JPH04230171A - Picture processor - Google Patents

Abstract

PURPOSE:To simply implement the multi-value processing at a high speed satisfying simultaneously resolution of a character picture and gradation of a photographic picture. CONSTITUTION:A picture characteristic is extracted from input picture information in which a character picture and a photographic picture exist in mixture and when the information is the character picture, no error is calculated and when the information is the photographic picture, error calculation circuits 22, 23 are activated. The error calculation circuits 22, 23 fetch a picture signal to calculate an error from only an input of a multiplexer circuit 20 and calculate an error of a preceding picture element to a noted error and an error of surrounding picture elements in a preceding line based on an error signal calculated in a prediction way and correct input picture information based thereon.

Description

[Detailed description of the invention]

0001

[Industrial Application Field] This invention is a multivalued image that satisfies both the resolution of the text portion and the gradation of the photographic portion in an image in which a text image and a gradation image such as a photograph are mixed. The present invention relates to an image processing device for performing image processing.

0002

[Prior Art] Generally, in a document image processing device that can handle not only code information but also image information, contrast image information such as characters and line drawings is fixed in a document read by a reading means such as a scanner. Simple binarization is performed using a threshold value, and image information having gradations such as photographs is binarized by pseudo gradation means such as a dither method. This is because when the read image information is simply binarized using a fixed threshold, the resolution is preserved in the text and line image areas, so no image quality deterioration occurs, but the image quality does not deteriorate in the photo image area. Since the tonality is not preserved, the resulting image has degraded image quality. On the other hand, if the read image information is subjected to gradation processing using systematic dithering, etc.
In areas of photographic images, gradation is preserved, so no deterioration in image quality occurs, but in areas of characters and line images, resolution decreases, resulting in an image with degraded image quality. That is, it is impossible to satisfy the image quality of each area having different characteristics at the same time by performing a single binarization process on the read image information.

However, the ``error diffusion method'' has been proposed as a binarization method that satisfies the gradation characteristics of photographic image areas and has better resolution for text/line image areas than the systematic dither method. There is. This "error diffusion method" is a method in which the density of the pixel of interest is multiplied by the multivalue error of surrounding pixels that have already been multivalued, multiplied by a predetermined weighting coefficient, and multivalued is performed using a fixed threshold. . FIG. 12 is a block diagram of the configuration of multivalue processing using the "error diffusion method." In FIG. 12, an input image signal read by an input device such as a scanner is supplied to one input terminal of an adder 41 via an input terminal 40. The previous line error signal is supplied from the adder 43 to the other input terminal of the adder 41 via the line buffer 42, and the input image signal with the previous line error corrected is input to one input of the adder 44 at the next stage. fed to the end. The other input terminal of this adder 44 is connected to a coefficient unit 45 for weighting error coefficient A for outputting the previous pixel error.
An input image signal in which the previous line error and the previous pixel error have been corrected is obtained from the adder 44 and supplied to the input end of the multi-value conversion circuit 46 . A large number of threshold values for multi-value conversion are generated from an external threshold generation means and supplied to the multi-value conversion circuit 46. In this multi-value conversion circuit 46, the corrected input image signal from the adder 44 is compared with multi-value threshold values Th1 to Thn, and a predetermined multi-value conversion output is outputted to an output terminal 47.

The multi-value output obtained from the multi-value conversion circuit 46 is supplied together with the corrected input image signal from the adder 44 to an error calculation circuit 48, and the difference between the two is output as an error signal. This error signal is supplied to coefficient units 49, 50, and 51, and multiplied by weighted error coefficients B, C, and D, respectively. Here, the coefficients of the coefficient units 45, 49, 50, and 51 are A=1/2, B=1/16, C=5/16, D=
It is set to 1/8. The weighted error signal from the coefficient unit 49 is added in an adder 53 to a signal obtained by delaying the weighted error signal from the coefficient unit 50 by one pixel in a delay circuit 52, and is further added to the output delay circuits 54, 55 of the coefficient unit 51. The adder 43 adds the signal delayed by two pixels. In this way, by delaying the output of the adder 43 by one line by the line buffer 42, the three-pixel error signal is supplied to the adder 41, and the input image signal is corrected.

[0005]

[Problem to be solved by the invention] The above “error diffusion method”
The method minimizes the error by diffusing the error caused by multilevel conversion of the pixel of interest to surrounding pixels and compensating for the error. Therefore, when calculating an error, after calculating the corrected pixel, it is necessary to compare and subtract from the pixel of interest, and then to perform addition and subtraction to the pixel of interest. Therefore,
It is extremely difficult to perform error diffusion processing at high speed.

Therefore, it is an object of the present invention to provide an image processing device that can perform multi-value processing that satisfies both the resolution of character images and the gradation of photographic images at high speed and easily. .

[0007]

[Means for Solving the Problems] An image processing device according to the present invention includes a multi-value conversion means for obtaining multi-value information by converting input image information including a pixel of interest into multi-value information, and a feature amount representing the characteristics of this input image information. an extraction means for extracting the error, a means for calculating an error caused by the multivalue processing by the multivalue conversion means using input image information and a feature amount, and an error obtained by the calculation means and extraction by the extraction means. The image forming apparatus is characterized by comprising a generating means for generating error correction information based on the generated feature quantity, and a means for correcting input image information using the error correction information generated by the generating means.

[0008]

[Operation] The input image information supplied to the multivalue conversion means and a determination signal indicating whether the input image is text information or gradation information such as a photograph are supplied to the prediction error calculation means, and a prediction error signal is obtained. . This prediction error signal is supplied to the previous pixel error calculation circuit and the previous line error calculation circuit, and the obtained previous pixel error and previous line error are used to calculate the error generated by the multi-level conversion means for the input image information. will be added to compensate.

[0009]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In FIG. 1, input image information read by an input device such as a scanner is supplied from an input terminal 11 to a line buffer 12 and temporarily stored therein. Input image information is read out from the line buffer 12 in synchronization with a predetermined clock, and is supplied to the delay circuit 13 and feature amount extraction circuit 14. The delay circuit 13 is used to synchronize the timing with the input signal from the line buffer 16 in the adder 15 at the next stage.

On the other hand, the feature extracting circuit 14 extracts the maximum and minimum values of image density within a (4×4) window including the pixel of interest * as shown in FIG. The extracted maximum value signal 14a and minimum value signal 14b are respectively sent to the image determination circuit 1.
7 is supplied. This image determination circuit 17 is, for example, a RAM.
The output signal 14a from the feature extraction circuit 14
, 14b are supplied as address signals, and a corresponding output is obtained. The output is set to be "1" when the input image is a text, and "0" when it is a photograph.

The input image signal delayed by the delay circuit 13 is added to the previous line error signal output from the line buffer 16, and is supplied to the adder 18 at the next stage. Adder 1
8 is also supplied with the previous pixel error signal output from the coefficient unit 19, and both are added to form a corrected image signal. This corrected image signal is supplied to the multi-value conversion circuit 20, converted into a multi-value signal, and outputted to the output terminal 21. The multi-level conversion circuit 20 is formed of, for example, a RAM, and converts the corrected image signal into this R.
It is supplied to the AM as an address signal, and a multivalued output corresponding to this address signal is read out from the RAM.

The corrected image signal obtained from the adder 18 is further supplied to a prediction error calculation circuit 22. This prediction error calculation circuit 22 is also formed of RAM, and the corrected image signal is combined with the output of the image determination circuit 17 and supplied to the RAM as an address signal. A specific example of this prediction error calculation circuit 22 is configured as shown in FIG. RAM22
a constitutes the prediction error calculation circuit 22, and the 10-bit corrected image signal from the adder 18 is input to its address input terminal A00.
~A09. Here, MS is connected to input terminal A09.
The sign bit of B is supplied, and the carry bit is supplied to input terminal A08. The RAM 22a also includes an address input terminal A10, to which "1" from the image determination circuit 17 is input.
” or “0” signal is supplied.The input terminal A10 is RA
This is the enable terminal of M22a, and as described above, when the input image is a character, "1" is output from the image determination circuit 17, and at this time, 0 is output as the error output. Terminal D0
0 to D08 are for data output, and a 9-bit error signal is output.

The prediction error signal thus obtained is supplied to a coefficient unit 19 and multiplied by a coefficient A to form a previous pixel error signal. This previous pixel error, as shown in FIG.
A previous pixel error signal is formed by multiplying by .

The output signal of the RAM 22a is also supplied to a three-pixel error calculation circuit 23. The three-pixel error means an error regarding the corresponding pixel one line before the pixel of interest * and the pixels before and after it in the error matrix shown in FIG.

FIG. 5 shows a specific example of the three-pixel error calculation circuit 23, in which the error signal from the prediction error calculation circuit 22 is sent to the coefficient multiplier 2.
31, 232, and 233, and are multiplied by weighting coefficients B, C, and D, respectively. The signal multiplied by the coefficient B is supplied to one input terminal of the adder 234, and the signal multiplied by the coefficient C is given a delay of one pixel by the delay circuit 235, and is supplied to one input terminal of the adder 236. supplied to The signal multiplied by the coefficient D is delayed by two pixels by delay circuits 237 and 238, and then supplied to the other input terminal of the adder 236. In the end, for the output signal of the coefficient unit 231, the output of the coefficient unit 232 corresponds to one pixel, and the output of the coefficient unit 233 corresponds to 2 pixels.
It appears at the output of adder 234 after being delayed by a pixel.

The weighting coefficients B, C, and D of these three pixels are B=1/16, C=5/16, and D=1/8, respectively.
These weight errors are calculated as eB-1 = B x EB, respectively, where EB is the prediction error signal.
(1)

[0017] eC0 = C×EB
(2) eD+1 = D×EB
(3) becomes. Therefore, 3 pixel error BL
E is BLE=eB-1+eC0+eD+1 (4
). Note that FIG. 6 shows the arrangement of weighted errors of each pixel corresponding to the error matrix of FIG. 4.

The three-pixel error BLE is sent to the line buffer 16 as error storage means and temporarily stored. The line buffer 16 is composed of line memories for two lines, the previous line and the current line.

By the way, the feature amount extraction circuit 14 for determining whether the input image is a character or a gradation image such as a photograph is as follows.
For example, it is configured as shown in FIG. As shown in FIG. 2, the feature extracting circuit 14 is configured to obtain the maximum and minimum values of density in each column in a (4×4) pixel region including the pixel of interest Pij.

For example, as shown in the timing chart of FIG.
The image information input to the input terminals I1 to I4 of the selector 141 in units of four pixels in the column direction in synchronization with the output of the counter 142 is composed of 8 bits per pixel. The image information input to the selector 141 is sent to the comparator 143
-146 and compared in the column direction in units of 4 pixels to determine the maximum density and minimum density in that column.

The comparators 147 and 148 at the next stage are the comparators 14
The outputs of 3-146 are respectively inputted at the timing of FTR1, and the maximum value and minimum value respectively determined in the column direction are determined. The maximum value 14a and the minimum value 14b are output at the timing of FTR2, and the difference between them is determined by the subtracter 149 at the next stage. Therefore, the subtracter 149 is a circuit included in the image determination circuit 17 of FIG.

In the adder 15 of FIG. 1, if the input image signal is corrected using the output of the three-pixel error calculation circuit 23, the resolution is maintained since there is no correction amount for the text area, and the resolution is maintained for the photographic area. Since the error can be compensated for using the multi-value conversion error of surrounding pixels, multi-value conversion processing can be performed with good gradation.

As described above, in the present invention, by setting the output value of the multi-value converting circuit 20 to a value of 2n-1+1 (n is an integer), the prediction error calculation means 22 uses the corrected pixel data obtained from the adder 18. This method predicts the output value and calculates the pixel error of interest. By setting the output value to 2n-1+1 (n is an integer), only the upper bits of each output become significant, and the lower bits become 0. From this, the error between the corrected pixel data pe and the output level value can be calculated by simply comparing the upper predetermined bits of the corrected pixel data, excluding the lower bits.

As shown in FIG. 9, if the sign bit of the corrected pixel data pe is 1, the corrected pixel data pe is a negative value (pe<0), so the output value is 0, and the error is due to the corrected pixel data. Output pe as is. Next, look at the carry bit of the corrected pixel data and if it is 1, the corrected pixel data pe is greater than ff(heX), so the output value is f
f, and the error is corrected pixel data pe-100 (heX)
Just output +1.

Next, if pe is 0≦pe<100 (heX), if 1 is set in the (8-n) bit of the corrected pixel data, the data below the (8-n) bit is treated as a negative value. Output. Figures 10 and 11 show values from binary to 9.
Shows data up to the value. Here, the case of five values will be explained.

The corrected pixel data is 10 bits, the MSB is a sign bit and the next bit is a carry. Output value is 5
value, since n=3, each threshold value is e0(
hex), a0 (hex), 60 (hex), 20 (h
ex), and the output values are ff(hex), c0(hex
), 80 (hex), 40 (heX), 00 (heX)
It is. For example, if the corrected pixel data is ae (heX), the output value will be c0 (heX), and the error will be (ae
-c0)=3ee. (8-3) bits, i.e. 5
Since the bit is 1, a value of 4 bits or less is output as a negative number. Correction data is 9e (heX)
, the output value will be 80 (heX) and the error will be (
9e-80)=01e(hex). Since the 5th bit is 0, this means that a value of 4 bits or less is output as a positive value.

To explain more generally, if the magnitude of a digital signal is expressed in the range 0x00 to Px100, the output level at which it is easy to calculate the error is (0x00, 0x80, 0
×100), (0×00, 0×40, 0×80, 0×A
0, 0x100), where the lower two bits are 0. When this output level is used, it is only necessary to compare the upper few bits and output the lower bits as they are. Such an output level is equal to 0x100 divided by two. If m=1, the output level is 3 levels, m=2
If so, it becomes 5 values, if m=3, it becomes 9 values, and so on, resulting in 2m+1 values. If m=n-1 is substituted, the value becomes 2n-1+1.

As described above, by predicting the output value for the corrected pixel data, the sign bit of the corrected pixel data,
By detecting the carry bit, the (8-n)th bit, the error can be calculated quickly and easily.

[0029]

As described above, according to the present invention, when image information including a pixel of interest is multivalued, errors can be reduced only by using corrected image information on the input side without using the output of the multivalue conversion means. Since the calculation is performed, there is no need for subtracters and comparators like in the past, and multi-value processing can be performed at high speed.
It is possible to provide an image processing device that satisfies both the resolution of character images and the gradation of gradation images such as photographs.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of the present invention.

FIG. 2 is a diagram of a 4×4 pixel window for feature extraction.

FIG. 3 is a specific block diagram of an error calculation circuit.

FIG. 4 is an error matrix diagram between a pixel of interest and surrounding pixels.

FIG. 5 is a block diagram of a three-pixel error calculation circuit in FIG. 1;

FIG. 6 is a layout diagram of weight errors corresponding to FIG. 4;

FIG. 7 is a specific block diagram of the feature amount extraction circuit in FIG. 1.

FIG. 8 is a timing chart of the circuit of FIG. 7;

FIG. 9 is a diagram showing the relationship between corrected pixel data, output, and error according to the present invention.

FIG. 10 is a diagram showing the relationship between output value and error.

FIG. 11 is a diagram showing the relationship between output values and errors.

FIG. 12 is a block diagram of a conventional error diffusion processing circuit.

[Explanation of symbols]

12, 16...Line buffer, 13...Delay circuit, 14...
Feature extraction circuit, 15, 18... Adder, 17... Image determination circuit, 19... Coefficient unit, 20... Multi-value conversion circuit, 22... Prediction error calculation circuit, 23... 3-pixel error calculation circuit.

Claims (1)

[Claims] 1. Multi-value conversion means for obtaining multi-value information by converting input image information including a pixel of interest into multi-value information, extraction means for extracting feature quantities representing features of the input image information, and said multi-value conversion means. means for calculating an error caused by multivalue processing using input image information and a feature quantity; and error correction information based on the error obtained by the calculation means and the feature quantity extracted by the extraction means. An image processing apparatus comprising: a generating means for generating error correction information; and a means for correcting input image information using error correction information generated by the generating means.