JPS6349428B2 - - Google Patents

Description

[Detailed description of the invention]

INDUSTRIAL APPLICATION FIELD The present invention is applicable to general image scanning/storage devices, such as facsimile transmission devices, which scan and decompose an image and then reconstruct the image again.
The present invention relates to an image signal processing method and an image signal processing device used in a display device. Configuration of conventional examples and their problems In recent years, the field of use of facsimile has been expanding more and more, and it has become desirable to transmit not only simple character images but also a wide range of images including gradations and periodic patterns. There is. In particular, when drawing lines in a document image have a relatively fine image period, such as a halftone photograph or a plaid pattern on clothing, moiré fringes occur due to interference between the image period and the scanning period of the original image.
This may result in significant deterioration of image quality and limit the scope of use of facsimile, or as a countermeasure, it is necessary to provide expensive equipment that scans images at a high scanning line density or is equipped with multiple scanning line densities. I'm doing it. As a conventional example, the manner in which moiré fringes occur when a halftone photograph is used as an original will be described below, and a method called the dither method as a typical example of binary expression of gradation will be explained. FIG. 1 shows how halftone dots are quantized. The areas of halftone dots 1 to 6 are all the same size, with the inside of the circle being black and the outside of the circle being white. Now, suppose that we obtain numerical values as shown in the figure by quantizing the black area while dividing it into square grids, as shown in the figure. The quantization value is a decimal number, with 100 when all the squares are black and 0 when all the squares are white. (In Figure 1, the value 0 is left blank.) For the quantization numbers in Figure 1, if we represent 50 or less as white and 51 or more as black, then the reproduced network in Figure 2 It looks like a binarized image of points. Referring to FIG. 2, the reproduced black halftone dots vary from one to four pixels. Originally, all the halftone dots in the original image have the same area, so the reproduced black halftone dots should be represented by a fixed number of pixels. A difference in the number of pixels in one halftone dot means that the average density around the halftone dot is different. As shown in FIG. 2, when the number of pixels in one halftone dot changes gradually on a plane and becomes periodic, this becomes visible to the human eye as moiré fringes. To explain the reason, the total number of quantizations around each halftone dot for halftone dots 1 to 6 in Figure 1 is 312, and when all of one pixel is black, the quantization value is Since it is set to 100, one black color when reproduced in Figure 2 also has a value of 100. On the playback side, since it is binary, there is no intermediate value, and everything is represented as 0 or 100. Looking at the example of halftone dot 3 in Figure 1, the quantization number 83 is reproduced as black.
17 becomes extra black, and the part with quantization number 41 is reproduced as white, so 41 becomes too white. Therefore, the former case will result in an error of 17, and the latter case will result in an error of -41. When the error for the entire halftone dot 3 in FIG. 1 is examined in this way, the reproduced black is 300 for 3 pixels, and the error for the original halftone dot is -12. That is, for binary reproduction, a difference within ±1/2 pixel in area with respect to the original halftone dot always occurs. Translated into a quantized value, it is ±50. Since each halftone dot in FIG. 1 has a total quantization number of 312, the reproduced halftone dot with the least error is 300, and it is best to reproduce with three blacks. However, when we compare the reproduced halftone dots in Figure 2 with the halftone dots in Figure 1, halftone dots 3 and 4 are just right, halftone dot 5 has one extra black, and halftone dots 1 and 6 have one black. There is less black, and halftone dot 2 is reproduced with less black than two, and these errors cause moiré fringes. Therefore, in order to make moiré fringes less visually noticeable, it is a good idea to make the area of the reproduced halftone dots constant when the halftone dot area of the original image is constant. In order to suppress the occurrence of moiré fringes, it is preferable to use scanning lines that are considerably finer than the number of lines of the target halftone dots to reduce deformation of the halftone dots. For example, if the size of the grid in Figure 1 is reduced to about 1/10, the area variation of the converted halftone dots will become very small and will no longer be perceived as moiré fringes. However, making the scanning lines finer means that the number of image data increases in proportion to the square of the scanning lines, which increases the processing time of image data, complicates device manufacturing, and reduces the efficiency of using electrical transmission lines. Economic efficiency, increased storage capacity for systems that store all image data, etc.
Many other problems arise. Therefore, although moire fringes do occur, efforts have been made to select an appropriate scanning line density that makes the moire fringes less noticeable. For example, in the example shown in Figure 1, if we consider that the positional relationship between the center of the original halftone dot and the lattice is misaligned, there is a difference in the area of the reproduced halftone dots. By determining the period of the grating at an interval of 1/2, and aligning the array direction of the halftone dots with the direction of the grating, this discrepancy can be eliminated and the areas of the reproduced halftone dots can be made the same. It becomes possible. However, the halftone dots in commonly used photographs vary in both shape and number of lines, and it is difficult to completely cover them with this method. Furthermore, recently, there has been a general desire to reproduce intermediate levels by digitally transmitting grayscale images with intermediate levels, even in the case of reproduction transmission. As a typical example of expressing a grayscale image with an intermediate level by binarizing it,
There is a method called the dither method. Figure 3A
~C is an example of a dither method for expressing intermediate density in binary form. Figure A shows the quantization number of image data in decimal notation, where the whitest level is 0 and the blackest level is 0.
Image data at levels 30 and 60 are given as examples out of image data in a range of 100. B in the figure is a threshold table for binarizing the image signal in A in the figure, and corresponds one-to-one with each pixel of the image data. In the example shown in FIG. 7B, the numerical array in the 4×4 matrix (unit matrix) in frame 7 is repeatedly expanded into two dimensions. The threshold is divided into 17 between 0 and 100.
It is set with 16 values. C in the same figure represents a binarized image, and the image data in A in the figure that is smaller than the threshold in B in the figure is colored white, and if it is larger than or equal to the threshold value in the same figure, it is colored in black. For example, image data 8 and 9 in Figure A are both 30, and are compared and determined using the threshold values 65 and 18 of the corresponding coordinate points in Figure B, respectively, and are represented in white and black, respectively, in Figure C. As the value of the image data becomes larger, the number of objects exceeding the threshold in the unit matrix increases, and therefore the number of black areas increases. In this way, the number of blacks proportional to the level of the image signal is generated for each unit matrix, and an average density is expressed. The dither method is good for expressing images with intermediate density, but it is also useful for expressing binary images (halftone images such as newspaper photographs, text, etc.).
It is unsuitable for scanning line drawings, etc.) to create binary data. FIG. 4 shows the result of binarizing the halftone image data of FIG. 1 using the threshold table of FIG. 3B. This result indicates that the area reproduction of halftone dots is not good, similar to the phenomenon explained above regarding moiré fringes. Furthermore, if a thin line such as a character or line drawing is close to the pixel width, the reproduced line will look like a dotted line, which is easily considered from the relationship between the quantization value of image scanning and the threshold value of the unit matrix. Purpose of the invention Therefore, the purpose of the present invention is to
Second, it is possible to obtain binarized data that does not cause moiré fringes or breaks in line drawings, secondly, it is possible to express grayscale images with intermediate levels using binarized image data, and thirdly, the above-mentioned 2. In other words, an image signal processing method and image signal that can effectively process images that include a mixture of binary images such as text and photographs and images with intermediate shading. The purpose of the present invention is to provide a processing device. Structure of the Invention In order to achieve the above object, the present invention sets a scanning window that is sequentially scanned with respect to the image signal level of each pixel obtained by scanning and decomposing the original image, and the image signal within this scanning window is set. When reproducing a new black pixel from the total level value, the error that occurred when reproducing the black pixel in the scanning window at the previous scanning position is calculated as the sum of the image signal levels within the scanning window at the next scanning position. It is added from time to time. DESCRIPTION OF EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. First, the principle of the present invention will be explained step by step with reference to FIG. Step (1) The image signal string obtained by scanning and decomposing the original image is stored in the image signal storage section G according to the main scanning direction and the sub-scanning direction, and each pixel P _i,j in the image signal storage section G is stored. The image signal level of (i=1 to I, j=1 to J) is defined as L _ij . Step (2) Collection of pixels

Set a scanning window W _i,j surrounding [Formula]. Step (3) Each pixel P _i+u,j+v of scanning window W _i, j (u=0~m,
The sum S _n of image signal levels L _i+u,j+v of V=0 to v),
The sum S with the error correction amount E generated on the scanning line W _i,j-1 is determined. Step (4) For a predetermined image signal level C, when O≦S≦C×(m+1)×(n+1), (a) S=C×N+A However, O≦N≦(m+1) ×(n+1) Find N and A such that O≦A<C, and (b) calculate the image signal level L i+u,j+j of each pixel P i+ _{u ,j+v} within the scanning window _{W i,j.} K descending or ascending values
(P _i+u,j+v ), (c) For the image signal level L _{i+u,j+v of each pixel P i+u,j+v} within the scanning window W _{i, j} , Using descending or ascending values K(P _i+u,j+v ), for K(P _i+u,j+v )≦N, L _i+u,j+v = C K(P _{i+u , j+v} )=N+1, perform the first permutation such that L _i+u,j+v =A K(P _i+u,j+v )>N+1, L _i+u,j+v =0, (d) the image signal level P _1ST (=L i,j ) of the first replacement of the pixel P _{i,j that} will not be included in the scanning window again due to the subsequent movement of the scanning window W _i,j ;
Predetermined binarization level V where O≦V≦C
When P _1ST (=L _i,j )>V, L _i,j = C, E=L _i,j −C P _1ST (=L _i,j )≦V, L _i,j =O, E=L _i,j is applied to the image signal level L _i,j, the result is set as P _2ND , and the error correction amount E is determined. When S<O, all the values of each image signal level L _i+u,j+v in the scanning window W _i ,j are replaced with the image signal level O, and the error correction amount E is saved as S. . When S≧C×(m+1)×(n+1), all the values of each image signal level L _i+u,j+v in the scanning window W _i, j are replaced with the image signal level C, and the error is The correction amount E is S−C×(m+1)×(n
+1). Step (5) In the main scanning direction, change i from 1 to (I-
m) and repeat steps (2) to (4). Step (6) In the sub-scanning direction, change j from 1 to (J-
Repeat steps (2) to (5) with changes up to n). In the above description, the case where the scanning window W _i,j is rectangular has been described, but this is also possible for any shape such as a circle, an ellipse, or a triangle. At this time, the operations performed in steps (4)-(d) are performed on pixels that are no longer included within the scanning window after processing in that scanning window and are no longer subjected to data redistribution conversion. Further, the predetermined value of the image signal level C may be the maximum image signal level or a value in the vicinity thereof. Furthermore, in the above explanation, all the image signal sequences are stored in one image signal storage section G, but it is also possible to store only the image signal sequences necessary for the scanning window W _i,j and replace them one after another as scanning is performed. be. Next, the present invention will be explained with examples using specific numerical values. Figure 6A is the same halftone dot as halftone dot 1 in Figure 1,
It is assumed that the range shown in FIG. 6A is the entire original image. Adding all the quantized numbers in FIG. 6A is 312, which is equivalent to three 100s and one 12. By distributing these three 100s and one 12 in descending order of the quantization number of the original image, they can be arranged as shown in FIG. 6B or FIG. 6C. Here, the fraction less than 100, 12, is placed at the end. When the same values exist in the quantization numbers of the original images, it is preferable to rank the same values in a predetermined order, for example, in the order in which they are examined for size comparison. For the values in FIG. 6B or C, if 51 or more is black and 50 or less is white, three blacks will be reproduced. Let us now examine the nature of the constant C in the equation in step (3) above. The reason why the constant C is set to be the same as the maximum quantization number is to make the halftone dot area of the reproduced image as close as possible to the halftone dot area of the original image. However, depending on the purpose, it may be possible to create reproduced halftone dots with larger or smaller halftone dots than the original image. In this case, the former case can be operated by setting C smaller than the maximum quantization number, and the latter case by setting C larger than the maximum quantization number.
For example, if C = 85, the redistributed values will be 3 85s and 1 57, and below 50 will be white,
If 51 or more is black, then 4 blacks will be played. Similarly, if C = 135, the redistributed values will be two 135s and one 42, and if 50 or less is white and 51 or more is black, two blacks will be reproduced. Become. Now, in the explanation of Figures 6A to 6C, an example was shown in which images are redistributed to the entire original image at once, but in general, actual image data is not a small number as shown in Figure 6A. The number is an order of magnitude larger, and such processing is unrealistic. Therefore, in the present invention, a scanning window is used to perform practical image processing.
The size of the scanning window should be decided based on the complexity and effectiveness of the image processing, so here we will explain, as an example, the operation of image processing using a scanning window consisting of a 3 x 3 pixel matrix. . FIGS. 7A to 7J are diagrams for explaining scanning by scanning lines and image redistribution. A in the same figure is original image data, which has the same numerical value as halftone dot 1 in FIG. Figure 7A
The original image data is main-scanned to the right and sub-scanned downward in a 3×3 pixel scanning window 10 shown by a thick frame in the figure. As the scanning window 10 scans, the results of image redistribution are sequentially obtained from B to J in the figure. First, the image redistribution described in FIGS. 6A to 6C is performed with respect to the data within the scanning window 10 of FIG. 6A. The result, as shown by the dotted line frame 11 in FIG. 7B, is the same as the data in FIG. 7A. Next, as shown in Figure B, the scanning window 10 is moved by one pixel in the main scanning direction and the pixels are redistributed within the scanning window 10 in Figure B, resulting in the result as shown in the dotted line frame 11 in Figure C. Thereafter, the scanning window 10 is moved and redistributed as shown in C in the same figure, and the result in the dotted line frame 11 in D in the same figure is obtained.The movement and redistribution of the scanning window 10 is then repeated, but the subsequent data is 0. Therefore, the redistribution result does not change. When the movement of the scanning window 10 in the main scanning direction is completed, the scanning window 10 is returned to the beginning of the main scanning and moved by one pixel in the sub-scanning direction, as shown in FIG. Hereinafter, the redistribution result of the scanning window 10 in D in the same figure is shown in the dotted line frame 11 in E in the same figure, the result of redistribution of the scanning line 10 in E in the same figure is shown in the dotted line frame 11 in F in the same figure, and the scanning window 10 in F in the figure Data conversion is performed as shown in the dotted line frame 11 in G in the same figure, and as shown in the dotted line frame 11 in H in the same figure, the reallocation result of the scanning window 10 in G in the same figure is shown in the dotted line frame 11 in FIG. After G in the figure, the scanning and redistribution results of the scanning window 10 remain unchanged. When the movement of the scanning window 10 in the main scanning direction is completed, the scanning window 10
returns to the beginning of the main scan again, as shown in the scanning window 10 in FIG. The redistribution result of the scanning window 10 in FIG. 2H is converted into data as shown in the dotted line frame 11 in FIG. 1, and the redistribution result of the scanning window 10 in FIG.
Thereafter, as scanning and reallocation of the scanning window 10 are repeated, the data 12 is converted once again when the scanning window 10 moves to the position shown in FIG.
At this time, if the binarization level V is 50, the data 12
is replaced with O, and the error correction amount E becomes 12, which is added at the same time when calculating the sum of the image signal levels within the scanning window at the next position. In addition, in this embodiment, the scanning windows 10 of A to J in the same figure
The same numbers did not appear except for 0 and 100, but when ranking the same values, for example, the first
It is sufficient to predetermine a procedure such as prioritizing the second value in the sub-scanning direction and prioritizing the second value in the main scanning direction to determine the order. Figure 8 shows that the original data in Figure 1 has been redistributed by scanning 3 x 3 pixel scanning lines, and then binarized with 51 and above being black and 50 and below being white. This is a reproduced image. However, the numerical values in the figure are the numerical values of the original image data, not the numerical values of the redistribution results. In the figure, all halftone dots are reproduced using three pixels, and it can be said that a reproduction result with minimal error was obtained. Next, an example of the case where the present invention is applied to a grayscale image having an intermediate level is shown in FIGS. 9A to 9Z. Figure A shows image data expressed in hexadecimal, with the blackest value being F and the whitest value being O. The image data is main-scanned in the right direction in the figure and sub-scanned in the downward direction using a 3×3 pixel scanning window.
In A to Z of the same figure, the solid line frame indicates the area of the scanning window 12 before the data is redistributed, and the dotted line frame 13 indicates the area after the data is redistributed. The area inside frame 14 is the error correction amount E.
The flag is cleared to O at the beginning of main scanning. here,
The maximum image signal level C in the above step (4) and the value of the binarization level V in step (4) are expressed in hexadecimal notation as C
The manner in which the above-mentioned steps (2) to (6) are executed with =F and V=7 will be described below. By adding the contents of the scanning window 12 and the contents of the frame 14 in Figure A and redistributing them, the result will be as shown in the dotted line frame 13 in Figure B. In this data conversion, the error correction amount E in the frame 14 is 0. Next, the contents of the scanning window 12 and the frame 14 in FIG. 1B are added and redistributed, resulting in the result shown in the dotted line frame 13 in FIG. The error correction amount E in the frame 14 is O. After moving the scanning window 12 and adding and redistributing data as shown in FIG. , and continue in the same way as shown in G to J in the same figure. The sub-scanning is further shifted to K in the same figure, L in the same figure, M in the same figure, and N in the same figure, but up to this point the error correction amount of the frame 14 remains 0. In N of the same figure, the added value of the scanning window 12 and the frame 14 is C, and this value is replaced twice in the above procedure (4). Specifically, the value of B of the scanning window 12 is C
is changed to P _1ST , and then F is replaced as P _2ND , resulting in a dotted line frame 13 in the same figure. At this time, the error correction amount E in the time frame 14 is -3. Adding the contents of the scanning window 12 and the contents of the frame 14 in the figure is -2
Therefore, the redistribution results are all 0 as shown in the dotted line box 13 in FIG. Thereafter, the data will be converted as shown in P to Z in the figure. Next, the image signal processing device of the present invention will be explained. FIG. 10 is a block diagram of a data conversion circuit, and is a diagram for explaining the outline of the basic operation for carrying out the present invention. Analog image signal 15 obtained by scanning the original image
is converted into a digital image signal by the A/D converter 16, and is stored in the image data storage device 18 through the gate circuit 17. The storage device 18 stores multiple lines of the original image in the main scanning direction (data conversion of the present invention is performed for three lines).
It has a storage capacity that can store image data for 3 lines if processing is performed using a scanning window of 3 pixels.
The storage address of the image data is specified by the address control circuit 19. The data addition circuit 20 obtains the data within the scanning window from the storage device 18 through the gate circuit 17, and calculates the sum of this data and the data obtained from the error correction amount designation circuit 21. The ranking circuit 22 obtains the data within the scanning window from the storage device 18 through the gate circuit 17, and stores the data in the storage device 18 in descending order of the data.
All the data addresses of the corresponding scanning window positions are determined and notified to the address control circuit 19. The redistribution circuit 23 creates converted data from the summation obtained by the addition circuit 20, and sequentially writes the converted data to the address of the storage device 18 specified by the address control circuit 19 through the gate circuit 17. At the same time, step (4) above
The error correction amount of pixel P _i,j at is also calculated and notified to the error correction amount designation circuit 21. The timing at which data is written to the pixel P _i,j is informed from the address control circuit 19 via the signal line 24 . The error correction amount designation circuit 21 determines the error correction amount based on the sum total from the addition circuit 20 and notifies the addition circuit 20 of the amount. This error is used when calculating the sum of the next scanning window. The data for which all data conversion processing for redistribution has been completed is read out from the storage device 18 through the gate circuit 17, passes through the binarization circuit 25, and becomes an output image signal 26 to be recorded by the image recording device. A timing signal generation circuit 27 sends a timing signal to each block to synchronize the entire block. Next, the ranking circuit 22, the redistribution circuit 23, and the error correction amount designation circuit 21 will be explained. FIG. 11 shows details of the ranking circuit 22 of FIG. 10. The nine pieces of data within the 3.times.3 pixel scanning window pass through the gate 29 from the terminal 28 and are stored in predetermined positions of the nine data registers 30 corresponding to the positions within the scanning window. The predetermined position at this time is the terminal 31
This is specified by setting the output of the address counter 32 that counts the timing signal input from the address counter 32 to the register 30 via the gate 33. terminal 3
The timing signal entering from 1 passes through gate 34,
It also serves as the data write clock for the register 30, and at the same time enters the timing control circuit 35, and is connected to the signal line 3.
A gate switching signal is sent to 6. The gate switching signal on the signal line 36 drives the gates 29, 33, and 34, creating an input mode state in which nine data input from the terminal 28 are input into the register 30. The maximum value detection circuit 37 has a register 30
The maximum value is detected for the nine data items, and the data address of the maximum value is output. At this time, the timing control circuit 35 drives the gates 29, 33, and 34 using the gate switching signal on the signal line 36, thereby creating a rewriting mode for the contents of the register 30. In this state, the data address of the maximum value is set in the register 30 via the gate 33, the contents of the negative data constant 38 are set in the register 30 via the gate 29, and The internal clock signal outputted through the signal line 39 is sent to the gate 34.
, the maximum value data in the register 30 is rewritten to negative data by using the register 30 as a data write clock. In this state, when nine internal clocks are output on the signal line 39, the contents of the register 30 will all change to negative values. The maximum value detection circuit 37
The data addresses corresponding to the data first fetched into the register 30 in descending order of magnitude are output as outputs. This address becomes write data for the nine address registers 40 and is stored in sequence. At this time, the internal clock on the signal line 39 becomes the write clock for the register 40 and simultaneously enters the address counter 41. The output of the address counter 41 passes through a gate 42 and specifies the location where address data is to be stored in the address register 40. At this time, the signal outputted from the signal line 43 of the timing control circuit 35 drives the gate 42, indicating the data write state, that is, the address counter 41.
The output of the address register 40 is given to the address register 40. After nine address data are written in the address register 40, the signal line 43 drives the gate 42,
The data of the address register 40 is set to a read state. After this, when the timing control circuit 35 outputs a read clock to the signal line 44, the address counter 45 counts this and supplies the output to the address register 40 through the gate 42 to specify the read position of the address data. . In this way, address data from the ranking circuit is output to terminal 46. FIG. 12 shows details of the redistribution circuit 23 of FIG. 10. The sum S of data within the scanning window is set in a register 49 from a terminal 47 via a gate 48. The timing signal at terminal 50 drives gate 48 to set sum S in register 49 at terminal 47.
, and the other signals are passed through the output signal of the subtracter 51. The timing for reading data into the register 49 is determined by a timing signal input from a terminal 52. The subtracter 51 subtracts the constant C of the register 53 from the contents of the register 49 and outputs the result.
Therefore, the output of the register 49 is sequentially subtracted by the constant C from the initial sum S every time a timing signal is input from the terminal 52. Comparator 54 is register 4
The contents of register 53 are compared with the contents of register 53, and if the contents of register 49 are larger or the same, gate 55 is driven and the contents of register 53 are output from gate 55, and when the contents of register 49 are small, gate 55 is driven.
is driven to make the contents of the register 49 the output of the gate 55. The output of gate 55 is output to terminal 58 via gate 56 and gate 57. The positive/negative determination circuit 59 drives the gate 56, and when the output of the register 49 is positive, the output of the gate 55 is set as the output of the gate 56, and when the output of the register 49 is negative, the constant O of the register 60 is set as the output of the gate 56. do. The output of gate 56 becomes the input of gate 57, while
A comparator 61 compares the value with a constant V of a register 62. The output of comparator 61 drives gate 63, and if the output of gate 56 is greater than the contents of register 62, the constant C of register 53 is driven by gate 63.
If it is not larger, the constant 0 of the register 60 is set as the output of the gate 63. The output of gate 63 becomes the input of gate 57. Gate 57 is terminal 6
4, the output of gate 56 or the output of gate 63 becomes the output of gate 57. The signal on the terminal 64 is the signal on the signal line 24 explained in FIG. 10, and instructs the timing of writing data to the above-mentioned pixel P _i,j . At this time, the output of gate 63 is used as the output of gate 57. The value obtained by subtracting the constant C of the register 53 from the output of the gate 56 by the subtracter 65 and the value obtained by subtracting the constant C of the register 53 from the output of the gate 56
The data of either one of the outputs passes through the gate 66 and is written into the register 67. The gate 66 is driven by the output of the comparator 61, and if the output of the gate 56 is greater than the constant V of the register 62, the subtracter 65 is driven.
The output of the gate 56 is taken as the output of the gate 66, and if it is not larger, the output of the gate 56 is taken as the output of the gate 66. The register 67 takes in the output of the gate 66 at the timing of writing data to the above-mentioned pixel P _i,j of the terminal 64 . The output of the register 67 is from the terminal 68 to the 10th
It is applied to the error correction amount designation circuit 21 explained in the figure. This is the value of the error correction amount E in step (4) described above. Figure 13 shows the error correction amount designation circuit 21 in Figure 10.
The details are as follows. Total sum S of data input from terminal 69
is subtracted by the contents of the register 71 in the subtracter 70. The contents of the register 71 are the values of C×(m+1)×(n+1) in the above procedure (4), and are 3× in this example.
For a 3-pixel scanning window, it is a constant of 9×C. Subtraction value 7
The output of 0 enters the positive/negative determining circuit 72 and the gate 73. When the positive/negative determining circuit 72 determines that the output of the subtracter 70 is positive or 0, it drives the gate 73 and makes the output of the subtracter 70 the output of the gate 73. terminal 69
The total sum S of data input from the gate 74 is also input to the positive/negative determination circuit 74 and the gate 75. In the positive/negative determining circuit 74, when the total sum S of data is negative, the outputs of the gates 73 and 75 pass through an OR circuit 76 and a gate 77 and become a signal at a terminal 78. The other input signal to the gate 77 is the error correction amount E in step (4) of the terminal 68 in FIG.
The value of is input from terminal 79. The output signals of the positive/negative determining circuit 72 and the positive/negative determining circuit 74 (signals that turn on the gates 73 and 75) drive the gate 77 through the OR circuit 80, and the output of the OR circuit 76 is used as the output of the gate 77. Therefore, when gate 73 and gate 75 are off, the signal at terminal 79 is
The output will be 7. The signal at terminal 78 specifies the amount of error correction to be used in the calculation of the next scanning window. Effects of the Invention As described above, according to the present invention, a scanning window is set to be sequentially scanned with respect to the image signal level of each pixel obtained by scanning and decomposing the original image, and When reproducing black pixels by redistributing the total image signal level, the error that occurred during the reproduction of black pixels in the scanning window at the previous scanning position is used as the image in the scanning window at the next scanning position. Since it is added at the same time when calculating the total signal level, the image after being binarized by redistribution becomes high-quality binarized image data without moiré fringes. Therefore, for the purpose of removing moiré, there is no need to perform higher-density scanning than necessary or change the scanning line density depending on the number of dots in the original image, which improves the economy and operability of equipment production and operation. can be achieved. The present invention also has the effect of improving the connection of thin lines in characters and line drawings.
Further, when the image processing of the present invention is applied to an original image having an intermediate density, it is possible to obtain binarized image data representing an average intermediate density including surrounding pixels. Therefore, even if the original image contains intermediate density level images and binary level images such as halftone dots and characters, high-quality image processing results can be obtained using the same processing method.

[Brief explanation of drawings]

FIG. 1 is a diagram showing quantization of halftone dots, FIG. 2 is a diagram showing a binarized image of halftone dots reproduced by the conventional method,
Figure 3 shows an example of the dither method for expressing intermediate density in binary form, and Figure 4 shows the result of binarizing the halftone image data in Figure 1 using the threshold table in Figure 3B. FIG. 5 is a diagram explaining the principle of the present invention, FIGS. 6 A to C are diagrams explaining redistribution of image data, and FIGS. 7 A to J are diagrams explaining scanning of the scanning window and redistribution of image data. FIG. 8 is a diagram illustrating a binarized image obtained by applying the image signal processing method of the present invention to the image data in FIG. 1, and FIGS. FIG. 10 is a block diagram showing an embodiment of the image signal processing device of the present invention, and FIG.
FIG. 1 is a block diagram showing a ranking circuit, FIG. 12 is a block diagram showing a redistribution circuit, and FIG. 13 is a block diagram showing an error correction amount designation circuit. 15...Analog image signal input terminal, 16...
A/D converter, 17...gate circuit, 18...storage device, 19...address control circuit, 20...addition circuit, 21...error correction amount designation circuit, 22...
Ranking circuit, 23...Redistribution circuit, 24...Pixel
Signal line for timing of writing data to P _i,j , 25
... Binarization circuit, 26 ... Image signal output terminal, 2
7...Timing signal generation circuit, 28...Data input terminal, 29...Gate circuit, 30...9 data registers, 31...Timing signal input terminal, 32...Address counter, 33, 34...
Gate circuit, 35... Timing control circuit, 36
... Gate switching signal line, 37 ... Maximum detection circuit, 38 ... Negative data constant register, 39 ...
Internal clock signal line, 40...9 address data storage registers, 41...address counter, 42...gate circuit, 43...gate switching signal line, 44...readout clock signal line, 45...
...address counter, 46...address data output terminal, 47...summation S input terminal, 48...gate circuit, 49...register, 50...timing signal input terminal, 51...subtractor, 52...timing Signal input terminal, 53...Register of constant C, 54...Comparator, 55, 56, 57...Gate circuit, 58...Redistribution data output terminal, 59...
...Positive/negative judgment circuit, 60...Register of constant O, 6
1...Comparator, 62...Register of constant V, 63
...Gate circuit, 64...Signal input terminal for timing of writing data to pixel P _i,j , 65...Subtractor,
66... Gate circuit, 67... Register for storing error correction amount, 68... Output terminal for error correction amount,
69...Input terminal for sum S, 70...Subtractor, 7
1... Constant C x (m+1) x (n+1) register,
72... Positive/negative judgment circuit, 73... Gate circuit, 7
4... Positive/negative judgment circuit, 75... Gate circuit, 76
... OR circuit, 77 ... Gate circuit, 78 ... Output terminal for error correction amount specified value, 79 ... Input terminal for error correction amount, 80 ... OR circuit.

Claims (1)

[Scope of Claims] 1. A first step of storing the image signal level of each pixel obtained by scanning and decomposing the original image in an image signal storage unit, and M adjacent pixels (M is a natural number) from the storage unit. The sum Sm of the image signal levels of all the pixels within the scanning window selected while moving the pixels by a predetermined pixel and the error correction amount E that occurred at the previous scanning window position.
and a third step of inputting the image signal levels of the M pixels from the scanning window and ranking the M pixels in descending or ascending order of the image signal levels. According to the procedure and the predetermined image signal level C, when O≦S≦C×M, S=C×N+A
(However, N is an integer such that O≦N≦M, and A is O≦
Find N and A such that A<C), and rank them in descending or ascending order according to the image signal level within the scanning window.
The 1st to Nth plane elements have an image signal level C, and the Nth
The +1st pixel has an image signal level of A, and N
The +2nd and subsequent pixels are subjected to primary replacement with the image signal level set to O, and then the pixels that are no longer included within the scanning window due to the movement of the scanning window are replaced after the primary replacement. By comparing the image signal level P _1ST with a predetermined binarization level V such that O≦V<C, if it is larger, C is replaced, and if it is not larger, O is replaced with the image signal level P _1ST after the second substitution. The image signal level P _2ND is given as the image signal level P 2ND, and the second substitution is performed, and then the sum of the differences between the image signal level P _1ST and the image signal level P _2ND is set as the error correction amount E at the next movement position of the scanning window. , when O>S, the M pixels in the scanning window are replaced with an image signal level of O, and the sum S is the error correction amount E at the next moving position of the scanning window, so that S>C×M. When the M pixels in the scanning window are replaced with an image signal level of C,
and a fourth step in which C×M is the error correction amount E at the next moving position of the scanning window. 2. Storage means for storing the image signal level of each pixel obtained by scanning and decomposing the original image, and scanning for selecting M adjacent pixels (M is a natural number) from the storage means while moving them by a predetermined pixel. a data adding means for calculating the sum Sm of the image signal levels of the M pixels selected by the scanning window and the sum S of the error correction amount E generated at the position of the previous scanning window; Input the image signal levels of M pixels, and select the M pixels in descending or ascending order of the image signal levels.
When O≦S≦C×M, S=C×N+A (however, N is O≦N≦M). an integer, A is O≦A
<C), and according to the ranking determined by the ranking means, the 1st to Nth pixels have a pixel signal level of C, and the (N+1)th pixel has a pixel signal level of A. , (N
+2) The pixels from the th to the Mth are subjected to a first permutation in which O is assigned as the image signal level, and when O>S, all M pixels are subjected to a second permutation to have an image signal level of O, Also, when S>C×M, all M pixels are C
redistribution means for performing a third replacement on the image signal level P 1ST, which is no longer included in the scanning window due to the next movement of the scanning window, and is replaced by the image signal level P _1ST ; , compare it with a predetermined binarization level V where O≦V≦C, and if the image signal level P _1ST is larger than the binarization level V, set C, and if not, set O
a binarization means for obtaining the image signal level P _2ND ;
When the redistribution means performs the first replacement, the sum of the differences between the image signal level P _1ST and the image signal level P _2ND is set as the error correction amount E at the next movement position of the scanning window, and When the redistribution means performs the second replacement, the sum S is set as the error correction amount E at the next movement position of the scanning window, and when the redistribution means performs the third replacement, and an error correction amount designation means for supplying S-C×N to the data addition means as an error correction amount E at the next moving position of the scanning window.