JPH07143339A - Image signal processor - Google Patents

Abstract

PURPOSE:To decide a combination of coefficients for each marked pixel in a small circuit scale and also to decide the distribution of coefficients without limiting the combination patterns of coefficients by selecting at random a coefficient for each peripheral pixel out of plural types of coefficients which are stored in a coefficient storage means. CONSTITUTION:The image density data on a marked pixel 1 is compared with the threshold value outputted from a threshold value generating part 5 by a comparator 6. Then an error ERR is calculated based on the result of comparison of the comparator 6. At an error distribution pattern deciding part 7, the random number outputted from a random number generating circuit 15 are inputted to the selectors 11-14 directly or via the exclusive circuits 16-18. The selectors 11-14 store the prepared coefficients K1-K4 respectively, and one of these stored coefficients is selected by each selector and outputted as the error distribution coefficients KA-KD. The coefficients KA-KD are added to the error ERR when the pixel 1 is threshold processed. So that an error to be assigned to each peripheral pixel is calculated.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for converting a multi-tone image data into a pseudo-tone representation by a binary image in order to make the multi-tone image data easy to handle when outputting the data for communication, printing and calculation. Regarding

[0002]

2. Description of the Related Art In recent years, due to the mechanization of office work and the development of information and communication networks, not only manuscripts represented only by black and white binary,
There is an increasing demand for processing originals expressed in multiple gradations.

Further, in order to correspond to the one having a simpler structure and easily available due to the amount of information, the communication mode, the image reproducing device such as a printer, etc., a multi-tone image is a high-quality binary image. Therefore, there has been a demand for a means of converting into a pseudo gradation expression, storing and transmitting. As a method of such pseudo gradation conversion, there is a dither method, but this method has a problem that a texture pattern called moire is generated by repeatedly using the contents of a dither matrix prepared in advance.

Therefore, an error diffusion method has been used as a method for obtaining a higher quality pseudo gradation image. In this method, a binarization error generated when a certain pixel of interest is binarized by a preset threshold value is multiplied by a certain coefficient and added to the surrounding pixels.
Also in this method, depending on the method of setting the coefficient for distributing the binarization error, a texture pattern may be generated by repeating the same distribution coefficient pattern. As a means for determining the distribution coefficient for removing such a texture pattern, several distribution patterns of the distribution coefficient to the peripheral pixel positions are prepared in advance, and the pattern is sequentially selected from among them to calculate the coefficient. The method of deciding a pattern is taken.

FIG. 3 is a diagram for explaining a conventional coefficient determination method as disclosed in Japanese Patent Laid-Open No. 63-102475.

In the error diffusion method, the value of the pixel of interest (pixel indicated by a star in FIG. 3) is compared with a threshold value to create binary data, and the error ERR generated at that time is represented by A to D in FIG. The pixel is distributed and added to the pixel at the position to be a new pixel of interest. As an error distribution method at that time, as shown in the error distribution determination means in FIG. 3, register sets 1 to n in which combination patterns of error distribution coefficients are set in advance are provided, and one of them is set by the register setting means. It is a proper (random) selection of register sets.

For example, in the case of this example, register sets 1 to
As shown in the figure, four coefficients K1 to K4 for distributing errors to the respective positions A to D are set in n in various orders. And register set 1
By selecting one of n, pixel positions A to D are selected.
The coefficients KA-KD corresponding to each of the above are determined. Then, from the coefficients KA to KD and the error value ERR when the pixel of interest is binarized, the distribution error to be added to each pixel data of the pixel positions A to D is obtained as follows.

Distribution error to A pixel position: ERR × KA Distribution error to B pixel position: ERR × KB Distribution error to C pixel position: ERR × KCD Distribution error to pixel position: ERR × KD

[0009]

However, when the coefficient distribution pattern is determined by the method as described above, the number of register sets increases and the circuit scale becomes very large when actually configuring the circuit. . For example, when the above-mentioned distribution pixels are set at four positions A to D and the coefficient is four, the number of distribution patterns is ₄ P ₄ = 24, and it is necessary to have 24 combinations of registers. There is a problem that the circuit scale becomes very large.

Therefore, in many cases, the number of combination patterns is usually limited to about several. For example, Japanese Patent Laid-Open No. 63-102475 discloses two examples. However, if the number of patterns of combinations is reduced in this way, the number of combinations that can be selected is reduced, and the clutter of error distribution is reduced, and it may not be possible to sufficiently suppress the occurrence of texture patterns. .

In view of the above problems, the present invention determines a combination of coefficients for each pixel of interest with a small circuit scale, and determines coefficient distribution without limiting the combination pattern. To aim.

[0012]

According to a first aspect of the present invention, when converting image data input at multiple gradation levels into a binary image, a pixel of interest is binarized by a preset threshold value. The binarization error generated at the time of binarization is distributed to the peripheral pixels located around the binarization based on the coefficient set for each peripheral pixel, and the error distributed from each target pixel is added. A coefficient storage unit that stores the plurality of types of coefficients for allocating the binarization error in the image signal processing device that stores the image data, reads the stored image data, and uses the image data as the next pixel to be binarized. A coefficient selecting means for randomly selecting a coefficient for each peripheral pixel from among a plurality of types of coefficients stored in the coefficient storing means when the binarization error of one pixel of interest is distributed to the peripheral pixels; Equipped with The features.

According to a second aspect of the present invention, the coefficient storage means is a means for storing coefficients corresponding to the number of peripheral pixels to which the binarization error is distributed, and the coefficient selection means is already different. It is characterized in that it is means for randomly selecting from the coefficients excluding the coefficients set in the peripheral pixels of.

[0014]

FIG. 1 is a diagram for explaining the operation of the present invention. This figure shows 2 of 1 of the target pixel (pixel indicated by star in the figure)
It shows a state in which the binarization error is distributed to the peripheral pixels A to D.

The coefficient storage means 2 stores a plurality of types of coefficients K.
1, K2 ... Are stored. Coefficient selection means ~
One of these coefficients K1, K2, ... Is randomly selected and the value is set as the coefficient of each pixel A to D. As a result, the combination pattern of the coefficients for the pixels A to D is changed each time. At this time,
The pixels A to D may have different coefficients as described in claim 2, but the same coefficient may be set to be overlapped.

As a means for randomly selecting the coefficient, for example, when the lower several bits can be expected to be random as in a natural image, information of the following bits of the pixel of interest is used. it can. A random number generator may be used when the lower bits cannot be expected to be random, such as an image created by a computer.

[0017]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 2 is a diagram showing the configuration of the error distribution pattern determination unit of the image signal processing apparatus according to the embodiment of the present invention. In this example, as in the case of FIG. 3, the binarization error of the target pixel 1 is distributed to the four surrounding pixels A to D.

In the figure, the image density data of the target pixel 1 is compared with the threshold value output from the threshold value generating section 5 by the comparator 6, and the error ERR is calculated based on the result. The error is weighted by the error distribution pattern determination unit 7 shown in the figure, and then distributed to each pixel position and added.

The configuration of the error distribution pattern determination unit 7 will be described.

In the figure, a random number generating circuit 15 is a circuit for generating a 2-bit random number. The number of bits is appropriately set according to the number of selected coefficients. The random number output from the random number generation circuit 15 may be the direct or exclusive circuit 1.
It is input to the selectors 11 to 14 via 6 to 18. The exclusive circuits 16 to 18 are circuits that prevent random numbers having the same value from being input to different selectors. The selectors 11 to 14 are circuits that determine weighting coefficients (error distribution coefficients) KA to KD of the respective peripheral pixels A to D, and the coefficients K1 to K4 are stored in advance. In addition,
In this embodiment, since there are four peripheral pixels, four types of coefficients K1 to K4 are stored, but the number of stored coefficients is appropriately set according to the number of peripheral pixels. The coefficients K1 to
K4 is a normalized value.

The selectors 11 to 14 are four prepared in advance.
One coefficient K1 to K4 is stored, and one of the stored coefficients is selected and output as error distribution coefficients KA to KD. For example, the selector 11 stores the coefficients K1 to K4 and outputs any one of them as the error diffusion coefficient KA of the peripheral pixel A. Similarly, the selector 12 sets one of the stored coefficients K1 to K4 as the error diffusion coefficient KB of the peripheral pixel B, and the selector 13 sets one of the coefficients K1 to K4 as the error diffusion coefficient KC of the peripheral pixel C. , The selector 14 outputs one of the coefficients K1 to K4 as the error diffusion coefficient KD of the peripheral pixel D, respectively.

Output error diffusion coefficients KA, KB, K
C and KD are the error E in binarizing the target pixel 1, respectively.
By being added to RR, the error distributed to each peripheral pixel is calculated.

The random number generating circuit 15 outputs three types of random numbers Rnd1, Rnd2 and Rnd3 each consisting of 2 bits. Of these, the random number Rnd1 is directly input to the selector 1. The selector 1 uses this random number Rnd1 to select one of four types of coefficients K1 to K4 stored in advance,
The error diffusion coefficient KA is output.

The random number Rnd2 is output to the exclusive circuit 16. The exclusive circuit 16 compares Rnd2 with the numerical value (Rnd1 in this case) input to the selector 11, and if they do not match, Rnd2 is directly used as Rnd2 ′.
Is output to the selector 12. On the other hand, if Rnd2 matches the numerical value input to the selector 11, then R
nd2 is changed, and if the changed value does not match the numerical value input to the selector 11, it is output as Rnd2 '. As a method of changing Rnd2, for example, Rn2
There are a method of adding a predetermined number (such as "1") to d2 and a method of generating a random number from the random number generation circuit 15 again. In this way, a value different from that of the selector 11 is input to the selector 12, and the error diffusion coefficient KB is selected from the coefficients K1 to K4 based on this.

Similarly, the exclusive circuit 17 has a random number Rnd3.
And when the numerical value (Rnd1) input to the selector 11 and the numerical value (Rnd2 ′) input to the selector 12 are input and Rnd3 is different from Rnd1 and Rnd2 ′, the value is regarded as Rnd3 ′. It is output and the error diffusion coefficient KC is selected based on this. In addition, R
R when nd3 has the same value as Rnd1 and Rnd2 '
The setting of nd3 'is performed in the same manner as above.

The exclusive circuit 18 includes Rnd1 and Rnd.
2'and Rnd3 'are input. The exclusive circuit 18 inputs to the selector 14 a 2-bit numerical value that does not have the same value. Then, the error diffusion coefficient KD is selected based on this.

The respective errors of the peripheral pixels A to D are calculated based on the error diffusion coefficients KA to KD selected as described above.

In the above embodiment, the exclusive circuits 16 to 1
Although 8 is provided so that the coefficient assigned to another pixel cannot be set, the same coefficient as the coefficient assigned to another pixel may be set without providing the exclusive circuit. However, in that case, it is necessary to normalize the coefficients of the peripheral pixels after selecting the coefficients, and the normalization process is performed after assigning the coefficients to each pixel.

[0030]

According to the present invention, the coefficients used for error distribution are stored in advance, and the coefficient is distributed for each pixel from among them. That is, although the coefficients themselves are stored in advance, the storage means may be small in scale because the coefficients are not stored in a combined state but in a single state. Then, the combination of coefficients for each pixel is decided at any given time, and the combination pattern is randomly selected from all the combination patterns for the coefficients. That is, the number of combination patterns of coefficients increases, and it is possible to efficiently suppress the texture pattern that occurs when a multi-tone image is converted into a binary pseudo-tone expression, and obtain a high-quality pseudo-tone. become.

[Brief description of drawings]

FIG. 1 is a diagram for explaining the operation of the present invention.

FIG. 2 is a diagram showing a main configuration of an image signal processing device according to an embodiment of the present invention.

FIG. 3 is a diagram showing a main configuration of a conventional image signal processing device. 1 Pixel of Interest 7 Error Distribution Pattern Generation Unit 11 to 14 Selector (Coefficient Storage Means) 15 Random Number Generation Circuit 16 to 18 Exclusive Circuit A to D Peripheral Pixels for Binarization Error Distribution

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI technical display location 4226-5C H04N 1/40 103 A

Claims (2)

[Claims] 1. Image data input at multi-gradation levels
When converting to a value image, the pixel of interest is binarized by a preset threshold value, and the binarization error generated at the time of binarization is set for each of the peripheral pixels located in the periphery of the binarized error. In the image signal processing device, which is sorted based on the calculated coefficient, the errors sorted from each target pixel are added and stored, and the stored image data is read out and used as the next binarized target pixel. Coefficient storage means for storing a plurality of types of coefficients for allocating the binarization error, and a plurality of types of coefficients stored in the coefficient storage means for allocating the binarization error of one pixel of interest to peripheral pixels. An image signal processing device, comprising: a coefficient selecting unit that randomly selects a coefficient for each peripheral pixel from among them. 2. The coefficient storage means is means for storing coefficients corresponding to the number of peripheral pixels to which the binarization error is distributed, and the coefficient selection means is already set to another peripheral pixel. The image signal processing apparatus according to claim 1, wherein the image signal processing apparatus is means for randomly selecting from coefficients excluding existing coefficients.