JPH02107063A - Picture signal processor - Google Patents

Abstract

PURPOSE:To obtain a smooth gradation characteristic and a high resolution characteristic by making an error-corrected input level into a multilevel with the comparison of plural thresholds, subtracting the value from the corrected level obtained with adding a noticed picture element and the integrated error of a noticed picture element position, end obtaining the multilevel error. CONSTITUTION:A corrected error arithmetic means 12, which addition-operates an integrated error Sxy corresponding to a noticed picture element position 3 in an error memory means 1 and the error of an periphery 2 and outputs the error correcting level, and an error distribution updating means 11, which calculates the error distribution value from the multivalued error from a differential arithmetic means 8 and an error distribution coefficient, adds the value to the integrated error at the corresponding picture element position in the error memory means 1, and again storing the value, are provided. Further, to a noticed picture element 3, a first input correction level obtained by taking into consideration the integrated error Sxy corresponding to the noticed picture position and the integrated error of the periphery 2 are made into the multilevel by the plural thresholds, and the new multilevel error is obtained from the sum between the noticed picture element 3 and the integrated error Sxy and the difference between the element 3 and a multilevel output Pxy. Thus, the gradation characteristic and resolution are improved and the high-speed picture processing is attained.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to an image signal processing device having a function of performing multi-value reproduction of image information including gradation images.

BACKGROUND OF THE INVENTION In recent years, with the mechanization of office processing and the rapid spread of image communication, there has been an increasing demand for high-quality image reproduction of gradation images and printed images in addition to conventional black-and-white binary originals.

In particular, pseudo gradation reproduction using a binary image of a gradation image is highly compatible with display devices and recording devices, and many proposals have been made.

The dither method is the most well-known method for reproducing these pseudo gradations. This method reproduces gradation based on the number of dots reproduced within a predetermined area, and compares the input image information with the threshold value prepared in the IJ box for each pixel. Binarization processing is then performed. In this method, the gradation characteristics and the resolution characteristics directly depend on the size of the dither matrix, and are incompatible with each other. Furthermore, when used for printed images, etc., the occurrence of moiré patterns in the reproduced image is unavoidable.

A random dither method has been proposed as a method that achieves both the above gradation characteristics and high resolution, and is highly effective in suppressing the occurrence of moiré patterns.A representative example of this method is the error diffusion method [R.F. Gray Scale” Varnish I Day 75 Digest 36-37 pages (Reference: R, FLOYD & L, 5TEINB
ERG, ” An Adaptive Al-gor
ithm for 5patiat Gray 5ca
le”, SID 75 DIGEST, pp36-
37) ) has been proposed.

FIG. 4 is a block diagram of the main parts for realizing the above error diffusion method. Let the coordinates of the pixel of interest in the current image be (x,
y), 401 is an error storage means, 402 is an unprocessed pixel area around the pixel of interest indicated by the error distribution coefficient matrix, and 403 is a coordinate (x.y).

pixel of interest for integration error Sxy at y), 404 is coordinate OL, input terminal for input level IXY at y),
405 is an input correction means for Ixy (= Ixy + 5xy), 406 is a binary level Pxy of output level O or R
, 407 is a signal terminal that applies a constant threshold R/2, 408 compares the input signal Ixy with the constant threshold R/l, and when Ixy≧R/2, Pxy=R, otherwise Pxy=O ? : Output binarization means, 409 is Exy
(= I xy - Pxy) This is a difference calculation means for calculating the binarization error for the pixel of interest.

Now, the integration error Sxy for the pixel of interest is the (1)th, (
2) It is expressed by the formula.

5xy=nKij 0 Ex-j-H, y-i-+-+
°−=(1) (However, i and J indicate the coordinates in the error distribution coefficient matrix.) This error distribution coefficient Klj is used to weight the distribution of the error Exy to the surrounding pixels of the pixel of interest. However, * indicates the position of the pixel of interest).

In the configuration shown in FIG. 4, the above calculation adds the binarization error Exy to the pixel of interest to each pixel A in the unprocessed surrounding pixel area 2.
-D is multiplied by the corresponding distribution coefficient, and the error storage means 401
This is realized by an error distribution calculating means 410 which adds the value to the value within and stores the corresponding position again. However, the integrated error at pixel position B in the error storage means 1 is cleared to 0 in advance.

Problems to be Solved by the Invention As already mentioned, the error diffusion method is superior to the systematic dither method in terms of gradation characteristics and resolution, and also eliminates moiré patterns when reproducing printed images. It has the characteristics of having a large suppressing effect.

However, depending on the structure and weighting of the error region, unique striped patterns and textures may occur depending on the coefficients, and resolution may still deteriorate. For example, if the structure of the error area is made large to make the error as small as possible on average, and to obtain smooth gradation characteristics, the resolution deteriorates. Further, such a structure has the problem that it requires a lot of calculations and the processing speed of the pixels is slow.

The present invention solves the problem of the error diffusion method described above by adding the accumulated error at the pixel position of interest calculated using a uniform weighting coefficient and the errors in its surroundings, thereby creating an error region that is a small error region but essentially functions as a large region. By doing this, the gradation characteristics are improved, and at the same time, the resolution is improved by reducing the amount of error correction for the pixel of interest, and the new error is calculated from the pixel of interest and the cumulative error to satisfy the equal density relationship. The present invention provides an image signal processing device that can easily realize several gradations of recording and display systems by producing multivalued output using a threshold level of be.

Means for Solving the Problems The present invention, when converting multi-level image signal levels sampled pixel by pixel, (A) makes the multi-level conversion error of the pixel of interest correspond to the surrounding pixel positions. (B) a correction error calculation means for adding the accumulated error corresponding to the pixel position of interest in the error storage means and the surrounding errors and outputting the error correction level; (C) the above-mentioned a first input correction means that adds the error correction level and the input level of the pixel of interest and outputs a first input correction level; (D) comparing the first input correction level and a plurality of threshold levels (E) a second input correction means that adds the input level of the pixel of interest and the integration error and outputs a second correction level; , (F) a difference calculating means for calculating a multi-value halftoning error from the difference between the second input correction level and the multi-level halftoning level, and (G) paying attention to the multi-level halftoning error and error distribution coefficient from the difference computing means. An image comprising: an error distribution updating unit that calculates an error distribution value corresponding to unprocessed pixels around the pixel, adds the error distribution value to the accumulated error of the corresponding pixel position in the error storage unit, and stores the result again. It is a signal processing device. In addition, the correction error calculation means multiplies the accumulated error corresponding to the pixel of interest and the sum of surrounding errors by a coefficient 1/2 or 1-1/2n (n is a positive integer) and performs an addition operation to correct the error. The above objective is achieved by determining the level.

Effects With the above configuration, the present invention multivalues the input correction level for the pixel of interest by using a plurality of threshold values, taking into consideration the integration error corresponding to the pixel position of interest and the surrounding integration errors, and generates a new multivalue error. This method improves the gradation characteristics and resolution by determining the difference between the sum of the pixel of interest and the integration error and the multilevel output, and also enables high-speed pixel processing by uniformly setting the error distribution coefficient to 1/2n. .

Further, the contour enhancement effect is controlled by appropriately selecting a calculation coefficient by the correction error calculation means.

Embodiment FIG. 1 shows the main block configuration of an image signal processing apparatus in an embodiment of the present invention.

In Figure 1, the coordinates of the pixel of interest in the original image are (x
, y), 1 is the error storage means, 2 is the unprocessed pixel area around the pixel of interest indicated by the error distribution coefficient matrix, and 3 is the integrated error Sxy at the coordinates (X, Y).
4 is the input terminal of the input level Ixy at the coordinates (X, y), and 5 is the input level Ixy and the correction error level exy? which is the output from the correction error calculation means 12.
: A first input correction means that inputs and outputs the first input correction level Ixxy, 6 is an output terminal for the multilevel signal Pnxy, and 7 is a first input correction level that compares the first input correction level with various threshold values. Multi-value conversion means outputs the multi-value signal Pn(x, y), simultaneously selects a multi-value conversion level corresponding to the multi-value output signal, and outputs it to the difference calculation means 8; 101 indicates the input level Ixy of the original image and the integration error Sxy; 8 is a multilevel halftoning error En(x, y)k, which is the difference between the second input correction level and the multilevel halftoning level. Difference calculation means for outputting, 10
is the result of calculating the error distribution coefficient and multi-value error corresponding to the unprocessed pixels surrounding the pixel of interest, and the accumulated error of the surrounding pixel area 2 so far, and the new accumulated error is stored again in the error storage means l. error distribution updating means for storing the error distribution at pixel positions A-D;
11 inputs the integration error Sxy corresponding to the target pixel position 3 and the integration error in the surrounding pixel area 2, and sets the error correction level e1.
．． This is a correction error calculation means that outputs.

Now, the operation of the configuration shown in FIG. 1 will be described in detail, taking quaternary output as an example.

The first input correction level output by the first input correction means 5 is the result of multiplying the integrated error Sxy corresponding to the pixel position of interest by the coefficient Ka by the correction error means 11, and also the result of multiplying the integrated error Sxy corresponding to the pixel position of interest by the coefficient Ka, and A, B, C
, D corresponding to the respective integration errors S^, SB,
Add each of Sc and So k and add the coefficient K to the result.
The error correction level exyk is obtained by adding the result of multiplying by b.
Output. The error correction level exyk input capel Ix
y and is added by the first input correction means 5 to output a first input correction level.

Next, the multivalue conversion means 7 and the difference calculation means 8 will be explained using FIG. 2. The first input correction level I+xy which is the output from the first input correction means 5 & each comparator 20
1.202.203, and compares them with preset thresholds T1, T2, T, to obtain multi-level signals A, B,
Output C.

When the first input correction level is smaller than the threshold value T, the multilevel signals A, B, and C are respectively A=B=C“0”.
'', equal to threshold T, or smaller than T, then A
= "1", B = C = "0", when equal to threshold T2 or less than Ts, A = B = "1", C = "0", when larger than threshold T, A =B=C=”l”
Output. The threshold value is set to the center value between the multi-value output levels. That is, if the maximum density signal is 255, the four-value output level is R, = O1 R1 = 85, R = 1
70, R, = 255, and the respective thresholds are T1 = (
0+85)/2=43,'L=(85+170
) / 2 = 127, T-= (170+255
)/2=212.

The multilevel output signal is a four-level amplitude modulation signal or a pulse width modulation signal shown in FIG. Now, the selector 204 inputs the multilevel signal and sets a preset multilevel output level Rnk.
It is selected and output based on the multivalued signal. For example, when the multi-value output signals A, B, and C are all "0", the multi-value output level R, = OY, 1'''0# When 0#, R, = 85, 1# "1" @0 ” when R snow = 170
, "1""1"-"1#, outputs R,=255.

The difference calculation means 8 uses the second input correction means 101 to calculate the integrated error Sxy at the position corresponding to the pixel of interest and the input level I.
The multilevel output level Rn is subtracted from the second input correction level obtained by adding
y)=Izxy Rn is output.

The multivalued error obtained here is further processed by the error distribution updating means 11.
The integration errors Sλ, Sc′, and S in the previous pixel processing process stored in the storage device corresponding to each position in the peripheral pixel area 2 are read out and the new integration errors S
^, S9, Sc, SD&calculate.

Then, the new integrated error is stored at pixel position A- in the error storage means.
Update processing is performed to store it in the storage device corresponding to D.

The above process of one pixel processing is performed by changing the first input correction level to ■
]Xy, the second input correction level l2XY, and the multi-level level of the image signal &Rn, I I Xy:I
XY + e t F=■■+Ka-8xy+Kb
(S^+SB+SD) ・・
...(3) (However, O<Ka<1.O<Kb<1)
I+xy≧T n ---pxy = RnI 1xy
<Tn------PXY=Rn-11= (
Ixy+ 5xy) -Pnxy-(5).

Furthermore, the input correction level hxy and the multilevel halftoning error Exy will be described in detail.

Now, the error distribution coefficient KA corresponding to each position in the peripheral pixel area
~KD f 1, coefficient Ka = Kb = 1, then
Correction error level e· corrected to input level Ixy.

The accumulation of errors in the previous processes is 5xy=
1/4 (Ex-+y-x+ Exy-x+ Ex+x
y-t+Ex-ty) SA = 1/4 (Ex y-1+Ex-4-+y-
1+Ex-+-z y-s) S8=O 3c = 1/4 (EX-1y) So = 1/4 (Ex-ty +Ey4-2 y
), so e−y = Sxy+ (SA + Ss + So
)=1/4 (EX-1y-t+2Exy-t+2Ex
-+-+y-1+3Ex-ty +Ex-Hy-x+E
x-2y)...(7) Therefore, in the correlation between the pixel of interest and the integrated error, a peripheral region of the pixel of interest as shown in FIG. 3, that is, an error filter structure is formed. Correcting the input correction level Ixy using an error filter formed based on such correlation means correcting error components around the pixel of interest on an average basis, resulting in a fine-grained reproduced image. can be obtained. Furthermore, by selecting the coefficients Ka and Kb to be small, the corrected scene in the processing of the pixel of interest becomes relatively small, and as a result, the pixel signal component of the eye pixel becomes large, and a reproduced image with an enhanced outline can be obtained.

Next, the new multilevel error, which is one of the features of the present invention, is
2 of the above-mentioned input correction level Izxy from the second input correction level Izxy obtained by adding the input levels Ixy.
Dilute the value by subtracting it.

The reason for this is that the error level e, which is corrected to the input level Ixy in the present invention, is a superposition of a part of the error component generated from the correlation between the pixel of interest and errors in its surroundings, and satisfies the density conservation system. It's not something you do. Therefore, the new error Exy for maintaining the density preservation system is the integration error in a system where the sum of the pixel of interest and the error distribution coefficient is 1, that is, 5xy=DKij・Ex −j +2 y−i−1−
1 and the input level Ixy.

Next, error correction level e8. When calculating, the integration error S
Coefficient Ka multiplied by xy and integration error S in the surrounding pixel area
For convenience of explanation, the coefficient Kb multiplied by the sum of ^, Sc, and SD was set as Ka=Kb=1, but Ka and Kb were each set to 0.
If it is made smaller in the range of <Ka<1.0<Kb<1, the image signal component of the pixel of interest becomes relatively large, and an output image with an enhanced outline can be obtained. For example, increase the coefficient Ka by increasing the coefficient K.
If b is made small, a precise reproduced image cannot be obtained. Furthermore, if both the coefficients Ka and Kb are set large, the correction ratio of the error component becomes positive feedback, resulting in a reproduced image that is rough overall. On the other hand, if both coefficients are made small, the reproduced image will have extremely emphasized contours. Therefore, the coefficient K
By decreasing a to relatively increase the image signal component of the pixel of interest and increasing the coefficient Kb to increase the proportion of error components around the pixel of interest, a reproduced image with enhanced contours and high precision can be obtained. . By setting these coefficients to 1/2 (n is an integer) or 1-1/2, logical operations are easy and high-speed processing is possible.

Effects of the Invention As described above, in the present invention, a plurality of input correction levels are corrected by adding a part of the integration error corresponding to the position of the pixel of interest and a part of the total sum of the surrounding integration errors to the pixel of interest. Performing multi-value conversion by comparing with a threshold value, and subtracting the multi-value conversion level corresponding to the multi-value conversion signal from the correction level obtained by adding the new multi-value conversion error to the pixel of interest and the accumulated error at the position of the pixel of interest. This method makes it possible to obtain reproduced images with smooth gradation characteristics and high resolution characteristics. Also, the correction error level e
Contour enhancement control is possible by appropriately selecting the coefficients Ka and Kb of x, and the coefficient is set to 172n or 1-1/2
By doing so, high-speed calculation is possible.

Even if the error distribution coefficient (weighting is a coefficient) is equal to -law,
It is possible to obtain a binary reproduced image that is dense, smooth, and has a high resolution without producing texture or unique striped patterns that occur with conventional error diffusion.

[Brief explanation of the drawing]

FIG. 1 is a block wiring diagram of an image signal processing device according to an embodiment of the present invention, FIGS. 2(a) and (b) are block wiring diagrams of multi-value conversion means in the main part of the same device, and FIG. FIG. 4 is a block diagram of a device for implementing the conventional error diffusion method, which is a diagram of the peripheral error region formed by the correction error calculation means in the main part of the device. 1... Error storage means, 2... Surrounding pixel area, 3...
・Positive pixel position, 4... Input terminal, 5... First input correction means, 6... Multi-value signal output terminal, 7... Multi-value conversion means, 8... Difference calculation means , 101... Second input correction means, 11... Error distribution updating means, 12...
- Correction error calculation means, 201, 202. 203...
Comparator, 204...Selector,...Modulator, 206...Recording system.

Claims (2)

[Claims] (1) Error storage means for storing the multi-value conversion error of the pixel of interest in correspondence with the surrounding pixel positions when converting the multi-level image signal level sampled in pixel units into multi-value data; and the error storage unit. a correction error calculation means for adding an integrated error corresponding to the pixel position of interest and errors around the pixel in the means and outputting an error correction level; a first input correction means for outputting an input correction level; and a first input correction means for inputting the first input correction level and comparing it with a plurality of threshold levels to output multi-value data, and converting the first input correction level into multi-value data. a multi-value conversion means for selectively outputting a corresponding multi-value conversion level; a second input correction means for adding the input level of the pixel of interest and the integration error and outputting a second correction level; Difference calculating means for calculating a multi-level halftoning error, which is the difference between the input correction level and the multi-level halftoning level, and an error corresponding to unprocessed pixels around the pixel of interest from the multi-level halftoning error and error distribution coefficient from the difference computing means. An image signal processing device comprising: an error distribution updating unit that calculates a distribution value, adds the error distribution value to the accumulated error of the corresponding pixel position in the error storage unit, and stores the result again. (2) The correction error calculation means multiplies the sum of the integrated error corresponding to the pixel of interest and the errors around it by a coefficient 1/2^n or 1-(1/2^n) (n is a positive integer), respectively. 2. The image signal processing apparatus according to claim 1, wherein the error correction level is determined by performing an addition operation.