JPH05183737A - Picture processor - Google Patents

Abstract

PURPOSE:To improve picture quality at a highlight part, and simultaneously, to improve the picture quality at a low density part as well. CONSTITUTION:The subject picture processor is to quantization-process multi- level picture data into binary data, and the remarked multi-level picture data is divided into high order data and low order data by a computing element 23. The low order data is binarized by a comparator 15 by making a normalized uniform random number generated from a random number generator 24 or a dither value from a dither generator 25 selected by a selector 22a threshold. Then, this binary data is added to the high order data of the picture signal of a remarked picture element by an adder 14, and is binarized by the comparator 18 on the basis of the binary data of the picture element binarized before the remarked picture element by an arithmetic unit 13 by making plural weighted mean values of plural picture elements in the vicinity of the remarked picture element the threshold. At that time, density is preserved by allotting a density error to be caused at the time of binarization by the adder 16 or a subtracter 17 to the adder 16, etc., to the unbinarized picture elements around the remarked picture element.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus for quantizing multi-valued image data into binary data.

[0002]

2. Description of the Related Art As a pseudo intermediate processing method used in digital copying machines, facsimile machines, etc.
An image processing apparatus such as "-210959" has been proposed. In this image processing apparatus, the pixel of interest is binarized into black or white by using the binary data of the pixel which has already been binarized in the vicinity of the pixel of interest, and the error generated when binarizing the pixel of interest is detected. An operation of adding to multi-valued data of unbinarized pixels in the vicinity is sequentially performed for each pixel.

That is, the average density is calculated using only the binary data that has been binarized, and the input multi-valued data is binarized using the average density as a threshold. Therefore, it is possible to perform binarization with a relatively small amount of processing, and since the error between the input multi-valued data and the average density that occurs when binarizing the input multi-valued data is corrected, an image with good gradation can be obtained. The binarization method has the advantage of being obtained.

[0004]

However, since the pixel of interest is binarized depending on the conditions of neighboring pixels, the following drawbacks occur. There was a problem that dots were connected in the highlight part and the image quality was degraded.

Further, in order to add the error generated at the time of binarization to the surrounding unbinarized pixels, it is necessary to have a memory for holding the error data.

[0006]

The present invention has been made for the purpose of solving the above-mentioned problems, and has the following constitution as one means for solving the above-mentioned problems. That is, n
In an image processing apparatus capable of binarizing multivalued image data represented by bit signals by at least two kinds of binarization methods for each pixel, based on the value of the upper 1 bit (l <n) of the multivalued image data. Selecting means for selecting a binarizing method; first binarizing means for binarizing the lower m bits (m <n) of the multi-valued image data by the selected binarizing method; and the lower bit. The multi-valued image data quantized by binarizing is binarized using a binarization method different from the binarization of the lower bits.
Second binarizing means for binarizing.

[0007]

In the above structure, the highlight portion stochastically quantizes the lower bits of the pixel data,
The binarization performed only with the peripheral pixel information is appropriately dispersed, and the deterioration of the image quality due to dot connection is improved.
In addition, in the case of the intermediate density where the roughness is conspicuous, by quantizing the lower bits of the pixel data with the dither threshold,
It is possible to suppress the occurrence of roughness.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment according to the present invention will be described in detail below with reference to the drawings. [First Embodiment] FIG. 1 is a block diagram of an image processing apparatus according to a first embodiment of the present invention. In FIG. 1, an input sensor unit 101 is composed of a photoelectric conversion element such as a CCD and a driving device that operates the photoelectric conversion element, and performs an operation of reading a document.
The document image data read by the input sensor unit 101 is multivalued analog image data. This multivalued analog image data is sequentially sent to the A / D converter 102, and the analog data of each pixel is converted into corresponding multivalued digital data. Next, in the correction circuit 103, correction processing such as shooting correction is performed in order to correct sensitivity unevenness of the CCD sensor and illuminance unevenness due to the illumination light source.

The multi-valued image data input to the subsequent binarization circuit 104 is quantized into binary data by a binarization method described later. The printer 105 is a printer configured by a laser beam or inkjet method,
The dot is ON / OFF controlled based on the binary data sent from the binarization circuit 104, and the image is reproduced on the recording paper. 2 and 3 show the detailed block configuration of the binarization circuit 104 in FIG.

In FIGS. 2 and 3, 1 and 2 are line memories for storing one line of binarized binary image data, and 3 to 12, 21 are D types for delaying the image data by one pixel. Flip-flop DF / F, 13 is an arithmetic unit for obtaining a weighted average value of a predetermined area from the binary image data around the pixel of interest, and 14 is quantization by adding the binarized lower data to the pixel upper bit data. An adder 15 for outputting the data is a comparator for binarizing the lower bit of the pixel data by comparing the lower bit of the pixel data with the threshold based on the threshold sent from the selector 22.

Reference numeral 16 is an adder for adding error data for density correction to the quantized data, and 17 is a subtractor for calculating a difference between the weighted average value output from the calculator 13 and the quantized data of the pixel of interest. Reference numeral 18 is a comparator for binarizing the pixel of interest by comparing the threshold value with the quantized data of the pixel of interest using the weighted average value output from the calculator 13 as a threshold, and 19 is the pixel of interest output from the subtractor 17. Error ROM for calculating error data on the basis of the difference between the quantized data of 1 and the weighted average value output from the calculator 13, and 20 is the error ROM 1
This is a line memory for storing the calculation error data from 9 for one line.

A selector 22 selects whether to output the generated random number from the random number generator 24 as a threshold value or the dither threshold value output from the dither generator 25 as a threshold value, based on the upper bit data of the pixel of interest. , 23 is an arithmetic unit for dividing the multi-valued pixel data input from the correction circuit 103 into upper bit data and lower bit data, 24 is a random number generator for generating a uniform random number for each pixel, and 25 is according to the pixel position. It is a dither generator that generates a dither threshold based on dither matrix. The principle of the binarization method in this embodiment having the above configuration will be described below with reference to FIG.

(1) shown in FIG. 4 shows the multi-value density for each pixel of the input image. In (1) of FIG. 4, f (i, j) represents multi-valued image data of the input image at the target pixel position to be binarized, and takes a value of "0 to 255" in an 8-bit configuration. Further, the image data at the pixel position above the broken line has already been binarized, and when the binarized image of the pixel of interest is performed,
After that, the same binarization process as f (i + 1, j), f (i + 2, j), ... Is sequentially performed.

FIG. 4B is a diagram showing the binarized image data. In FIG. 4B, B (i, j) is 2 of the target pixel.
The data (0 or 1) after binarization is shown. The portion surrounded by the broken line is pixel data that has already been binarized when the pixel of interest is processed, and these are used in the binarization process of the pixel of interest. FIG. 4C is a diagram showing weighting. R is an example of a weighting mask for obtaining the average density, and 5
It is represented by a matrix of × 3 size.

Here, as the weight for the unbinarized pixel, R (0,0) = R (1,0) = R (2,0) = 0 is used. Then, first, the input image data f (i, j) is converted into the upper 4 bit data.
It is divided into fH (i, j) and lower 4 bit data fL (i, j). In the highlight part where the upper data fH (i, j) is "0", the lower data is "0" or "1" with a uniform random number that values (0 to 15) are generated with the same probability. Is binarized to. Further, in the portion where the high-order data fH (i, j) is “1” and the roughness is conspicuous, the low-order data is 4 × as shown in FIG.
It is binarized using four size diamatrics as a threshold.

Further, in a portion where the dither texture is conspicuous in which the upper data is (2 to 15), the lower data fL (i, j)
Is binarized to “0” or “1” using a uniform random number as a threshold, as in the highlight part. Then, the binarized lower data BL (i, j) is added to the upper fH (i, j) to obtain the quantized data h (i, j). Next, the weighted average density m (i, j) in the vicinity of the target pixel is calculated from the pixel data that has already been binarized by the following equation (1).

[0017] The quantized data h (i, j) of the pixel of interest includes m (i, j) calculated by the above equation and the binarized correction value E assigned to the pixel of interest.
It is binarized according to the following equation (2) using (i, j).

That is, when h (i, j) + E (i, j)> m (i, j), B (i, j) = 1│h (i, j) + E (i, j) ≤m ( When i, j) B (i, j) = 0 | Equation (2) Further, the correction value of the binarization error generated at this time is also calculated by the following Equation (3) at the same time. E1 (i + 1, j) = E2 (i, j + 1) = {h (i, j) + E (i, j) -m (i, j)} / 2 ... Equation (3) E1 (i + 1, j) is a correction value assigned to the pixel (i + 1, j) next to the pixel of interest, and E2 (i, j + 1) is the pixel (i, j +) one line after the pixel of interest. This is the correction value assigned to 1).

Binarized correction value E assigned to the pixel of interest
(i, j) is the error E1 (i, j) generated when the pixel (i-1, j) that is one pixel before the pixel of interest is binarized by the above method,
Sum of error (E2 (i, j)) generated when the pixel (i, j-1) one line before the pixel of interest is binarized, E (i, j) = E1 (i, j) + E
It is 2 (i, j). By sequentially performing the above operation for each pixel, the entire image is binarized.

As described above, in the density where the regular texture such as the highlight portion is conspicuous, the lower bits are quantized by using the random number as the threshold, and in the density range where the random roughness is conspicuous, the lower bits are quantized by the dither threshold. , It is possible to prevent the dot connection in the low density area and prevent the occurrence of texture and roughness in each density area. Further, since the density of the lower bit data is the quantization probability, the density is statistically stored. In the upper bit, binarization is performed based on the peripheral pixel information. By correcting the binarization error, the density is preserved and an image excellent in gradation and resolution is obtained. Further, since the correction amount of the binarization error only needs to be stored for the upper bits, the memory can be saved. In the configuration based on the binarization principle of the present embodiment described above, the pixel of interest
The operation when binarizing (i, j) will be described.

The line memories 1 and 2 shown in FIGS. 2 and 3 store binary data of pixels which have been binarized before the pixel of interest from the comparator 18. When the pixel of interest is binarized, the line memory 2 outputs the binary data B (i + 2, j-2) one line before, and the line memo 1 outputs the binary data B (i +) two lines before. 2, j-2) is output. DF / F 3 to 12 each output data delayed by one pixel. That is, DF / F 3 is B (i + 1, j-2), DF / F 4 is B (i, j-2), DF /
F 5 is B (i-1, j-2), DF / F 6 is B (i-2, j-2), DF / F 7 is
B (i + 1, j-1), DF / F 8 is B (i, j-1), DF / F 9 is B (i-1, j-)
1), DF / F 10 is B (i-2, j-1), DF / F 11 is B (i-1, j),
The DF / F 12 outputs B (i-2, j), respectively.

The binary data from these DF / Fs 3 to 12 are input to the calculator 13. The computing unit 13 computes the weighted average density value m (i, j) indicated by the symbol A based on each weight mask from the binary data in the vicinity of the pixel of interest, and outputs it to the subtractor 17 and the comparator 18. The adder 14 adds the high-order bit data fH (i, j) and the low-order binarized data BL (i, j) binarized to "1" or "0" to obtain the quantized data h.
Output (i, j). The adder 14 outputs the upper bit data fH
Quantized data h (i, j) is output by adding the lower binarized data BL (i, j) binarized to (i, j) and "1" or "0".

The comparator 15 compares the threshold value selected by the selector 22 with the lower 4-bit data fL (i, j) to binarize it, and outputs the lower binary data BL (i, j). Selector 2
2 selects the random number threshold value from the random number generator 24 as the threshold value when the upper fH (i, j) is “0” or the value of (2 to 15), and the upper data fH (i, j) is “2”. In the case of 1 ", the dither threshold value from the dither generator 25 is selected as the threshold value.

On the other hand, the multi-valued data f (i, j) of the pixel of interest sent from the correction circuit 103 is input to the calculator 23. The calculator 23 divides the input 8-bit multivalued data f (i, j) of the target pixel into upper 4 bit data fH (i, j) and lower 4 bit data fL (i, j) and outputs them. The random number generator 24 also generates a uniform random number having a value of (0 to 15) for each pixel. The dither generator 25 generates a dither threshold based on the dither matrix according to the pixel position.

The adder 16 adds the error correction data E1 (i, j) and E2 (i, j) to the quantized data h (i, j) to obtain h (i, j) +
Output E (i, j). The subtractor 17 is h (i, j) + E (i, j)
And the difference between m (i, j) is calculated. Further, the comparator 18 uses m (i,
j () is used as a threshold value, and binarization of h (i, j) + E (i, j) is performed,
The binary data B (i, j) is output. The binary data B (i, j) from the comparator 18 is sent to the printer 105.

The binary data B (i, j) is also input to the line memory 2 and the DF / F 11 and serves as peripheral pixel information for pixels to be subsequently binarized. On the other hand, subtracter 1
H (i, j) + E (i, j) -m (i, j) output from 7 is input to the error ROM 19. The error ROM 19 stores the error data E1 (i + 1, j), E2 (i, j + 1) according to the above-mentioned equation (3).
Is output. E1 (i + 1, j) is delayed by one pixel by the DF / F 21 and assigned to pixel (i + 1, j). E2 (i, j + 1) is stored in the line memory 20 and stored in one line. The next pixel (i, j + 1)
Assigned to.

As described above, the binarization processing of the image data is sequentially performed for each pixel by repeating the series of processing. As described above, according to the present embodiment, the highlight portion stochastically quantizes the lower bits of the pixel data, so that the binarization performed only by the peripheral pixel information is appropriately dispersed. The deterioration of image quality due to dot connection is improved. Further, at the intermediate density where the roughness is easily noticeable, the lower bit of the pixel data is quantized by the dither threshold, so that the occurrence of the roughness can be suppressed.

[Second Embodiment] Next, a second embodiment according to the present invention will be described in detail with reference to FIGS. 6 and 7. The basic structure of the second embodiment is similar to that of the first embodiment shown in FIG. 1 described above, and the structure of the binarization circuit 104 is different. The detailed configuration of the binarizing circuit of the second embodiment is shown in FIGS. In the second embodiment, the same components as those in the first embodiment are designated by the same reference numerals, and detailed description thereof will be omitted.

6 and 7 are different from the first embodiment shown in FIGS. 2 and 3 in the following points. In the first embodiment, the correction value for the binarization error output from the error ROM 19, that is, E1 (i + 1, j) = E2 (i, j + 1) = {h (i, j).
The value of E2 including + E (i, j) -m (i, j)} / 2 including the code is stored in the line memory 20 for one line. This is because E2 can be negative.
Then, the output of the line memory 20 was directly input to the adder 16.

However, in the second embodiment shown in FIGS. 6 and 7, the output of the error ROM 19 is the line memory 52.
0, and only the absolute value of E2 is input to the line memory 52.
Stored in 0. The output of the line memory 520 is input to one of the arithmetic units 526, and the output of DF / F 8 is input to the other of the arithmetic units 526. Therefore, the capacity of the line memory 520 can be smaller than that of the line memory 20.

Then, the correction value E added to the pixel (i, j)
The code of 2 (i, j) is coded by the calculator 526 and sent to the adder 16. In the computing unit 526, B (i, j
If -1) = 1, a plus sign is given, and if B (i, j-1) = 0, a minus sign is attached. The value of B (i, j-1) is h (i, j-
1) + E (i, j-1) is determined by the magnitude relation of m (i, j-1), so it corresponds to the sign of E2 (i, j).

With the above configuration, the memory capacity of the line memory 520 can be saved. [Third Embodiment] Next, a second embodiment of the present invention will be described in detail with reference to FIGS. The basic configuration of the third embodiment is similar to that of the first embodiment shown in FIG. 1 described above, and the configuration of the binarization circuit 104 is different. The detailed structure of the binarizing circuit of the second embodiment is shown in FIGS. In the third embodiment, the same components as those in the first embodiment are designated by the same reference numerals, and detailed description thereof will be omitted.

The third embodiment is an application example using the error diffusion method instead of using the error ROM 19. 8 and 9 are different from the first embodiment shown in FIGS. 2 and 3 in the following points. That is, 613 is an arithmetic unit that outputs the accumulated error Err (i, j) of the pixel that has already been binarized,
Reference numeral 614 is an adder for adding a cumulative error to the quantized data, and 6
Reference numeral 15 is a comparator for binarizing the quantized data after the error correction using a predetermined value α as a threshold value, 616 is a multiplier for multiplying the binary data by a predetermined value, and 617 is the quantized data after the error correction and the binary value. It is a subtracter that calculates the difference from the converted data.

A case where the pixel of interest is binarized with the above configuration will be described. First, the input target pixel data f (i,
j) is divided into upper bit data and lower bit data by the arithmetic unit 23. The random number generator 24 outputs a random number value, and the dither generator 25 outputs a dither threshold value according to the pixel position. The selector 22 selects either a random number value or a dither threshold value based on the value of the upper bit data. The comparator 15 binarizes the data based on the selected threshold value.
The adder 14 adds the binarized lower bit data to the upper bit data and outputs quantized data h (i, j).
The configuration and operation of these are almost the same as in the first embodiment described above.

The arithmetic unit 613 also outputs the cumulative error Err (i, j) of the pixel which has already been binarized. The adder 614 adds the accumulated Err (i, j) to the quantized data h (i, j). The comparator 615 sets h (i,
j) + Err (i, j) is binarized by the following formula. If h (i, j) + Err (i, j)> α, B (i, j) = 1 If h (i, j) + Err (i, j) ≦ α, B (i, j) = 0 Comparator The binarized signal B (i, j) from the printer 615 is the printer 10
5 and also to the multiplier 616. In the multiplier 616, the binarized data B input to
The maximum value Dmax that the quantized data h (i, j) can take in (i, j)
And outputs it to the subtractor 617. The subtractor 617 calculates the binarization error e (i, j) according to the following equation and outputs it to the DF / F 11.

E (i, j) = h (i, j) + Err (i, j) -Dmax.B (i, j) where e (i, j) is the line memories 1, 2, DF / F 3 ~ 12
And becomes the source of the accumulated error data for the pixel of interest, which is delayed by each of the above and is binarized later. Cumulative error is 5 × 3
The size weighting coefficient matrix S is calculated as follows.

[0037] Hereinafter, the above process is repeated to sequentially binarize the pixels.
As described above, according to the third embodiment, as in the first embodiment, the highlight part is stochastically quantized with the lower bits of the pixel data, so that only the peripheral pixel information is used. The binarization is dispersed appropriately, and the deterioration of the image quality due to dot connection is improved. Further, at the intermediate density where the roughness is easily noticeable, the lower bit of the pixel data is quantized by the dither threshold, so that the occurrence of the roughness can be suppressed.

According to each of the embodiments described above, the highlight portion stochastically quantizes the lower bits of the pixel data, so that the binarization performed only by the peripheral pixel information is appropriately dispersed. , Degradation of image quality due to dot connection is improved. Further, at the intermediate density where the roughness is easily noticeable, the lower bit of the pixel data is quantized by the dither threshold, so that the occurrence of the roughness can be suppressed.
Moreover, since the density is statistically stored, the gradation is not deteriorated and the error is corrected only by the upper bits.
It also saves memory.

That is, the selection means for selecting the binarization method based on the value of the upper bit of the multi-valued image data and the lower bit of the multi-valued image data are selected by the selected binarization method.
By binarizing the binarizing means for binarizing and the lower bit, the binarization of binarizing the quantized multi-valued image data using a binarization method different from the binarization of the lower bit. And means.

The present invention may be applied to a system composed of a plurality of devices or an apparatus composed of a single device. Further, it goes without saying that the present invention can be applied to the case where it is achieved by supplying a program to a system or an apparatus.

[0041]

As described above, according to the present invention, the lower bit of the pixel of interest is selectively used by a binarization method having a simple and regular or random characteristic according to the density range. By quantizing, it is possible to prevent deterioration of image quality due to regular texture and also to prevent deterioration of image quality due to roughness. Furthermore, according to the present invention, the memory capacity for storing the binarization error correction amount can be saved.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of an image processing apparatus according to an embodiment of the present invention.

[Fig. 2]

3 is a block diagram showing a detailed configuration of the binarization circuit shown in FIG. 1. FIG.

FIG. 4 is a diagram showing a multi-valued image (1), a binary image (2) and a weighting mask (3) for calculating an average density for each pixel according to the present embodiment.

FIG. 5 is a diagram showing an example of a dithermatrix according to the present embodiment.

FIG.

FIG. 7 is a block diagram showing a detailed configuration of a binarizing circuit according to a second exemplary embodiment of the present invention.

FIG. 8

FIG. 9 is a block diagram showing a detailed configuration of a binarizing circuit according to a third exemplary embodiment of the present invention.

[Explanation of symbols]

 1, 2, 20, 520 Line memory 3-12, 21 D type flip-flop DF / F 613 Arithmetic unit 13, 23, 526 Arithmetic unit 14, 16, 23, 614 Adder 15, 18, 615, 425 Comparator 17 , 617 Subtractor 19 Error ROM 22 Selector 24 Random number generator 101 Input sensor unit 102 A / D converter 103 Correction circuit 104 Binarization circuit 105 Printer 617 Subtractor.

Claims (5)

[Claims] 1. An image processing apparatus capable of binarizing multivalued image data represented by an n-bit signal by at least two kinds of binarization methods for each pixel. selecting means for selecting a binarization method based on the value of n), and a first binary value for binarizing the lower m bits (m <n) of the multivalued image data by the selected binarization method. Binarizing means, and second binarizing means for binarizing the multivalued image data quantized by binarizing the lower bits using a binarization method different from the binarization of the lower bits. An image processing apparatus comprising: 2. The first binarizing means includes a random number generating means for generating a uniform random number standardized for each pixel, and two lower bits based on the random number generated by the random number generating means.
The image processing apparatus according to claim 1, further comprising a computing unit that digitizes the image. 3. The first binarization unit uses multi-valued image data as a threshold value by using dithermatrix.
The image processing apparatus according to claim 1, further comprising calculation means for digitizing. 4. The second binarizing means, an average value computing means for obtaining a weighted average value in the vicinity of the pixel of interest based on binary data of pixels already binarized before the pixel of interest. The image processing apparatus according to claim 1, further comprising: a binarization arithmetic unit that binarizes the multivalued image data quantized based on the average value calculated by the average value arithmetic unit. 5. The image processing apparatus according to claim 1, wherein the second binarizing unit binarizes the quantized multi-valued image using an error diffusion method.