JPH09205558A - Image processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To sufficiently eliminate a pseudo medium tone pattern and to excellently restore gradation while suppressing a fogged image in the case of estimating original gradation information from image data subjected to limited gradation. SOLUTION: Image data are received by an image input section 1 and stored in an image storage section 2. A feature value computing section 11 reads an image of a prescribed area including an attentional picture element from the image storage section 2 and calculates its featured value for each picture element in the vicinity areas set around the picture element of interest. A weight setting section 12 sets a weight of each picture element based on the featured value obtained by the featured value arithmetic section 11. A mean value calculation section 13 conducts arithmetic mean processing based on the weight and each picture element set by a weight setting section 12 to obtain a decoded value of the picture element of interest. The obtained decoded value is fed sequentially to an image output section 3, from which the value is outputted.

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 processing a digitized image, and in particular, an image for restoring original image by estimating original tone information from the image subjected to limited tone processing. The present invention relates to a processing device.

[0002]

In order to reduce the capacity of the storage device and the transmission time when storing / transmitting a digital halftone image, 256
An image expressed in multiple gradations such as values may be converted into a small number of gradations such as binary or quaternary by a pseudo halftone process such as a dither method or an error diffusion method. However, such a pseudo-halftone image may generate a pattern called moire, which is not included in the original image, when the image is enlarged or reduced. Further, when using an output device capable of multi-value recording at the time of output, there is a problem that the performance of the output device cannot be utilized. Therefore, as one of means for solving these problems, a method of restoring the original gradation information from the pseudo halftone image has been studied.

Conventionally, as a method for restoring an original image from an image subjected to limited gradation processing, for example, Japanese Patent Publication No. 5-1146.
As described in Japanese Patent Publication No. 5 and the like, it is most common to smooth an image by using a fixed size aperture. However, in such a method, if the opening is large, even an area where density changes, such as an edge, is uniformly smoothed, and the image is blurred. Further, if the aperture is made small, a pseudo halftone pattern remains, or the number of gradations after restoration becomes insufficient. For example, even if a binary image is processed with an opening having a size of 3 × 3, it can be restored to at most 10 values.

Therefore, as disclosed in Japanese Patent Publication No. 6-24006, Japanese Patent Publication No. 4-31466, Japanese Patent Publication No. 2-76370, etc., a plurality of apertures are prepared and the apertures are opened under various conditions. As described in the selection method, IEICE Technical Research Report, IE93-13, “Estimation of Halftone Image from Pseudo Halftone Image”, etc., a regression line is obtained from the average value of the neighborhood area and restored. The method of doing is proposed.

However, in the method of selecting a plurality of apertures, a small aperture is not selected and the pseudo-halftone pattern is not removed due to an edge or fluctuation of the pseudo-halftone pattern, or the image quality is abruptly changed at the switching portion of the plurality of apertures. Due to the change, the image may become unnatural. Further, since the smoothing is basically performed by the opening, there is a problem that the number of gradations after restoration cannot be sufficiently obtained. Further, the method using the regression line is an estimation by a straight line, so that it is difficult to follow a sharp change in density, and the reconstructed values are separately obtained from the regression lines in the 8 or 4 directions around the pixel of interest, and the average is taken. Blurring may occur due to averaging.

[0006]

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned problems of the existing method, and when estimating the original gradation information from the limited gradation-converted image data, the pseudo halftone is used. An object of the present invention is to provide an image processing device capable of sufficiently removing a pattern and capable of restoring good gradation while suppressing blurring of an image.

[0007]

According to a first aspect of the present invention, in an image processing apparatus, a characteristic amount calculating means for obtaining a characteristic amount of each pixel in an input image, and a characteristic obtained by the characteristic amount calculating means. Weighted value setting means for obtaining a weighted value for the pixel of interest and its neighboring pixels based on the amount, and mean value calculation means for obtaining a weighted average value from the pixel values and weighted values corresponding to the pixel of interest and its neighboring pixels It is characterized in that each pixel in the obtained image is used as a pixel of interest, and gradation restoration processing is performed by the feature amount calculation means, the weight value setting means, and the average value calculation means.

According to a second aspect of the present invention, in the image processing apparatus according to the first aspect, the weight value setting means is based on a threshold value determined in advance according to the relative position between each neighboring pixel and the target pixel. Then, the weight value is obtained based on the comparison result of the feature amount difference between the target pixel and the neighboring pixel obtained by the feature amount calculating means and the threshold value.

According to a third aspect of the present invention, in the image processing apparatus according to the first aspect, the weight value setting means has a value predetermined according to the relative position between each neighboring pixel and the target pixel, It is characterized in that the weighted value of each pixel is obtained by an operation using the feature amount of each neighboring pixel obtained by the feature amount calculating means.

According to a fourth aspect of the present invention, in the image processing apparatus according to the first aspect, the feature amount computing means is based on a second derivative value of the input image or a value obtained by smoothing the second derivative value. It is characterized in that an output value is generated in.

According to a fifth aspect of the present invention, in the image processing apparatus according to the first aspect, the feature amount computing means is based on a first derivative value of the input image or a value obtained by smoothing the first derivative value. It is characterized in that an output value is generated in.

[0012]

FIG. 1 is a block diagram showing an embodiment of an image processing apparatus according to the present invention. In the figure, 1 is an image input unit, 2 is an image storage unit, 3 is an image output unit, 4 is a gradation restoration processing unit, 5 is a control unit, 11 is a feature amount calculation unit, 12 is a weight value setting unit, and 13 is This is an average value calculation unit. Image input unit 1
Inputs the transmitted or accumulated image data. The input image data is an image represented by limited gradation. The image storage unit 2 stores the image data input from the image input unit 1. The image output unit 3 outputs a processing result. The gradation restoration processing unit 4 performs gradation restoration processing. The control unit 5 controls each unit.

The tone restoration processing unit 4 has a feature amount calculation unit 11, a weight value setting unit 12, and an average value calculation unit 13. The feature amount calculation unit 11 calculates the feature amounts of the pixel of interest and its neighboring pixels. The weight value setting unit 12 sets the weight value of the target pixel and each neighboring pixel based on the feature amount of each pixel calculated by the feature amount calculation unit 11. The average value calculation unit 13 performs a weighted averaging process on the basis of the pixel values of the target pixel and neighboring pixels and the weight values corresponding to each pixel set by the weight value setting unit 12 to obtain the restored value of the target pixel. .

In FIG. 1, image data represented by limited gradation is input from the image input unit 1, stored in the image storage unit 2, and read out to the gradation restoration processing unit 4. In the gradation restoration processing unit 4, the read image data is input to the feature amount calculation unit 11, and each pixel is sequentially set as a target pixel, and the feature amount of the target pixel and its neighboring pixels is calculated. This feature amount is sent to the weight value setting unit 12, and the weight value of the target pixel and each neighboring pixel is calculated based on the value. The pixel value and weight value of the pixel of interest and each neighboring pixel are calculated by the average value calculation unit 1
Sent to 3. The average value calculation unit 13 obtains a weighted average value from the plurality of pixel values and the weighted value, and outputs the weighted average value as the restored value of the pixel of interest. The restoration values obtained by the tone restoration processing unit 4 are sent to the image output unit 3 and output in the order of processing.

FIG. 2 is a flowchart showing an example of the processing flow of the gradation restoration processing unit 4 in the embodiment of the image processing apparatus of the present invention. In FIG. 2, the width and height of the image are W pixels × H lines, and the upper left of the image is (0,
0) and the lower right is (W-1, H-1). In S21, the target pixel is set to the upper left pixel of the input image, and the gradation restoration process is started.

In step S22, the image data of p × q pixels centering on the pixel of interest is read from the image storage unit 2 into the feature amount calculation unit 11. The area of the image data of p × q pixels to be read is necessary for calculating the feature amount of each pixel in the neighboring area when the area including the target pixel and the neighboring pixels around the target pixel is set as the neighboring area. It must be large enough to contain pixels. For example, if 5 × 5 pixels centered on the pixel of interest are used as the neighborhood area and the secondary differential value due to the opening of 9 × 9 pixels shown in FIG. q = 13. Feature calculation unit 1
In 1, the feature amount is calculated and output for each pixel in the neighborhood area.

In S23, the plurality of feature amounts obtained in S22 are transferred to the weight value setting unit 12, and the weight value setting unit 12
The weighted value for each pixel is calculated with. The method of setting the weight value will be described later.

In step S24, the weighted value for each pixel obtained in step S23 is transferred to the average value calculation unit 13, and a weighted averaging operation is performed from each pixel value and the weighted value to calculate the restored value of the pixel of interest. Then, in S25, the restoration value of the pixel of interest obtained in S24 is sent to the image output unit 3 and sequentially output to a multi-value printer or the like.

In S26, it is determined whether or not the process for one line of pixels has been completed. If the line is being processed, the pixel of interest is changed to the pixel to the right of the current pixel in S27, Return to S22. If the processing of the line is completed, that is, if the current pixel of interest is the rightmost edge,
In S28, it is determined whether the processing has been performed up to the final line. If it is not the final line, the leftmost pixel one line below is taken as the pixel of interest in S29, and the process returns to S22. When the processing of the final line is completed, the processing of the gradation restoration processing unit 4 is completed.

An example of the operation of the embodiment of the image processing apparatus of the present invention will be described below based on a specific example. In the following description, the case where binary data is restored to 256 values will be described as an example. However, even when data of three or more values is input as input, and when the gradation of output data is other than 256 values, It can be processed similarly.

FIG. 3 is an explanatory diagram of an example of the neighborhood area.
In FIG. 3, the pixel indicated by the thick frame at the center is the pixel of interest. Here, the 5 × 5 pixel area shown in FIG. 3 is set as the neighboring area. Here, the pixel values of the respective pixels in the 5 × 5 pixel neighborhood area will be referred to as p ₀₀ to p ₄₄ .

FIG. 4 is an explanatory diagram of an example of the characteristic amount calculation. In the figure, pixels surrounded by thick lines are pixels for calculating the feature amount. The coefficient values are shown with different hatching. The feature amount calculation unit 11 uses the pixel values of the surrounding 9 × 9 pixels for each pixel in the 5 × 5 pixel area shown in FIG. 3 to generate the coefficient matrix shown in FIG. 4 (A) or (B). Perform the calculation used. In the calculation using the coefficient matrix shown in FIG. 4A, it is possible to perform the secondary differential processing according to the distance from the pixel of interest. The coefficient can be set, for example, to 8 for a blank pixel and -1 for a pixel hatched to the lower right. FIG.
In the calculation using the coefficient matrix shown in (B), it is possible to perform the differential processing for each of the vertical and horizontal directions.
As the coefficient, for example, a pixel hatched to the lower right can be −1, and a pixel having cross hatching can be 2. Here, it is assumed that the second-order differential processing is performed using a coefficient matrix as shown in FIG. 4A, and the absolute value thereof is output as the feature amount of each pixel. here,
The feature amount of each pixel in the neighborhood area of 5 × 5 pixels is set to k ₀₀ to k _44.
I will call it.

FIG. 5 is an explanatory diagram of an example of the threshold value matrix and the weight value matrix used in the weight value setting unit 12. 25 feature quantities k0 calculated by the feature quantity calculation unit 11
0 to k44 are transferred to the weight value setting unit 12. In the weight value setting unit 12, for example, a threshold matrix as shown in FIG. 5 (A) is defined. If the absolute value of the difference between the feature amount k _{22 of the} pixel of interest and the feature amount k _ij (i = 0 to 4, j = 0 to 4) of each neighboring pixel is less than or equal to the threshold value of FIG.
The value at the position (i, j) in the matrix shown in (B) is a weighted value, and when the value is larger than the threshold value, the weighted value is 0.
Is set. Here, the weighted values of each pixel will be referred to as w ₀₀ to w ₄₄ .

The 25 weight values w _{00 to} w ₄₄ obtained by the weight value setting unit 12 are transferred to the average value calculation unit 13. In the average value calculation unit 12, the pixel values p _{00 to} p ₄₄ and the weight values w ₀₀ to
The weighted average value is obtained from w ₄₄ by the following equation (1), and the result is output to the image output unit 3 as the restored value of the pixel of interest. (Reconstruction _{value) = (w 00 × p 00} + w 01 × p 01 + ... + w 43 × p 43 + w 44 × p 44) / (w 00 + w 01 + ... + w 43 + w 44) × (255/25) - ( 1)

By restoring the pixel value of the target pixel in this way, two pixels can belong to the same density change region when the difference in feature amount between the target pixel and neighboring pixels is small. Since it is considered that the property is high, smooth multilevel restoration can be performed by performing smoothing. Further, when the difference between the feature amounts is large, it is considered that the two pixels are likely to belong to different density change regions. Therefore, the neighboring pixels are not used for smoothing to save the edge. To be done. In this way, the neighboring 1
Good restoration can be realized by performing the smoothing in which the weight value is controlled for each pixel.

In this example, the characteristic amount is obtained by the coefficient matrix shown in FIG. 4A using the 5 × 5 neighborhood area shown in FIG. 3, and the threshold value shown in FIG. Although the processing is performed using the weighted value matrix shown in), it is also possible to use a configuration in which these are modified within a range that produces a similar result. For example, instead of the filter of FIG. 4 (A) used to obtain the feature amount, a plurality of filters are prepared for each direction as shown in FIG. 4 (B), and each filter output has the largest absolute value. Is output as shown in FIG.
It is possible to obtain a finer feature amount than the filter of (A). Further, since the second derivative filter has a sharp reaction to noise, a modification such as smoothing to suppress it and outputting it may be considered.

Further, as another modification, in this example, when the absolute value of the difference between the feature amounts is larger than the threshold value, 0 is set as the weight value, but 1/2 of the value at the position of the weight value matrix is set. A configuration in which various weighting values are set, such as setting or multiplying the value of the position in the weighting value matrix by a value that is inversely proportional to the absolute value of the difference between the feature amounts, is also conceivable.

An example of the operation of the embodiment of the image processing apparatus of the present invention will be described below based on another specific example.
In the above-described specific example, the weighted value is set by comparing the absolute value of the difference in feature amount between the target pixel and each neighboring pixel with the threshold value. Here, a method of changing the weight value by calculating the feature amount of each neighboring pixel and the weight value matrix will be described.

FIG. 6 is an explanatory diagram of an example of a coefficient matrix of the first-order differential filter for obtaining the characteristic amount. In the figure,
Pixels surrounded by thick lines are pixels for which the feature amount is obtained. The feature amount calculation unit 11 obtains the first-order differential values in the horizontal and vertical directions using the coefficient matrices shown in FIGS. 6 (A) and 6 (B). As the coefficient, for example, the pixel hatched to the lower right may be 1, the blank pixel may be 0, and the cross-hatched pixel may be -1.

FIG. 7 is an explanatory diagram showing an example of a conversion table for converting the absolute value of the output of the primary differential filter. The feature amount calculation unit 11 obtains the maximum of the absolute values of the first-order differential values obtained using the coefficient matrix as shown in FIG. 6, and has the functional relationship as shown in FIG. Draw the conversion table. Then, the obtained conversion result is output. The filtering result of the limited gradation image often has a fluctuation due to the influence of the limited gradation in addition to the density change of the original image. Therefore, the output using the conversion table shown in FIG. It is possible to suppress fine fluctuations near the maximum / minimum values. Th1 and Th2 in the figure are appropriately determined in advance so that such a fine fluctuation can be suppressed. Further, the magnitude relationship is inverted because the feature amount is converted so that the output value becomes larger in a flatter region, and the weight value setting unit 12 described next.
This is for performing the processing in. Of course, it is also possible to configure without using the conversion table as shown in FIG.

If the neighborhood area shown in FIG. 3 is used also in this specific example, the 25 feature quantities k ₀₀
~ K ₄₄ is calculated and output to the weight value setting unit 12.
In this example, since the feature amount of the target pixel is not used for setting the weight value as described later, the feature amount calculation unit 11 may omit the calculation of the feature amount k ₂₂ of the target pixel as necessary.

FIG. 8 is an explanatory diagram of another example of the weight matrix. The weight value setting unit 12 multiplies the weight value matrix set in advance as shown in FIG. 8A by the corresponding feature amount, and as shown in FIG. 8B, the weight value for each neighboring pixel. seek w ₀₀ ~w _44. The weight value of the pixel of interest is fixed at 128 here.

Weight value w ₀₀ obtained by the weight value setting section 12
~ W ₄₄ is transferred to the average value calculation unit 13. In the average value calculation unit 12, the pixel values p _{00 to} p of the neighboring area including the target pixel are calculated.
₄₄ and a weighted value w ₀₀ to w _44, obtains a weighted average value by the above equation (1), and outputs to the image output unit 3 the result as the restored value of the pixel of interest.

In the average value calculation unit 13, besides the weighted average value is obtained by using the above-mentioned formula (1), it is also possible to use a configuration in which these are modified within a range that produces a similar result. For example, in each of the specific examples described above, the average value calculation unit 13 performs normalization based on the sum of weight values and multiplication for setting the output value range to 0 to 255. It can also be done at 12. Specifically, the weight value setting unit 12 obtains the weight values w ₀₀ to w _44, and then w _ij ′ = w _ij / (w ₀₀ + w ₀₁ + ... + w ₄₃ + w ₄₄ ) × (255/25) (i, j = 0 to 4)-(2) is calculated and converted into each weighted value w _ij , and the average value calculation unit 13 calculates (restored value) = w ₀₀ '× p ₀₀ + w ₀₁ ' × p ₀₁ + ... Even if the reconstruction value is obtained by + w ₄₃ ′ × p ₄₃ + w ₄₄ ′ × p ₄₄ − (3), the same result as the equation (1) can be obtained.

FIG. 9 is an explanatory diagram of another example of the coefficient matrix of the primary differential filter. In the above-described specific example, the first-order differential filter for detecting the density boundaries in the vertical direction and the horizontal direction as shown in FIG. 6 is used. However, for example, a filter corresponding to the oblique direction as shown in FIG. Different filters can be added. By using such a filter, a finer feature amount can be detected and the weight value can be controlled to obtain a better restored image.

[0036]

As is apparent from the above description, according to the present invention, the averaging process is performed using the weighted values corresponding to the feature amounts of the target pixel and its neighboring pixels to restore the gradation value. Therefore, as compared with the conventional simple averaging, it is possible to perform good gradation restoration in which blurring of the edge portion is suppressed. Further, since the weighted averaging process is performed, there is an effect that it is possible to realize an image processing device capable of sufficient gradation reproduction as compared with restoration by simple averaging.

Further, as in the invention described in claim 2, the target pixel obtained by the feature amount calculating means is based on a threshold value which is predetermined in accordance with the relative position between each neighboring pixel and the target pixel. By setting the weighted value based on the comparison result between the feature amount of the neighboring pixel and the threshold value, the gradation restoration processing is performed using only the neighboring pixel whose feature is similar to that of the pixel of interest, and there is little blurring. It is possible to perform various gradation restoration processing.

Alternatively, as in the third aspect of the present invention, a value determined in advance according to the relative position between each neighboring pixel and the target pixel and the characteristic of each neighboring pixel obtained by the characteristic amount calculating means. By calculating the amount and the weighted value of each pixel, it is possible to perform gradation restoration processing corresponding to a complicated feature change in the image.

The feature amount used in these processings is calculated based on the first-order differential value or the second-order differential value of the image as in the invention described in claim 4 or 5, so that the edge of the image is calculated. It is possible to perform good gradation restoration processing that reflects the existence and the degree thereof.

[Brief description of drawings]

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

FIG. 2 is a flowchart showing an example of a processing flow of a gradation restoration processing unit 4 in the embodiment of the image processing apparatus of the present invention.

FIG. 3 is an explanatory diagram of an example of a near area.

FIG. 4 is an explanatory diagram of an example of feature amount calculation.

5 is an explanatory diagram of an example of a threshold value matrix and a weight value matrix used in the weight value setting unit 12. FIG.

FIG. 6 is an explanatory diagram of an example of a coefficient matrix of a first-order differential filter for obtaining a feature amount.

FIG. 7 is an explanatory diagram of an example of a conversion table that converts the absolute value of the output of the first-order differential filter.

FIG. 8 is an explanatory diagram of another example of the weight value matrix.

FIG. 9 is an explanatory diagram of another example of the coefficient matrix of the first-order differential filter.

[Explanation of symbols]

1 ... Image input unit, 2 ... Image storage unit, 3 ... Image output unit, 4
... gradation restoration processing unit, 5 ... control unit, 11 ... feature amount calculation unit,
12 ... Weight value setting unit, 13 ... Average value calculation unit.

Claims (5)

[Claims] 1. A feature amount calculating means for obtaining a feature amount of each pixel in an input image, and a weight value setting means for obtaining a weight value for a target pixel and its neighboring pixels based on the feature amount obtained by the feature amount calculating means. And an average value calculation means for calculating a weighted average value from pixel values and weighted values corresponding to the target pixel and its neighboring pixels, and the feature amount calculation means as each pixel in the given image as the target pixel, An image processing apparatus, characterized in that gradation restoration processing is performed by the weight value setting means and the average value calculation means. 2. The feature of the pixel of interest and the neighboring pixel calculated by the feature amount computing unit based on a threshold value determined in advance according to the relative position of each neighboring pixel and the pixel of interest The image processing apparatus according to claim 1, wherein the weighted value is obtained based on a comparison result between the amount difference and the threshold value. 3. The weight value setting means uses a predetermined value according to the relative position of each neighboring pixel and a target pixel and the characteristic amount of each neighboring pixel calculated by the characteristic amount calculating means. The image processing apparatus according to claim 1, wherein the weighted value of each pixel is obtained by the above calculation. 4. The image according to claim 1, wherein the feature amount calculation means generates an output value based on a second derivative value of the input image or a value obtained by smoothing the second derivative value. Processing equipment. 5. The image according to claim 1, wherein the feature amount calculation means generates an output value based on a first derivative value of the input image or a value obtained by smoothing the first derivative value. Processing equipment.