no restriction on the size of N and N'). Also, normally N=2.3.4 values, but M>N.

If N', there is no numerical restriction.

(Δ) Image restoration method (a) Restoration using the fiI degree pattern method Now, let us consider restoring an M-valued halftone image from the binarized image (original image) shown in FIG. It is assumed that the original image shown in FIG. 1 was created using the density patterns shown in FIGS. 2(a) and 2(b). The original picture shown in Figure 1 is divided into 2 x 2 blocks (indicated by thick circles) as shown in the figure, and the number of black pixels in each block is calculated as shown in Figure 2 (B). Then, density data as shown in FIG. 3 is obtained.

This 11 degree data is used as the estimated value of the original image. The image data shown in Figure 3 consists of 5 levels (5 levels) from O to 4, and 2
This means that it has been restored from the digitized image.

(b) Restoration from dithered image FIG. 4 is a flowchart 1- showing another embodiment of the method of the present invention. This flow will be explained below from +1-1. For a binary image, a scanning aperture of the same size as the dither matrix is prepared (state (3)).Here, as a binary image, for example, as shown in FIG. The one shown in (a) is used, and the scanning amount [1 (, 1, for example, L as shown in FIG. 5 (b)]) is used.Then, the scanning aperture is shifted one pixel at a time. The number of black pixels in each scanner is counted (step 1).The detailed flowchart of step 1 is shown in FIG. 6. Here,
The relationship between the lines and columns of the binarized image is as shown in FIG. 5(a). FIG. 7 is a diagram 1J showing how the scanning aperture moves across the binarized image.

The large frame of the figure is the scanner 1:1, and the numbers on the black pixel number side within the 1° scanning aperture are sequentially calculated as the representative density values (811
degree level) (step ■). FIG. 8 is a diagram showing an example of the original halftone image obtained in this way. The restored image in the figure is a 1elta image restored to 16 values from level 1 to level 16.

A 4×4 scanning aperture is used here for the following reasons. If a size of 4×4 or more is used, the blur of the image will be significant, and if a size of 4×4 or less, for example, 2×2, the resolution is good, but there will be a problem that many gradation levels cannot be obtained. Therefore, the optimum size of the scanning aperture is 4×4.

However, if the resolution of recording or display is 10 dots/mm or more, the blur becomes less noticeable as the recording density increases, and it becomes possible to use a scanning aperture of a larger size.

(C) Restoration using scanning aperture variation method Here, a method of restoring a halftone image from a binarized or ternary image using a dither method will be described. The method of the present invention is characterized in that the scanning amount is changed depending on the image. FIG. 9 is a flow chart showing one embodiment of the method of the present invention.

First, a plurality of dither matrices of different sizes are prepared in advance (step 2). FIG. 10 is a diagram showing examples of dither matrices of various sizes like this. In the figure, G is the maximum size di 1f71~8x8 size.
Hereafter, the discrimination order is F (4x8) → E (8
x4)→r)(4x4)→C(2X4)→B(/I×2
)) △(2X2). Taking the lifting platforms of 71 to Rix G as an example, the 8×8 matrix has three types: Bayer type (A), halftone dot type (B), and high-resolution halftone type (H), as shown in Figure 11. ), but here we used the pay V type shown in (a). In addition, when using a high-resolution dot type threshold matrix as shown in FIG. 11(C), the size type is as shown in FIG. 12 in the discrimination order G-F→E.
-+D IJ length is used.

Next, the largest dither matrix and the dither image are compared to obtain an m-degree pattern (step 2).

First, the reason why matrices of large size are discriminated is based on the fact that there is a high possibility that a matrix of large size will be selected. FIG. 13(a) shows the operation of step (2). To be more specific, in the maximum width G, the number of white pixels of the dithered image within the width is calculated by [-shi], fill in the same number of white pixels as the dithered image in descending order of threshold value, and fill in the remaining black pixels to create a density pattern. Figure 13 (a) shows the density pattern created in this way.Next, it is determined whether the density pattern created in this way matches the pattern of the dithered image (step ). It is clear from Figure 13 (a) that they do not match. If there is a discrepancy, compare the dithered image with the dither matrix F of the next largest size as shown in Figure 13 (b). to obtain a density pattern (step 2). Once the density pattern is obtained, step 2 is performed again to determine whether the density pattern matches the pattern of the dithered image. This time, as shown in Figure J: Both patterns match.

When the two patterns match, the halftone estimated value is determined by multiplying the number of white pixels in the frontage by a gain (step 2). The number of white pixels within the frontage F is 18.

When calculating the gain, there is a relationship of FX2=G, so the gain is -
It is 2. Therefore, the number of white pixels is 18 and the gain is 2.
6 becomes halftone 11 (fixed value). In the example shown in FIG. 13, the dithered image and the density pattern matched when width F was used. Even if the image is progressive, the image and the density pattern may not match (step ■
). In such a case, the number of white pixels within the width of the minimum dither matrix plus the gain G-J is used as the halftone J.
Set fl to a constant value (step ■). The method of the present invention is advantageous in that the calculation for determining the frontage is simple. Furthermore, instead of counting the number of white pixels, the number of black pixels may be counted.

As described above, the binary or ternary image J: halftone image can be restored easily. Therefore, 1 pit 1~(binary)/D
ot image J: Su4pil-(161+ri) /pe1
Both images can be obtained, and it becomes possible to use a quaternary recording device instead of a dual-image recording device. Therefore, by making use of conventional image data such as binary data with a small gradation level per pixel, multiple image recording devices (thermal transfer recording devices, inkjet recording devices, semiconductor laser recording devices) that have begun to appear in recent years have been developed. ) can record this image.

<d) Processing for character images The present invention can also be applied to binary images of line drawings as shown in FIG. 14(a). Now, as shown in Figure 14 (a),
Let us consider restoring a halftone image from a binarized image using the scanning aperture method. Therefore, when a halftone image is restored using a scanning width of 4×4 as shown in FIG. 14(b), a halftone image of tj; is restored as shown in FIG. 15.

The restored image shown in FIG. 15 obtained in this way is 4
Due to scanning with a scanning width of x4, there is a slight blurring effect at the edges. Therefore, when the image shown in FIG. 15 is binarized, an image with fewer irregularities than the original image shown in FIG. 14(a) should be obtained. For example, for the restored image shown in FIG. 15, the threshold value 7 shown in FIG.
When binarization is performed using 1 to 6 riks, a re-binarized image as shown in FIG. 16 (b) is obtained. It can be seen that the unevenness is reduced compared to the original binarized image shown in FIG. 14(a).

Various image restoration processes have been explained above in (a) to (rl):l. According to the present invention, it is possible to restore an image M having a greater number of gradations M from a binary or rL/I value image, regardless of whether it is a line drawing (character drawing) or a halftone image. Also, this restoration method la
The method is not limited to the IJ method shown in this example.

(e) In the explanation of the image processing method up to the restoration processing platform, which is based on analog method, the case where digital processing is performed using digital image data was explained as an example, but similar image processing can be performed using analog method. It can also be realized by using

FIGS. 17 and 18 are diagrams showing an image processing method using an analog method. In FIG. 17, optical information of a binary image (original image) 1 is collected by a lens 2 and irradiated onto a two-dimensional image sensor 3 to obtain an image signal photoelectrically converted for each element. In this case, each of the N pixels of the original image 1 is made to correspond to one pixel of the two-dimensional image sensor 3, so that the average value of the N pixels is projected onto the image sensor 3. Here, it is assumed that the original image 1 is represented by nxn (=N) pixels, and the magnification of the lens 2 must be set to 1/n. That is, data of one pixel of the image sensor corresponds to data of one pixel of the M-ary image.

On the other hand, the embodiment shown in FIG. 18 uses a linear image sensor 4 using photoelectric conversion elements of Nxl pixels for one line as an image sensor.

In this embodiment, only one line of the image can be read, so it is necessary to move the original image 1 one line at a time in the direction of the arrow in the figure each time one line of the image is finished reading. After the binarized image is restored to a halftone image in this way, various image processing can be performed on the restored halftone image.

FIG. 19 is a dimensional diagram showing various examples of image processing.

In the embodiment shown in (A), the analog image signal is sent to the amplifier 11.
After amplifying the data, the A/D converter 12 converts the data into digital data, and the image processing circuit 13 performs seed Z image processing, which is then output as image data. Examples of image processing include density conversion and binarization processing. (a) In the example, Ana [1gu drawing (Na!
4r; After converting 'l into digital data with △/D converter 1β unit 12, gradation processing is performed with ROMI/I, converting to binarized data with binarization circuit 15'r'', and output as image data. Here, the ROM 14 stores, for example, data on nonlinear input/output conversion characteristics, and examples of the binarization processing method of the binarization circuit 15 include the 1fii degree pattern method and the dither matrix method.

In the embodiment shown in (C), an analog image signal is converted into digital data by an A/D converter 12, then filtered by a subsequent filtering circuit 1G, and the filtered data is binarized by a binarization circuit 15. Convert to data and output as image data. Actual door 1 shown in <2)
In the example, an analog image signal is converted into digital data by the A/D converter 12. The converted digital data is temporarily stored in the storage device 17. Enlarge/reduce circuit 1
Reference numeral 8 performs enlargement processing or reduction processing on the output of the A/D converter 12 as it is, or on the image data stored in the storage device 17. The image data processed by the enlargement/reduction circuit 18 is converted into binary data by the subsequent binarization circuit 15, and output as image data. In the above description, the case where binary conversion processing is performed is taken as an example, but multi-level conversion processing such as ternary or quaternary processing may also be performed.

In the embodiments shown in FIGS. 17 and 18, a color filter is attached between the lens 2 and the image sensors 3 and 4, and the color binary image is separated by the color filter, and each color is separated into a binary image. It is also possible to perform restoration processing on the data. As a color filter, for example, in the case of the embodiment shown in FIG. 18, it is possible to use an optical system as shown in FIG. 20. In the same figure, (a) shows the image 22 of the photographic lens 21 separated into three colors using a plurality of relay lenses 23 to 26 and the gicroic mirrors 27 and 27', and then redistributed on the C0Ds 28 to 30 respectively. It is configured to form an image.

In the example shown in (b), the photographing lens 21 and each CCD 28 to 3 are
0 and a plurality of prisms 31 with special shapes
34, and guichroic mirrors 27° and 27' are arranged between prisms 31 and 32 and between prisms 33 and 34 to separate the colors into three colors.

(c) shows three prisms with acute apex angles 35, 36.36
' are fitted together to form a triangle ABC as shown in the figure, and gicroic mirrors 37 and 38 are formed on the boundary surfaces of each prism to perform three-color separation. The example shown in (d) has a configuration in which the prism of the example shown in (c) is exactly turned over. tn of each prism! 1
Gychroic mirrors 39 and 40 are formed in each of the mirrors 9 and 9, respectively.

(B) Image processing method In (A), the method of restoring an M-valued halftone image from an N-valued image was mainly explained. Here, an image processing method and an N'-value processing method for processing the restored M-value halftone image will be described.

(a> Gradation processing The restored halftone image (In image) shown in FIG. 8 is subjected to tone conversion based on the halftone conversion characteristic curve as shown in FIG. 21. is the input, and the vertical axis is the output level.Corresponding to the fact that the halftone image shown in Figure 8 has 16 levels of gradation, both the input and output levels are 1.
There are 6 gradations. fA, tB, and fc in the figure are tone conversion characteristic curves, respectively. There is no concentration conversion in the curve 10, and the fA convex curve is for high concentration conversion, and fB is for low concentration conversion.

For example, in the case of gradation conversion using the A curve, the input level ``7'' becomes ``11'' as is clear from the figure.
converted to level. That is, it can be seen that the fA convex curve is converted to a higher density. Conversely, in the f curve, the density is converted to be lower. Here, when the halftone image shown in FIG. 8 is tone-converted using the fA convex curve, a halftone image as shown in FIG. 22 is obtained. Comparing the halftone image shown in FIG. 22 with that shown in FIG. 8, it can be seen that the level of scratches is higher.

Next, a case where halftone image restoration and gradation processing are performed simultaneously from the ternary image shown in FIG. 23 will be described. Furthermore, the second
Figure 3 is a ternarized image in which one part of the mesh is at a black level and the shaded part is at a gray level. Now, from this ternary image, Figure 2/I (a)
, (b) Let us consider reproducing a halftone image fil using a 4×4 density pattern as shown in (b). The numbers in Figure 24 (b) indicate m-degree bells, and the small black circles indicate gray dots.
Large black circles indicate black dots. If you restore the halftone image 1 ((() by using the density pattern of Figure 2/I in Figure 23, the second
A halftone image as shown in Figure 5 (a) is obtained.

If the restoration level at this time is reduced by 1, it is expressed as L = [number of gray dots x W + number of black dots XWe]
1, We-2. From the above equation, it can be seen that by changing the weights of W and WB, gradation processing becomes possible at the same time as restoration. Transform equation (1) into the following J: sea urchin.

From the formula (2), L=(W/WB)Xnumber of gray dots+number of black dots.If W/WB is made larger than 1/2, the density is restored to high 11, and if it is made smaller, the density is restored to low.

For example, if W /W day day - 34, Figure 25 (b)
It is restored as shown in . It can be seen that compared to case (a), there is a shift to the higher concentration side.

FIG. 26 [J] is a diagram showing this gradation conversion characteristic, in which the horizontal axis is the gradation level of the original image, and the vertical axis is the gradation level on the gradation side. fA
is W /VANs = 1/2 fB is W /Wa
= 3/4 respectively.

It can be seen that fB is converted to a higher concentration than fA. It goes without saying that this example can be applied not only to density patterns but also to restoration from ternary dithered images.

(b) Enlarging/reducing processing Next, a method of enlarging/reducing the restored halftone image will be explained. In the case of enlargement/reduction, the screen size is constant but the number of pixel data varies, so it is necessary to vary the pixel interval. Figure 27 is a diagram showing pixel positions during enlargement/reduction, (A) is when enlarged to 4/3 times, (B) is reduced to 3/4 times.
Each shows pixel tearing. In either case, the x mark indicates the pixel position when the magnification is the same, and the ○ mark indicates the pixel position when the magnification is changed. When enlarging, the pixel spacing becomes narrower than when it is at full magnification.
At the time of reduction, the pixel interval becomes wider than when it is equal to 18.

Here, there are several methods of enlarging/reducing, but the main ones are (a) a method in which the data of the pixel at the same magnification that is closest to the pixel position of the magnification is used as the pixel data of the magnification (nearest neighbor method, etc.) (0) A method of calculating logic or numerical calculations from the same-magnification pixel data surrounding a variable-magnification pixel (9 divisions).

logical sum, projection, reinforcement method, etc.) It is divided into two parts.

In (b), if the pixel density of the original image is high, there is not much of a problem, but when it is very low, such as around 10 dots, the unevenness especially in the black area becomes noticeable when enlarged. On the other hand, the processing is very simple. (B) has less image deterioration than (B), but has the disadvantage that processing tends to become complex. FIG. 28 is a J-diagram showing an example of an image obtained by enlarging the restored image shown in FIG. 8 to 4/3 times. FIG. 29 is a diagram showing an example of an image obtained by enlarging the restored image of the line drawing shown in FIG. 15 by 4/3 times. The above description has been given using an example of image enlargement, but image reduction can also be performed in the same manner.

As described above, according to the present invention, enlargement/reduction is possible, but the method of enlargement/reduction is not limited to this example. In the present invention, once the image is restored to a halftone image, it is enlarged and
In order to perform reduction, it is possible to solve the moiré problem that conventionally appears when scaling a 2 (or 3.4) value halftone image.

(C) Image enhancement processing Now, image enhancement is performed on the restored image shown in FIG.
I'll try to fix the blur. As image enhancement methods, there are several methods such as digital filters. Here, edge enhancement will be explained. The digital filter for edge emphasis is [1,8] when N=2, and [1,a when N=3.

1] is a one-dimensional filter. Other possible one-dimensional filters include filters such as N-5 and N-7.

Note that a is a constant that determines the spatial frequency characteristics, and depending on the application, a = 1.2 Low bass fill a = Q Fill to band cut 1 a − -1
, -2 bypass filter. Here, a constant for the bypass filter is used for edge enhancement.

As a two-dimensional filter, (31, not shown in Figure 30) is used.Of these, (A) is a bypass filter in the X-axis direction, (B) is a bypass filter in the X- and Y-axis directions, and (C) is a bypass filter in the X- and Y-axis directions. , (d) G31 with my correction filter, (d)
The constant a is preferably a≦−12. Here, when the filter shown in FIG. 31 is applied to the restored image shown in FIG. 15, the filtered image (also shown in FIG. 32:
The outside of Figure 5 is set to O, and negative values that appear in the calculation results are also set to O.

In this way, when image enhancement processing is applied to a line drawing, edges are particularly emphasized. On the other hand, as the digital filter, not only the one used in this example but also various types of filters can be used. Furthermore, although the image restoration method shown in (d) of (A) has a problem in that the blur is easily noticeable, it is possible to obtain a good image by improving it by the image enhancement processing described here.

Note that when performing image enhancement as described above, it is also possible to perform image enhancement simultaneously with image restoration.

In this case, the image shown in FIG. 15 may be subjected to filtering processing as shown in FIG. 33.

(d>Multi-level conversion processing The restored image shown in Fig. 32 is re-doubled using a threshold matrix.
It can be converted into value. For example, when the image is binarized using the threshold matrix shown in FIG. 34(a), a binarized image shown in FIG. 34(b) is obtained.

Incidentally, when the original image shown in FIG. 15 is binarized using the threshold matrix shown in FIG. 35(a), a binarized image as shown in FIG. 35(b) is obtained. In this way, the re-binarized image has the characteristic that the unevenness in the shaded area is less noticeable compared to the original binarized image shown in FIG. 14(a). The Δ portion in FIG. 3/I (b) is a portion with poor reproducibility due to the adverse effect of setting the outside of FIG. 15 to O during filtering during image enhancement.

FIG. 36 shows a binarized image obtained by binarizing the single-line and both halftone images shown in FIG. 29 using the threshold matrix shown in FIG. 35(a). Generally, when re-binarizing a binarized image, if there are irregularities in the black area of the original binarized image, the enlargement will be
・This unevenness is usually enlarged and reproduced in 1. In the case of the method of the present invention, as can be seen by comparing FIG. 14(a) and FIG. 36, the unevenness is reduced and enlarged.

This also makes clear the advantages of the method of the present invention.

Next, the case of ternarization will be explained. The image shown in FIG. 32 is ternarized using the threshold values shown in FIGS. 37(a) and 37(b). (a) uses white-gray threshold matrices, and (b) uses gray-black threshold matrices.

By using these threshold values, it is possible to obtain ternarized data of i11 degree level: (1) level from 0 to 6, (2) level from 6 to 10, and (2) level from 10 to 16. FIG. 37(c) shows the ternarized image obtained in this manner. In the figure, the shaded area is the gray level, the shaded area is the black level, and the other areas are the white level. The ternarized image shown in FIG. 38(c) is
This is a ternarized image in which the original image shown in the figure is ternarized using the threshold matrix shown in FIGS. 38(a) and 38(b). 37th
Comparing Figure 38(C) with Figure 38(C), even though Figure 37(C) has a higher gray-black threshold and the gray level is more likely to appear, it is better than Figure 38(C). The ash level is also low. This is because the image in FIG. 32 is emphasized.

Next, we will describe how the gradation-processed image shown in FIG. 22 is binarized using the systematic dither method. The halftone image shown in FIG. 22 (obtained from the restored image in FIG. 8 from the binary image in FIG. 5 as described above) is converted into a 4×4 threshold matrix shown in FIG. 39 (a). When binarized using
A binarized image as shown in Figure (b) is obtained. Figure 5 (
It can be seen that the image 1b shown in FIG. 39 (1), which was restored to halftone and then subjected to gradation processing, has a higher density than the binarized image shown in b). Note that it is of course possible to perform not only two-value conversion but also multi-value conversion such as three-value conversion and four-value conversion.

FIG. 22 shows the gradation-processed image shown in FIG. 40 (a).

When ternarized using the density pattern shown in (b), the fourth
It is possible to obtain a 3-valued image shown in Figure 1. Figure 40 (a) is the white-gray density pattern, Figure 40 (b)
shows a gray-black density pattern.

(e) Pattern change processing According to the present invention, it is possible to change the halftone expression method of an original binary image or a multivalued image. Here, an example will be explained in which a binarized image is used as the original image. As a method of expressing a halftone image (re-multilevel conversion), the following J: method can be considered.

■ N1iCt image (original image) by systematic digital method -M
l-valued image→N'-valued image by random D1f method Here, as the random D+7' method, for example, the average error method or the least squares method can be considered.

■ N-valued image (original image) by systematic dithering → M-valued image → N'-valued image by reciprocal dithering In this case, if the original image is Bayer dithering, For the converted image, a method such as a halftone pattern method can be considered.

■ N-valued image (original image) by systematic dithering → M-valued image N'-valued image by density pattern method The above three cases were explained as examples, but if restoration is possible, the reverse process can be used. It also holds true. As an example, consider a restored image as shown in FIG.

This restored image is restored from image data that has been binarized using a 4×4 Bayer dither matrix shown in FIG. The restored image shown in FIG.
The 4×4 halftone matrix shown in Figure (a) or the 43rd
When re-binarization is performed using the 2×2 pay V-shaped dither matrix shown in FIG. 44 (b) with different matrix sizes shown in the figure, the binarization shown in FIGS. 45 (a) and (b), respectively. An image is obtained. According to the present invention, the multi-value quantization method and multi-value quantization pattern can be changed depending on the application. Furthermore, the halftone expression method of the present invention is not limited to the methods (1) to (4) described above.

(f) Processing of ternary dithered image The ternary dithered FIN image as shown in Fig. 46 is moved pixel by pixel using a 2x2 scanning aperture, and the number of black pixels and device pixels within the scanning aperture are counted. A multivalued image is restored by changing the weight of −2 and elementary element = 1.

As a result, the J:Una multivalued image shown in FIG. 47 is restored. For the multivalued image obtained in this way,
When binarization is performed using the threshold matrix η゛J: shown in Figure (a), a binarized image as shown in Figure 48 (b) is obtained. The a-valued image may be reduced by 1q using the threshold for ternarization.Also, the threshold h
Even if n71-Rikus uses the density pattern method instead of Didi, J: Yes. For example, for the restored original image in Figure 47,
When the density pattern shown in Figure (a) is used to ternarize, that is, the value of one pixel in Figure 47 is expressed by a 2 x 2 ternary matrix, the result is as shown in Figure 49 (B). Na 3
A valued image is obtained.

((+) Processing for color images Various image processing methods for monochrome images have been explained up to now, but the same can be done for color images as well. In other words, for example, if a color image is processed using three primary color signals R,
It is assumed that the colors are separated into G and B. Each primary color signal R, G, B
can be considered in the same way as monochrome images, so all of the image processing methods described up to now and the image processing methods described below can be applied. Figure 50.

FIG. 51 is a flowchart showing an example of application of the method of the present invention to color images. First, FIG. 50 will be explained.

A color original is color-separated into R, G, and 83 primary colors using a color imaging system (step 2). As a color separation method, for example, R
, G, and B filters may be used to separate the color document information into colors. Next, for the color-separated R, G, and B signals,
Binarization processing is performed for each signal (step ■). The method of binarization processing is as described above in 1. : Binarization methods can be considered using various threshold matrices. The lζ binarized image data obtained by the binarization process is stored (step ■). Note that when storing the raw data, the raw data may be stored as is, or the data may be encoded and compressed before being stored.

Compress and store in memory! can save memory.

The stored image data is read out and restored (step 2). 19-dimensional processing methods include dividing the binarized image into blocks of a predetermined size and increasing the number of black pixels in each block to 4, and sequentially shifting the scanning width one pixel at a time.
IJsa1! One possible method is to count the number of black pixels within the scanning amount. Next is the restored Chumon S! ! I 1il
ii image is subjected to image processing such as color correction (no C color correction), Y, M
, Ci! ii Obtain an image (step ■).

As the color correction method, a normal method may be used. Other types of image processing may include, for example, image enhancement, enlarging/reducing one image, and γ conversion. This J: Obtained by /
, :Y, M, C11ji image signals can be input to an image output device such as a printer to reproduce a color image.

The flowchart shown in FIG. 51 is for converting into R, G, and B signals and then converting into hit Y, M, and C signals. The subsequent sequence is shown in FIG. 50 - it is the same, and the explanation will be omitted. As shown in Figures 50 and 51, color images differ from normal monochrome images in that R, G, B
Each image signal is normally separated into colors. Therefore, the amount of data is three times that of monochrome data. Also, when recording
.

If we consider four colors, M, C, and Black, the required data will be four times as large. Therefore, when attempting to store a color image, the amount of data is large, posing a major problem in terms of memory capacity. For example, △4(210X297111+11>
If each size image is captured at 16 dots/n+m, the amount of data will be:

210X297X16X16X3 (R,G,B)=4
7900160 (bits)" = approximately 6 Mbytes is required. Therefore, at the current level, it is difficult to store data, and magnetic disks or magnetic tapes must be used. On the other hand, the restoration method of the present invention is used. Then, after binarization, it is memorized and 6MbN = 748, which requires only 1. Also, 2
For example, if the cultivation data is coded and η is encoded for each l, the capacity will be smaller than this, which is very effective. In terms of recording and display, the image is encoded and displayed as binary data or as a halftone image, or it can be processed by color correction etc.
, M, C or Y, M, C, Black (It is also possible to create a number and record it as a binary or 3.4 value (
Y, M, C or Y, M, C, Bl! Of course, it is also possible to obtain the signal and then store it as shown in J: in Fig. 51).

Note that the image processing method described in the present invention is not limited to the example of the present invention, and other processes or processing methods can be applied. Further, the image restoration method of the present invention is not limited to the example of the present invention, and may also be performed optically (processing).

(Effects of the Invention) As described in detail above, according to the present invention, by performing the N-ary image → M conversion (restoration processing) → image processing → N'1 rff conversion process, the multi-value image 19 excellent reproduction images for data (2 to 8 values), especially binary and ternary image data
It will be done.

More specifically, ■2 (It is possible to perform gradation processing 9 image enhancement on 3, 4> value images, and it is also possible to solve moiré that occurs when changing the magnification of halftone images. Become.

■It is possible to obtain 3-value, 4-value, etc. images rather than 22-value images, so image data that lacks one gradation level can be recorded using a device that has a large number of recording levels.

Display becomes possible.

■When enlarging a binary line drawing, there is a problem that the unevenness of the diagonal lines stands out, but in order to return to a charcoal halftone image and enlarge it, we need to perform multilevel conversion again and reduce the increase in unevenness during enlargement. It is possible to prevent this.

By restoring the 0 image, the same or t
, t It is possible to use a different N' value conversion method/pattern. This makes it effective for applications such as printing. Also, the net pattern and angle change.

Line number change processing becomes possible.

■N value Since it is possible to store, for example, a binary image and perform various image processing on this image, there is no need to store, for example, a 256-value image as in the past.
This is very advantageous in terms of saving storage capacity (lowering the price).

[Brief explanation of drawings]

1 to 3 are explanatory diagrams of the method of the present invention, FIG. 4 is a flowchart showing an embodiment of the method of the present invention, and FIGS. 6 and 9
Figures 50 and 51 are diagrams showing other embodiments of the method of the present invention. FIGS. 7, 8, and 10 to 49 are explanatory diagrams of the method of the present invention. 1... Binarized image 2... Lens 3... Two-dimensional image sensor 1) 4... Linear image Hinley 11... Amplifier 12... Δ/D converter 13
...Image processing circuit 14...ROM 15...Binarization circuit 16...Filtering circuit 17...Storage device 18...Enlargement/reduction circuit Patent applicant Roku Konishi Photo Industry Co., Ltd. Agent
Patent Attorney Fuji Ijima 1 off-site person 2 years of practice (A) (B) Fig. 3 Fig. 4 Start ■ 2 (1141ZM"゛7°"'"0 size scanning aperture prepared Scanning aperture 1 pixel) ) ■ Count the number of black pixels in each aperture while shifting ■ Belt Dictionary Figure 5 (A) -> Column 4x4 Scanning Diarrhea 6 Figure χ B Figure 21 Figure 22 Figure 23 Figure 24 (I) ) (B) Hehehe Heguhe III l l Fig. 33 Rod 32 Fig. 35 boxes (A) (B) 21st Fig. 36 Fig. 37 (A) (B) (C) Diarrhea Fig. 38 (A) ) (B) (C) No. (A) 2i i1 A and A-Bi Nao Fig. 40 (A) (B) White - Gray Gray - Male Fig. 39 (B) ) Hi-Kokui Minoru Fig. 41 Fig. 31 Image enhancement 42 areas Kusara Ryo Tanen (A) Figure 43 Figure 44 (B) Matrix angle rumor (A) 4x4 Halftone dot shape Figure 46 Figure 3 Prejudice image 45 (B) 2x2 de 'Izamatrix Figure 47 Restored surface angle figure 48 (A) (B) Figure 49 (A) (B) Figure 50

Claims (2)

[Claims] (1) An image processing method characterized in that an M-valued halftone image is restored from an N-valued image, and a predetermined process is performed on the restored halftone image to obtain an N'-valued image. (2) The image processing method according to claim 1, wherein the integers N, M, and N' are set to have the following relationships: M≧N, M≧N'.