JPH0646263A - Picture processing method - Google Patents

Abstract

PURPOSE:To provide the picture processing method in which a characteristic of a picture element is discriminated with high accuracy while preventing deterio ration in the gradation. CONSTITUTION:A multi-value input signal corresponding to each picture element obtained by reading an original picture is inputted respectively to a gamma correction section 1 and a picture element characteristic extract section 12. The gamma correction section 1 applies prescribed correction to the input signal to improve the linearity of the area gradation and the picture element characteristic extract section 12 extracts the characteristic of the picture element based on the input signal.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method,
In particular, the present invention relates to an image processing method for processing using an area gradation method such as an error diffusion method.

[0002]

2. Description of the Related Art An error diffusion method is used for gradation expression when outputting multivalued image data to a binary printer. This method is an excellent method that achieves both gradation and resolution, but it also extracts the features of the original image and controls the error diffusion parameters according to the features to perform more accurate and high-quality image processing. The treatment method is disclosed in JP-A-63-214073.

[0003]

In the conventional image processing method as described above, the original image is read, and after A / D conversion of the original image, the correction circuit corrects the sensitivity unevenness of the CCD sensor and the illumination unevenness due to the illumination light source. Shading correction and the like are performed. Then, the feature extraction unit extracts the feature (edge, density, etc.) of the pixel of interest based on the corrected data, and the feedback amount of the error is controlled thereby.

However, in such an image processing method, when outputting to a printer or CRT having a low dot density resolution, the dot density becomes high with high density data,
Gradation of dots or the like deteriorates the gradation in the shadow region more than the theoretical value.

On the other hand, if the image data is corrected to prevent the deterioration of the gradation and the feature of the pixel is extracted based on the corrected data, the feature determination with low accuracy is performed.

The present invention has been made in order to solve the above problems, and provides an image processing method capable of preventing deterioration of gradation and accurately discriminating pixel features. With the goal.

[0007]

According to an image processing method of the present invention, by sequentially reading density data of continuous pixels forming an image and applying an area gradation method to the read density data, In the image processing method for obtaining multi-valued image data from, the predetermined density is added to the read density data in order to improve the linearity of the area gradation property, and the pixel data is corrected based on the density data before the correction. The feature is that the feature is extracted.

[0008]

According to the present invention, the read density data is subjected to the predetermined correction, and the characteristics of the pixel are extracted from the density data before the correction.

[0009]

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

The configuration and the function of each block will be described below with reference to the drawings. First, the image signal converted from the original image into multi-valued (M-valued) data by an image reader (IR) or the like is corrected in advance by the γ correction unit 1 for the non-linearity of gradation expression due to dot collapse or the like. This correction is performed when dots are printed at too high a density when outputting N-valued data to a printer or CRT due to the characteristics of the output device. Is to prevent this.

FIG. 2 is a diagram showing the characteristics of the ROM table used for this γ correction. In the figure, the broken line shows the ideal characteristic, the solid line shows the characteristic of the actually output data, and the dashed-dotted line shows the γ correction characteristic. That is, even if density data of 256 gradations is obtained in the input image, if the dots are printed in accordance with the output characteristics as they are in the output characteristics, the dots are separated from each other in the high density portion. Due to the close contact, the gradation characteristics cannot be reproduced. Therefore, the processing of applying the dynamic range conversion is performed without using the data of the portion indicated by D in the figure (high density data area). By performing the dynamic range conversion in this way, the 8-bit data of the obtained image data becomes substantially 7-bit data, and the deterioration of about 1-bit data cannot be avoided.

Therefore, in FIG. 1, the output of the γ correction unit 1 is deteriorated in accuracy as image data. Therefore, outputting the data to the pixel feature extraction unit 12 using the data cannot extract the pixel feature with high accuracy. Therefore, in this embodiment, the image data (multi-valued input signal) before being input to the γ correction unit 1 is directly input to the pixel feature extraction unit 12, which enables highly accurate extraction.

Next, the γ-corrected data in the γ-correction unit 1 is input to the error-correction unit 2 where an N value (M ≧ N).
The error component caused by the conversion is corrected by using the error diffusion method, and further converted into an N-value signal in the N-value conversion processing unit 3.

Here, a method of setting a threshold value used for the N-value conversion process will be described.

The recording system has multi-valued levels D _i = i / (q-
1), (i = 0, 1, 2, ..., Q-1: q is the number of gradations), the target pixel Fm, n whose error is corrected is set to a threshold value T _i for _i. = (D _i −D ₀ ) /
In comparison with 2, the multilevel output level Pm, n is determined by the following equation.

Pm, n = D _i : Fm, n ≧ T _i Pm, n = D _{i −1} : T _{i −1} ≦ Fm, n <T _i Here, N = 4, that is, multi-valued four-value output A method of determining the multilevel halftoning output level will be described with reference to FIG.

Here, the output value is D ₃ = 255, D ₂ = 1
Given 75, D ₁ = 85 and D ₀ = 0, the thresholds are T ₃ = 128, T ₂ = 85 and T ₁ = 43 based on the above equation.

By setting the threshold value in this way, the range of input image values that the multilevel halftoning output level Pm, n can take is as follows.

Pm, n = 255: 128 ≦ Fm, n ≦ 255 Pm, n = 175: 85 ≦ Fm, n <128 Pm, n = 85: 43 ≦ Fm, n <85 Pm, n = 0: 0 ≦ Fm, n <43. As a result, the range of the absolute value of the N-valued error Em, n is: Pm, n = 255: 0 <| Em, n | ≦ 128 Pm, n = 175: 0 <| Em, n | <42 Pm, n = 85: 0 <| Em, n | <42 Pm, n = 0: 0 <| Em, n | <42. From this result, the N-value conversion error E is detected for the image data in which all the density data levels appear evenly.
It is shown that the probabilities of occurrence of m and n differ depending on the output value. As a result, the error for each pixel is dispersed, and the pseudo gradation generated when the input image value Fm, n is close to the multi-value output level Pm, n with respect to the halftone image having a large density change such as a photographic image. Can be suppressed.

In the figure, the solid line corresponds to the output values of D _{0 to} D ₃ , and the broken line corresponds to the threshold values T _{1 to} T ₃ . For example, if the input image value is 255
And the threshold value T ₃ or 128, the value of the multi-valued output level Pmn is D ₃ or 2
The value is 55.

FIG. 4 shows an N-value conversion processing unit 3 for N (4) values.
It is a circuit diagram showing the specific content of.

This circuit converts 256-value 8-bit data into 4-value 2-bit data.

In the figure, the input M-value image data is input to each terminal A of the comparators 21, 22 and 23. Data “128” is input to the terminal B of the comparator 21, data “85” is input to the terminal B of the comparator 22, and the comparator 2
Data “43” is input to the terminal B of No. 3. On the other hand, the outputs of the comparators 21, 22, and 23 are input to the terminals a, b, and c of the decoder 24, respectively. In the comparator, the data input to the terminal A is compared with the data input to the terminal B. When A> B, the data “1” is output to the decoder 24, and when A ≦ B, The data “0” is output to the decoder 24. FIG. 5 shows a decoder truth table of the decoder 24.

By using such a circuit configuration, the input image data of 256 gradations is output as 4-valued data.

Next, based on the error-corrected image data output from the error correction unit 2 and the N-valued signal output from the N-valued processing unit 3, the N-valued error when N-valued Is calculated in the error calculation unit 4.

FIG. 6 is a diagram showing a specific circuit configuration of the error calculating section 4 in the case where a four-valued process in which N is 4 is performed.

In this circuit, specifically, N (4) valued error = Fm, n-Pm, n is calculated.

In the figure, error-corrected image data Fm, n
Is input to the terminal A of the subtractor 32. Data "0", "85", "175", and "255" data are input to the multiplexer 31, and based on the 2-bit data D0 and D1 shown in FIG. 4, based on the truth table of FIG. Corresponding to the input data (X1-X
4) is selected. The data corresponding to the selected terminal is input to the terminal B of the subtractor 32. In the subtractor 32, the difference between the data input to the terminal A and the data input to the terminal B is obtained. Then, the subtractor 32
Based on the result of the difference, calculation error data including a sign indicating whether the difference value is positive or negative and a calculation error absolute value indicating the absolute value of the difference is output.

On the other hand, the multi-valued input signal before being input to the γ correction unit 1 is also input to the pixel feature extraction unit 12 and the features (characters / photographs, edges, etc.) in pixel units are extracted. As this extraction method, a known technique such as a MAX-MIN detection circuit or a first-order differential filter may be used.

FIG. 8 is a diagram showing a specific structure of the MAX-MIN detection circuit. In the figure, the pixel data in the peripheral area of the pixel of interest is the MAX detection circuit 41 and MI.
It is input to the N detection circuit 42. The MAX detection circuit 41 is
The MIN detection circuit is a circuit that detects the maximum pixel value MAX in the peripheral area of the target pixel, and the MIN detection circuit is a circuit that detects the minimum pixel value MIN in the peripheral area of the target pixel. The pixel value MAX output from the MAX detection circuit 41 is input to the terminal A of the subtractor 43, and the pixel value MIN output from the MIN detection circuit 42 is input to the terminal B of the subtractor 43. Subtractor 43
At, the difference between the pixel value input to the terminal A and the pixel value input to the terminal B is calculated, and the difference value (MAX-
MIN) is output.

FIG. 9 determines the characteristics of the original in pixel units based on the maximum pixel value MAX, the minimum pixel value MIN, and the difference value (MAX-MIN) obtained by the circuit of FIG. FIG. 6 is a diagram showing characteristics of a basic type of document for the purpose of FIG.

That is, according to the values of the maximum pixel value MAX and the minimum pixel value MIN, it is determined that the original is a character area, a photograph area, a character / photo intermediate area, a dark character / photo intermediate area or a light character / photo intermediate area. can do.

The calculated error data calculated by the error calculation unit 4 is responsive to the output of the pixel feature extraction unit 12 to generate the first error data.
The gain calculation unit 5 performs a gain calculation process according to a rule according to the feature of the pixel of interest. At this time, as shown in FIG. 6, it is known whether the calculation error data is a positive component or a negative component. Therefore, by performing gain calculation by dividing these positive and negative components, highly accurate control is performed. . Specifically, this rule is as follows: (1) If the pixel of interest is a character region, it is preferable to prioritize the resolution. Therefore, the gain error is 0, and the calculated error data for the positive component and the negative component are completely in error diffusion processing. Avoid giving feedback.

(2) If the pixel of interest is a photographic region, it is better to attach importance to gradation, so that gain = 1 is set, and the calculated error data for both positive and negative components is completely fed back to the error diffusion process.

(3) When the pixel of interest is in the character / photo intermediate region, it is preferable to balance the contrast and gradation, so that the gain error is set to 0.5 and the calculation error data is reduced to half for both positive and negative components. It is fed back to the error diffusion process.

(4) When the pixel of interest is a dark character / photo intermediate region, it is preferable that the high-density portion emphasizes the resolution and the low-density portion emphasizes the gradation. Therefore, the positive component gain of the calculation error data is obtained. = 0.5, negative component gain = 0, and as a result, only the positive component error data is fed back to the error diffusion processing.

(5) When the pixel of interest is a light character / photograph intermediate region, it is preferable to emphasize the gradation in the high density portion and emphasize the resolution in the low density portion. = 0, negative component gain = 0.5, and as a result, only the negative component error data is fed back to the error diffusion processing.

By doing so, it is possible to perform an appropriate error diffusion process according to the characteristics of the original of the pixel of interest, and it is possible to obtain a high-quality and high-quality image.

Next, an attribute flag indicating the feature of the image extracted by the pixel feature extraction unit 12 is added to the stored error data output from the first gain calculation unit 5. This is done in order to clarify what kind of original the error was caused by the attribute flag and to perform highly accurate error correction in the subsequent processing.

The stored error data itself consists of 8-bit data, and is treated as 9-bit data as indicated by E in FIG. 10 by using 1-bit sign bit. However, even if the accuracy of the image data is originally 8 bits, the accuracy of the data is degraded by about 1 to 2 bits due to the shaling correction and other processing as described above. About 7 bits including the sign bit is sufficient. Therefore, in the 8-bit memory area, the remaining 1
It is possible to assign the above attribute flags to the bits. The attribute flag is calculated by the following formula.

Attribute flag = 1: Th ≦ MAX-MIN (character image area) Attribute flag = 0: Th> MAX-MIN (photo image area) (Th: threshold value) By defining the attribute flag in this way, The error data with an attribute flag in the example of FIG. 10 represents the data of the character image area, and the sign bit is stored in the last 8th bit data.

FIG. 11 is a diagram showing a specific example of a circuit in which the first gain calculating section 5 and the attribute flag adding section 6 are combined.

As shown in the figure, such an operation has one-to-one correspondence with the output with respect to the condition. Therefore, by storing all the combinations in the ROM table 51,
It can be easily achieved. The ROM table 5
The sign input to 1 and the absolute value of the calculation error are shown in FIG.
In the case of the example, it corresponds to the portion indicated by the data E.

The stored error data with the attribute flag output from the attribute flag addition unit 6 is held in the error storage unit 7 for several lines, and further, by the attribute flag separation unit 8.
It is separated into error data and attribute flags. Then, the attribute flags separated by the attribute flag separation unit 8 are input to the area feature extraction unit 11, where the area-based features are identified based on the features of the attribute flag of the pixel of interest and the features of the attribute flags of the surrounding pixels. To be extracted.

This extraction operation will be described below by showing a specific example. As shown in FIG. 13, the error data with the attribute flag is held in the FIFO memory of the error storage unit 7 for several lines, the output thereof is separated by the attribute flag separation unit 8, and the data of the attribute flag is an area formed by a digital filter. It is input to the feature extraction unit 11. FIG. 12 shows the specific contents of this digital filter. That is, the feature quantities of the attribute flags of the pixel of interest and its peripheral pixels are weighted by this attribute flag weighting digital filter, and the result is output as multi-valued (0 to 33) as the area attribute determination result. As is clear from FIG. 12, the weighting of this digital filter is larger as it is closer to the pixel of interest (corresponding to the circled coefficient), and erroneous discrimination of area features is suppressed. Since the attribute flag "1" indicates a character image area, the closer the area attribute determination is to 33, the closer the original image is to the character image area, and the closer the value is to 0, the closer to the photographic image area. There is.

On the other hand, the error data separated by the attribute flag separation unit 8 is weighted to the error data of the peripheral pixels by the error weighting digital filter of the coefficient corresponding to each attribute based on the above area attribute discrimination result. To be done. The selection criteria of this digital filter are as follows: 21 ≦ area attribute determination result → digital filter coefficient = character area 11 ≦ area attribute determination result <21 → digital filter coefficient = character / photo intermediate area 0 ≦ area attribute determination result <11 → Digital filter coefficient = for photo area. A specific example of the coefficients of this weighted digital filter is shown in FIG.

That is, FIG. 14 shows specific coefficients of the error filter for the photograph area. Note that * is a portion corresponding to the pixel of interest, and in this way, in the photographic region, the error data of a wide area around the pixel is considered in order to emphasize the gradation.

FIG. 14 shows an example of specific coefficients of the character / photo intermediate area error filter. Here, since it is the character / photo intermediate region, error data of peripheral pixels in a narrower range than the error filter for the photo region is considered in order to balance gradation and resolution.

FIG. 14 shows an example of specific coefficients of the character area error filter. Here, since it is a character area, only the error data of the peripheral pixels in a very narrow range is considered in order to emphasize the resolution.

FIG. 15 is a block diagram showing a specific structure of the error weighting filter 9.

In the figure, the error data output from the attribute flag separating unit 8 is the photographic region error weighting filter 6
1, the character / photo intermediate region error weighting filter 62 and the character region error weighting filter 63 are input and filtered, and their respective outputs are input to the filter selection circuit 64. The area selection result output from the area feature extraction unit 11 is input to the filter selection circuit 64, and the output of one of the weighting filters is selected based on the result, and the content thereof is the second gain calculation unit. It is output to 10.

Next, the error data for correction calculated by the error weighting filter 9 is subjected to error gain calculation in the second gain calculating section 10 based on the area attribute discrimination result by the output of the area feature extracting section 11. Be done. That is, if the pixel of interest is included in the character image area in area, simple N-value conversion is performed with the gain = 0 and the correction error data set to zero. On the other hand, if the pixel of interest is included in the character / photo intermediate image area areawise, gain =
The error diffusion process is performed by setting 0.5 and using only half the correction error data. Further, if the target pixel is included in the photographic image area in an area, the gain is set to 1, and the error diffusion process is performed using all the correction error data.

In summary, the gain calculation at the time of storing the error is performed by the feature of each pixel of interest, while the output of the error weighting filter for calculating the correction error data of the pixel of interest is calculated by the error data of the peripheral pixel. The gain calculation is performed based on the area attribute discrimination result that also takes the pixel characteristics into consideration. By doing so, it is possible to perform error diffusion processing by highly accurate area discrimination with few misjudgments.

FIG. 16 is a diagram showing a specific configuration of the second gain calculation section 10. In the figure, the output of the error weighting filter and the area attribute discrimination result are input to the ROM table 71, and the error data calculated based on the gain is output to the error correction unit 2.

In this way, the error correction unit 2 corrects the error based on the error data output from the second gain calculation unit 10.

In the above embodiment, the error diffusion method is applied, but the area gradation method can also be applied to the image processing such as the dither method.

[0057]

As described above, according to the present invention, since the read density data is subjected to the predetermined correction and the characteristics of the pixel are extracted from the density data before the correction, the image processing with high image quality and high accuracy can be performed. Is possible.

[Brief description of drawings]

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

FIG. 2 is a graph showing correction characteristics in the γ correction unit 1 of FIG.

FIG. 3 is a diagram showing the contents of 4-valued processing shown as a specific example in the N-valued processing unit 3 of FIG.

FIG. 4 is a diagram showing a specific configuration of an N-value conversion processing unit 3 corresponding to the specific example of FIG.

5 is a decoder truth table of the decoder shown in FIG.

FIG. 6 is a diagram showing a configuration of a specific example of an error calculating section 4 in FIG.

7 is a multiplexer truth table for the multiplexer of FIG.

FIG. 8 is a diagram showing specific contents of a pixel feature extraction unit 12 of FIG.

9 is a diagram showing the characteristics of an image area for determining the characteristics of a document image based on the output of FIG.

10 is a diagram showing a method of generating error data with an attribute flag in the attribute flag addition section of FIG.

11 is a diagram showing a specific configuration of an attribute flag adding section 6 of FIG.

12 is a diagram showing a specific example of coefficients of an attribute flag weighting digital filter used in the area feature extraction unit of FIG.

13 is a diagram showing a specific configuration of an error storage unit 7, an attribute flag separation unit 8 and an area feature extraction unit 11 in FIG.

14 is a diagram showing a specific example of an error filter corresponding to each area characteristic of the error weighting filter 9 of FIG.

15 is a diagram showing a specific configuration of the error weighting filter 9 of FIG.

16 is a diagram showing a specific configuration of a second gain calculation section 10 in FIG.

[Explanation of symbols]

 1 γ correction unit 2 error correction unit 3 N-value conversion processing unit 4 error calculation unit 5 first gain calculation unit 6 attribute flag addition unit 7 error storage unit 8 attribute flag separation unit 9 error weighting filter 10 second gain calculation unit 11 area Feature Extraction Unit 12 Pixel Feature Extraction Unit In the drawings, the same reference numerals indicate the same or corresponding parts.

Claims (1)

[Claims] 1. An image processing method for obtaining multi-valued image data from an image by sequentially reading density data of continuous pixels forming an image and applying an area gradation method to the read density data. In the image processing, the image data is characterized by adding predetermined correction to the read density data in order to improve the linearity of the area gradation property, and extracting the feature of the pixel based on the density data before the correction. Method.