JP2002335400A - Image processing unit, image processing method and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image processing unit, an image processing method and a recording medium that can reconfigure a half-tone image generated from a multi-gradation image into the multi-gradation image with high accuracy. SOLUTION: In the image processing unit 1, a feature detection means 12 detects a feature of a multi-gradation reduced image generated from a half-tone image, an image conversion means 13 divides the half-tone image into a plurality of sub band images with different frequencies, and a low frequency reconfiguration means 15 reconfigures low frequency components including fundamental information to decode the image from the sub band division images on the basis of the multi-gradation reduced image detected by the feature detection means 12. A contour correction means corrects high frequency components of the sub band images on the basis of the low frequency components of the sub band images and an image re-conversion means converts the sub band images into a multi-gradation image. Thus, the multi-gradation image can be restored with high accuracy.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to an image processing apparatus, an image processing method, and a recording medium for converting a halftone image generated from a multi-tone image into a multi-tone image.

[0002]

2. Description of the Related Art Conventionally, in image processing for converting a halftone image generated from a multi-tone image to a multi-tone image, characteristics of the halftone process for generating a half-tone image from the multi-tone image are considered. There are many things. For example,
When dither processing is used as halftone processing,
The halftone image is converted into a multi-tone image in consideration of the dither matrix. When the error diffusion processing is used as the halftone processing, the halftone image is converted into a multi-tone image in consideration of the error diffusion characteristics, for example, the error diffusion coefficient. An image processing apparatus for converting a halftone image into a multi-tone image is disclosed in Japanese Patent Laid-Open No. Hei 10-98628. In the image processing apparatus disclosed in Japanese Patent Application Laid-Open No. Hei 10-98628, a halftone image is decomposed into a plurality of sub-band images, and at least one of high-frequency component noise and a pattern is extracted from the plurality of sub-band images. Is selectively removed, and a multi-tone image is created from a halftone image by reconstructing a plurality of sub-band images into a multi-tone image.

[0003]

However, in the conventional image processing in consideration of the halftone processing, only a specific halftone image can be accurately converted to a multi-tone image. For example, in an image processing apparatus on the premise that dither processing is used as halftone processing, it is difficult to convert a halftone image using error diffusion processing as halftone processing into a multi-tone image with high accuracy. Conversely, it is difficult for an image processing apparatus on the assumption that error diffusion processing is used as halftone processing to convert a halftone image using dither processing as halftone processing to a multi-tone image with high accuracy.

In the image processing apparatus disclosed in Japanese Patent Application Laid-Open No. 10-98628, it is possible to remove a high-frequency pattern such as a halftone pattern, but a low-frequency component (LL component) of a sub-band image is used. No processing is performed for. Therefore, after the noise is removed, the histogram distribution of the restored multi-tone image has a strong lingering halftone image. For this reason, it is necessary to perform a smoothing process on the restored multi-tone image, so that it is difficult to preserve edge components such as contours.

An object of the present invention is to provide an image processing apparatus, an image processing method, and a recording medium for converting a halftone image generated from a multi-tone image into a multi-tone image with high accuracy.

[0006]

SUMMARY OF THE INVENTION According to the present invention, there is provided an image processing apparatus for converting an input halftone image into a multitone image, generating a multitone reduced image from the halftone image, and A feature detection unit that detects a feature, an image conversion unit that divides a halftone image into a plurality of sub-band images having different frequencies, and a feature of the reduced image detected by the feature detection unit. An image processing apparatus comprising: a low-frequency component reconstructing unit configured to reconstruct a low-frequency component of a divided sub-band image; and an image re-converting unit configured to form a multi-tone image from the sub-band image. It is.

According to the present invention, a multi-tone reduced image is generated from a half-tone image, the features of the multi-tone reduced image are detected, and the half-tone image is converted into a plurality of sub-frequency images having different frequencies by image conversion means. The image is divided into band images, and a low-frequency component including information that is basic for reconstructing a multi-tone image among the sub-band images is reconstructed by a multi-tone reduced image. Since the sub-band image is converted to a multi-tone image by the image re-conversion means using the reconstructed low frequency components, the reconstructed multi-tone image can suppress the lingering of the half-tone image, and the half-tone image , A multi-tone image can be accurately restored.

Further, the present invention is characterized by comprising a low-frequency component noise removing means for removing low-frequency component noise from the subband image divided by the image converting means.

According to the present invention, it is possible to remove the noise of the low frequency component of the sub-band image by the low frequency component noise removing means, and to accurately restore the multi-tone image from the halftone image. Can be.

Further, the present invention provides a high frequency component noise removing means for removing noise of a high frequency component in the subband image divided by the image converting means, and the divided frequency band converted by the image converting means. And a contour correcting means for correcting a contour by correcting a high-frequency component of the sub-band image based on a low-frequency component of the sub-band image.

According to the present invention, the high frequency component noise removing means removes the noise of the high frequency component of the subband image, and the contour correcting means removes the noise of the high frequency component based on the low frequency component of the subband image. Since the high-frequency component of the sub-band image is corrected, even if necessary information is removed by the high-frequency component noise removing unit, the contour of the image is corrected because the high-frequency component is corrected by the contour correcting unit. Is done. For this reason, it is possible to suppress a loss of an edge necessary for an image, and to accurately restore a multi-tone image from a halftone image.

Further, in the present invention, the low frequency component noise elimination means extracts a block composed of the pixel of interest and its neighboring pixels, and converts the density of the pixel of interest so as to approach the average value of the density of the neighboring pixels. The characteristic is that noise is removed by performing the above.

According to the present invention, a block consisting of a pixel of interest and its neighboring pixels is extracted, and the density of the pixel of interest is converted so as to approach the average value of the density of the neighboring pixels. The absolute value of the difference between the density of the pixel of interest and the absolute value of the density of the target pixel is determined when the number of neighboring pixels whose absolute value is larger than a predetermined threshold value is equal to or greater than a predetermined number. By replacing the density with an average value of the density in the vicinity that becomes larger than a predetermined threshold value, it is possible to remove noise such as an isolated point while keeping the edge.

Further, in the present invention, the high frequency component noise removing means obtains a frequency spectrum of the divided subband image, and removes a high frequency component noise from the frequency spectrum according to the band of the subband image. It is characterized by doing.

According to the present invention, the frequency spectrum of the divided sub-band image is obtained, and the noise of the high-frequency component is removed from the frequency spectrum according to the band of the sub-band image. it can.

Further, the present invention is characterized in that the feature detecting means obtains a density histogram from a multi-gradation reduced image.

According to the present invention, since the characteristic detecting means obtains the density histogram from the multi-tone reduced image, the histogram distribution of the restored multi-tone image can be estimated.

Further, according to the present invention, in an image processing method for converting an input halftone image into a multi-tone image, a multi-tone reduced image is generated from the halftone image, and the feature of the reduced image is detected. A detecting step, an image converting step of dividing the halftone image into a plurality of sub-band images of different frequencies, and using the characteristics of the reduced image detected by the characteristic detecting step, the sub-band divided in the image converting step. An image processing method comprising: a low-frequency component reconstructing step of reconstructing a low-frequency component of an image; and an image re-converting step of forming a multi-tone image from a sub-band image.

According to the present invention, a multi-tone reduced image is generated from a half-tone image, features of the multi-tone reduced image are detected, and the half-tone image is converted into a plurality of sub-band images of different frequencies. After reconstructing a low-frequency component containing basic information for reconstructing a multi-tone image from the sub-band image by using a multi-tone reduced image, the sub-band image Is converted into a multi-tone image, so that the reverberation of the half-tone image can be suppressed in the reconstructed multi-tone image, and the multi-tone image can be accurately restored from the half-tone image.

Further, the present invention is characterized by including a low frequency component noise removing step of removing noise of a low frequency component in the divided subband images.

According to the present invention, it is possible to remove the noise of the low frequency component of the sub-band image and to accurately restore the multi-tone image from the halftone image.

Further, according to the present invention, there is provided a high-frequency component noise removing step for removing noise of a high-frequency component in a divided sub-band image and a low-frequency component based on a low-frequency component in the divided sub-band image. A contour correcting step of correcting a contour by correcting a high-frequency component of the sub-band image.

According to the present invention, the noise of the high frequency component of the subband image is removed, and the high frequency component of the subband image is corrected based on the low frequency component of the subband image. , The outline of the image is corrected. Therefore, it is possible to suppress the loss of the edge required for the image, and to accurately restore the multi-tone image from the halftone image.

Also, according to the present invention, when converting an input halftone image into a multi-tone image, a feature detection step of generating a multi-tone reduced image from the halftone image and detecting a feature of the reduced image And an image conversion step of dividing the halftone image into a plurality of sub-band images of different frequencies, and using the features of the reduced image detected by the feature detection step to generate the sub-band images divided in the image conversion step. Computer-readable recording of a program for causing a computer to execute an image processing method including a low-frequency component reconstruction step for reconstructing low-frequency components and an image conversion step for constructing a multi-tone image from a sub-band image It is a possible recording medium.

A multi-gradation reduced image is generated from the halftone image, and the characteristics of the multi-gradation reduced image are detected.
The halftone image is divided into a plurality of sub-band images of different frequencies, and the low-frequency components containing the basic information for reconstructing the multi-gradation image among the sub-band images are divided into multi-gradation images. After reconstructing with the reduced image, the computer can read and execute an image processing method for converting the sub-band image into a multi-tone image. Thereby, in the reconstructed multi-tone image, the lingering of the half-tone image can be suppressed, and the multi-tone image can be accurately restored from the half-tone image.

According to the present invention, there is provided a computer storing a program for causing a computer to execute an image processing method including a low frequency component noise removing step of removing low frequency component noise in a divided subband image. It is a readable recording medium.

According to the present invention, a computer can read and execute an image processing method for removing noise of low frequency components in a subband image, and accurately restore a multi-tone image from a halftone image. be able to.

Further, the present invention provides a high frequency component noise removing step for removing noise of a high frequency component in a divided sub-band image, and a process for removing a high frequency component noise in the divided sub-band image. And a contour correcting step of correcting a contour by correcting a high-frequency component of a sub-band image.

According to the present invention, an image for removing noise of a high frequency component of a subband image and correcting a high frequency component of the subband image based on a low frequency component of the subband image is provided. The processing method can be read and executed by a computer, and the outline of the image is corrected. Therefore, it is possible to suppress the loss of the edge required for the image, and to accurately restore the multi-tone image from the halftone image.

[0030]

FIG. 1 is a block diagram showing a configuration of an image processing apparatus 1 according to an embodiment of the present invention. The image processing device 1 converts a halftone image input from the image input device 11 as shown in FIG.
This is a device for converting into a multi-tone image as shown in FIG. The image processing apparatus 1 includes a feature detecting unit 12, an image converting unit 13, a high frequency component noise removing unit 14, a low frequency component reconstructing unit 15, a low frequency component noise removing unit 16, a contour correcting unit 17, and an image It includes the re-conversion means 18.

The image input device 11 is, for example, a scanner or a digital camera.
(Magneto-optical disk) and DVD-RAM (Digital Ve
When a halftone image is recorded on a recording medium such as a rsatileDisk (Random Access Memory), an apparatus that reads the image from these storage media may be used. Further, the image input device 11 may be a communication means such as a modem when capturing a halftone image via a network or the like.

The image output device 18 is, for example, an image display device such as a liquid crystal display device or a CRT (Cathode Ray Tube),
An image forming apparatus using an electrophotographic method or an ink jet method.

Hereinafter, the configuration of the image processing apparatus 1 will be described with reference to a flowchart showing the operation processing of the image processing apparatus 1 shown in FIG. An image input device 11 for the image processing device 1
When a halftone image is input from the
2 forms a multi-tone reduced image from the halftone image and creates a histogram distribution which is a feature of the reduced image (step S1). Next, the image conversion unit 13 divides the halftone image into a plurality of sub-band images having different frequencies (Step S2). Next, the high frequency component noise removing unit 14 removes noise (such as an edge of a dot seen after halftone processing) such as a high frequency pattern existing in the high frequency component from the subband image divided by the image converting unit 13. It is removed (step S3). Next, the low-frequency component reconstructing means 15 reconstructs the low-frequency component based on the features of the reduced image detected by the feature detecting means 12 (Step S4). Next, the low frequency component noise removing means 16 removes noise such as an isolated point (step S5). Next, the contour correcting means 17 corrects the contour of the high frequency component (step S6). Finally, the image re-conversion unit 18 reconstructs the sub-band image into a multi-tone image (Step S7).

Hereinafter, the structure of the image processing apparatus 1 will be described in more detail. The feature detection unit 12 generates a reduced image with multiple gradations from the halftone image. When reconstructing a halftone image into an image of 256 gradations, a block of 16 × 16 pixels (256 pixels) of the input halftone image is regarded as one pixel, and the sum of the pixel values in the block is reduced. By treating it as the pixel value of the image, 25 times 1/16 times in the horizontal and vertical directions respectively.
A 6-gradation reduced image is obtained. Then, a histogram indicating a pixel distribution of each gradation is obtained as a feature of the reduced image of 256 gradations.

The image conversion means 13 divides the halftone image into a plurality of sub-band images having different frequencies. In the case where the halftone image is reconstructed into an image of 256 gradations, the pixels having the gradation frequencies of 0 and 1 in the halftone image are replaced so that the gradation frequencies are 0 and 255, respectively. A sub-band image is generated by performing a wavelet transform. In the present embodiment, the discrete wavelet transform is used. In addition to the discrete wavelet transform, a QMF (Quadrature Mirror Filter) bank, a CQF (Conjugate Quadrature Filter) bank,
Or, after converting to the frequency domain using a band division method using subbands such as SSKF (Symmetric Short Kernel Filter) bank, discrete cosine transform, discrete sine transform, or discrete Fourier transform, the frequency domain (band frequency) The sub-band image may be generated using a method such as a band division method of dividing the image data every time.

Here, the wavelet transform will be described. FIG. 4 is a block diagram showing an octave division process and a reconstruction process based on a tree structure of the wavelet transform. Hi (Z) is the input image (halftone image)
Is divided into sub-band images, and Fi (Z) is a filter bank used when reconstructing an image (continuous tone image) from the sub-band image. ↑ 2 represents upsampling (image division processing), and ↓ 2 represents downsampling (image reconstruction processing). Also, here, the tree configuration has two stages. A two-divided filter bank Hi (i = 0, 1) is used as a filter bank, H0 is a low-pass filter, and H1 is a high-pass filter. Image conversion means 13
Block 20 indicates that the subband output after passing through Hi (Z) is down-sampled.
1 indicates that the subband output after passing Fi (Z) is upsampled. At this time, the halftone image can be divided into a plurality of subband images by the wavelet transform processing shown in FIG.

FIG. 5 is a diagram showing an octave dividing process by a tree structure of a two-dimensional discrete wavelet transform in which the image converting means 13 actually divides an input halftone image into subband images. First, using a vertical low-pass filter 31 and a vertical high-pass filter 32, a vertical low-pass component is calculated as a vertical L component and a vertical high-pass component is calculated as a vertical H component, and the halftone image is divided into two in the vertical direction. . Further, each component is divided into two using a horizontal low-pass filter 33 and a horizontal high-pass filter 34 for the vertical L component and a horizontal low-pass filter 35 and a horizontal high-pass filter 36 for the vertical H component. I do. In the image conversion means 13, the vertical low-pass component and the horizontal low-pass component are the LL component, the vertical high-pass component and the horizontal low-pass component are the HL component, the vertical low-pass component and the horizontal high-pass component are the LH component, and the vertical low-pass component is the LH component. A high-pass component and a horizontal high-pass component as an HH component of 4
Calculate the two components.

FIG. 6 is a diagram showing the result of performing an octave division process on the input halftone image by a tree structure of the wavelet transform. In the present embodiment, the halftone image is divided into four components of the LL component, the HL component, the LH component, and the HH component.
Although the subband image is divided into three subband images, the LL component of the subband image may be further divided into LLLL components, LLHL components, LLLH components, and LLHH components by wavelet transform, as shown in FIG. The image reconverting means 19 for reconstructing a multi-tone image from the divided subband images will be described later.

Next, the high frequency component removing means 14 will be described. H of a subband image including high frequency components
It is expected that the L component, the LH component, and the HH component include a lot of noise (for example, a high-frequency pattern). Therefore, these noises are removed by the high frequency component removing means 14. Hereinafter, a specific method for removing noise such as a high-frequency pattern will be described. First, discrete Fourier transform is performed on the HL component, LH component, and HH component of the sub-band image divided by the image conversion means 13 to obtain a frequency spectrum of each component.

The equation of the discrete Fourier transform is shown in the following equation (1). F (u, v) = ΣmΣnf (m, n) exp (−2πmuj / N1) exp (−2πmvj / N2) (1) where m is the horizontal coordinate of the image (m = 0, 1,..., N1
-1), n is the vertical coordinate of the image (n = 0, 1,...,
N2-1), N1 and N2 represent the image size for performing the discrete Fourier transform, f (m, n) represents the image signal, and F (u, v) represents the Fourier matrix. Also, since F (u, v) is a complex number,
F (u, v) is represented by equation (2). F (u, v) = a (u, v) + ib (u, v) (2) where i represents an imaginary number. Therefore, the frequency spectrum |
F (u, v) | is represented by equation (3). | F (u, v) | = √ (a (u, v) ² + b (u, v) ² ) (3)

As described above, when the Fourier transform is performed, the horizontal coordinate of the image is set as m, and the vertical coordinate of the image is set as n, and the frequency spectrum is obtained. As shown in FIG. The central part 41 is a high frequency component, a diagonal direction 42
Becomes a low frequency component. Therefore, when the frequency spectrum regions of the high frequency component in the horizontal direction, the high frequency component in the vertical direction, and the high frequency component in the horizontal and vertical directions are simulated, they are the regions indicated by H in FIGS. 8, 9, and 10, respectively.

Therefore, for example, in the HL component of the subband image, the high frequency component in the vertical direction (FIG. 9) of the frequency spectrum included in the HL component from which the high frequency component in the horizontal direction is originally desired to be extracted By replacing the frequency spectrum of the portion indicated by H of the high-frequency component (FIG. 10) with 0, it is possible to roughly remove high-frequency noise. Similarly, for the LH component, the frequency spectrum of the portion indicated by H of the horizontal high frequency component (FIG. 8) and the horizontal vertical high frequency component (FIG. 10) included in the LH component is replaced with 0. To remove high frequency noise. As for the HH component, the frequency spectrum of the portion indicated by H of the horizontal high frequency component (FIG. 8) and the vertical high frequency component (FIG. 9) included in the HH component is replaced with 0 to remove high frequency noise. Is removed.

Next, the HL component, LH component, and HH component of the sub-band image from which high-frequency noise has been removed are obtained by performing a discrete inverse Fourier transform. The expression of the discrete inverse Fourier transform is shown in the following expression (4). f (m, n) = ΣuΣvF (u, v) exp (−2πmuj / N1) exp (−2πmvj / N2) (4) where u is the horizontal space coordinate, v is the vertical space coordinate,
N1 and N2 are the image sizes for performing the discrete inverse Fourier transform, f
(m, n) represents an image signal, and F (u, v) represents a Fourier matrix. By performing the above processing, unnecessary high frequency components can be removed from the HL component, LH component, and HH component of the subband image.

On the other hand, since the LL component of the subband image contains information that is basic for restoring the image, the HL component, the LH component, and the processing performed on the HH component as described above are not performed. Instead, another process is performed by a low-frequency component reconstructing unit 15 and a low-frequency component noise removing unit 16, which will be described later.

Next, the low frequency component reconstruction means 15 will be described in detail. Since the reduced image generated by the feature detecting means 12 is an image suitable for representing the characteristics of the image, the histogram of the image reconstructed into the multi-tone image may be similar to the histogram of the reduced image. desirable. Therefore, first, a density histogram of the LL component of the subband image is generated. And L of the subband image
The density histogram of the LL component is converted so that the density histogram of the L component has the same distribution as the histogram of the reduced image obtained by the feature detection unit 12.

FIG. 11 shows a low-frequency component reconstructing means 15.
10 is a flowchart showing the conversion processing operation of the histogram of the LL component of the sub-band image. The processing operation of the low frequency component reconstructing means 15 shown in FIG. 11 is performed in step S4 in FIG. Here, m and n are the horizontal position and the vertical position, respectively, and the LL component is LL (m,
n), and a histogram Hist (i) (i = 0 to N−1, where N represents the number of gradations) indicating the frequency of each gradation value of the LL component is obtained. Now, a histogram indicating the frequency for each gradation value of the reduced image obtained by the feature calculation unit 12 is represented by hi
Assuming that sts (i) (i = 0 to N−1, where N represents the number of gradations), the sum of hists (i) (i = 0 to N−1) is Hist (i) (i = 0 / N-1), which is 4 / N of the total sum (in the case of 1-octave division).

First, in step S41, i = 0, count
= 0, sum = Hist (0), sums = 0, each parameter is set to an initial value, and the process proceeds to step S42. Step S
At 42, it is determined whether or not sum is greater than hists [count] × N / 4. Here, the frequency of the LL component of the gradation number 0 and the frequency of the reduced image are compared. Where sum is hist
If it is smaller than s [count] × N / 4, step S
Proceed to 43. In step S43, LL in the LL component
Of the LL components satisfying (m, n) = i, all values not yet replaced with count are replaced with count, and the replaced image is added to sums. Next, the process proceeds to step S44, i is increased by 1, Hist (i), which is the frequency of the histogram of the number of gradations i, is added to sum, and the process proceeds to step S45.

On the other hand, in step S42, sum is set to hists [co
unt] × N / 4 or more, the process proceeds to step S46.
In step S46, LL (m, n) = i in the LL component
Of the components satisfying, hists [count] × N / 4−sums pixels are replaced with count, and the process proceeds to step S47. In step S47, sum is sum and hists [count] × N / 4
, The count is increased by 1, the sums is set to 0, and the process proceeds to step S45.

In step S45, it is determined whether or not count is equal to N-1. Here, if it is determined that the count is not N-1, the process proceeds to step S42, where the count is N
The above operation is repeated until the value becomes -1. On the other hand, if it is determined in step S45 that the count is N−1, among the LL components satisfying LL (m, n) = i,
Replace all values not replaced with count with count and end the processing operation.

The conversion processing operation of the LL component histogram of the above-described subband image will be specifically described using, for example, an example of reconstructing the LL component image into a 4-gradation image. In this example, the size of the original image is 128 pixels ×
Assume 128 pixels.

First, the histogram of the reduced image of four gradations is
hists (0) = 1024, hists (1) = 768, hist
It is assumed that s (2) = 1280 and hists (3) = 1024. At this time, the sum of the pixels of the reduced image is 4096 (6
4 × 64). The histogram of this reduced image is shown in FIG.
2 (a) is shown.

Therefore, the histogram of the target LL component is Hist (0) = 1024 × N / 4, Hist (1)
= 768 × N / 4, Hist (2) = 1280 × N / 4, Hi
st (3) = 1024 × N / 4. Now, since the number of gradations is 4, Hist (0) = 1024, Hist (1) = 76
8, Hist (2) = 1280 and Hist (3) = 1024.

The histogram of the LL component actually obtained is Hist (0) = 1280, Hist (1) = 768, Hi
It is assumed that st (2) = 768 and Hist (3) = 1280. FIG. 12B shows a histogram of the LL component.

First, in step S41, i = 0, count
= 0, sum = Hist (0), sums = 0, each parameter is set to an initial value, and the process proceeds to step S42. Step S
At 42, it is determined whether or not sum is greater than hists [count] × N / 4. Here, i = 0 and Hist (0) = 1
280, which is larger than the target histogram value 1024, and the process proceeds to step S46. In step S46, among LL components satisfying LL (m, n) = i = 0, 1024 (hists [count] × N / 4−sums = 1
024 × 4 / 4-0 = 1024) pixels are all counted
And the process proceeds to step S47. In step S47, sum is calculated as sum− (hists [count] × N / 4−sums) = 2
56, count is set to 1, and sums = 0. Next, the process proceeds to step S45, and since the count is not 3, the process returns to step S42.

In step S42, sum is again hists [co
unt] × N / 4 is determined. Now
sum = 256, and the target histogram value 10
Since it is smaller than 24, the process proceeds to step S43. In step S43, LL (m,
n) = 256 pixels among the pixels satisfying i = 0 that have not yet been replaced by count (in the above-described step S46,
1024 of the pixels satisfying LL (m, n) = 0
Since pixels are replaced by count, pixels that are not replaced by count are 256).
Is added to sums, and sums = 256. Next, the process proceeds to step S44, where i is set to 1 and sum
Is added to Hist (i) = 768 to obtain sum = 1024.
Proceed to step S45. In step S45, since i is not 3, the process returns to step S42.

In step S42, since sum = 1024, which is larger than hists [count] × N / 4 = 768, the process proceeds to step S46, and 512 (hists) of the LL components satisfying LL (m, n) = 1 [1] × N / 4-sums
= 768-256 = 512) pixels are replaced with count = 1. Then, sum is 256 (sum-hists [1] × N
/ 4 = 1024-768 = 256), the count is set to 2, and the sums are updated to 0.

Hereinafter, each step between steps S42 to S45 is repeated. Now, step S45
And when count = 3, the other parameter is sum
= 1024, sums = 0, i = 3, and step S6
In the LL component satisfying LL (m, n) = 3, 256 pixels are replaced with count = 2. Therefore, in step S48, 1024 pixels out of 1280 pixels of the LL component satisfying LL (m, n) = 3
The replacement process ends with count = 3. As a result, LL
The histogram of the components is Hist (0) = 1024, Hist
(1) = 768, Hist (2) = 1280, Hist (3) =
1024, which can be converted into a histogram distribution having the same shape as the reduced image. FIG. 12C shows a histogram of the LL component after the conversion. As described above, in the histogram conversion, the frequency of the LL component histogram is sequentially compared with the frequency of the histogram obtained by multiplying the reduced image by N / 4 (when divided by one octave), and the LL component histogram is formed into the same shape as the reduced image histogram. A conversion process is performed so that In the same gradation value (density value), a pixel to be replaced with another gradation value may be randomly selected, or
The processing may be performed sequentially from the pixels located above the image, and the pixels appearing first may be replaced with priority.

Since the LL component of the subband image is obtained by downsampling the original image (halftone image) input from the image input device 11, it is substantially equivalent to a reduced image obtained by reducing the original image to a multi-tone image. It is considered to be.
For this reason, the LL component is considered to be an image close to a multi-tone image. However, in reality, the histogram shape of the LL component is still in a biased state, leaving a reverberation of the binarized image. Therefore, by performing the histogram conversion on the LL component, it is possible to perform conversion into a multi-tone image with higher accuracy.

Next, the low frequency component noise removing means 16 will be described. The low frequency component noise removing means 16
For the LL component of the sub-band image subjected to the histogram conversion by the low frequency component reconstructing means 15, first, a block including a target pixel and its neighboring pixels is extracted. And
The density difference between the pixel of interest and its neighboring pixels is obtained. Replace with the average value of the density. When the low-frequency component reconstruction unit 15 performs histogram conversion of the LL component of the sub-band image, the contrast is enhanced, and thus the isolated point is further enhanced. In contrast, noise can be removed while preserving edges. In the present embodiment, as a block including the target pixel and its neighboring pixels, using a 5 × 5 mask size, T = 128, K
= 16.

Next, the contour correcting means 17 will be described. The contour correction unit 17 performs a differential filter process, which is a process of extracting a contour (edge), on the LL component from which noise has been removed by the low frequency component noise removal unit 16.
FIG. 13A is a diagram illustrating an example of a horizontal differential filter, and FIG. 13B is a diagram illustrating an example of a vertical differential filter. First, the horizontal differential filter processing result dH (m, n) for the LL component is obtained. Then, the HL component is corrected from the horizontal differential filter processing result dH (m, n) and the HL component HL (m, n) by the following equation (5). HL (m, n) = HL (m, n) + dH (m, n) / 4 (5)

Next, a result dV (m, n) of the vertical differential filter processing for the LL component is obtained. Then, the LH component is corrected from the vertical differential filter processing result dV (m, n) and the LH component LH (m, n) by the following equation (6). LH (m, n) = LH (m, n) + dH (m, n) / 4 (6)

Further, the horizontal differential filter processing result d
H (m, n) and vertical differential filter processing result dV
From (m, n), the horizontal and vertical differential processing result dHV (m, n) is obtained by the following equation (7). dHV (m, n) = √ (dH (m, n) ² + dV (m, n) ² ) (7)

Then, the horizontal and vertical differential processing results dHV
From (m, n) and the HH component HH (m, n), HH
The components are corrected by the following equation (8). HH (m, n) = HH (m, n) + dHV (m, n) / 4 (8)

In the image processing apparatus 1 of this embodiment, the high frequency component noise removing means 14 removes high frequency components which are considered unnecessary as HL component, LH component and HH component from the frequency spectrum of the subband image. Removed as noise. For this reason, in the processing of the high frequency component noise removing means 14, there is a possibility that the actually necessary high frequency components are also removed. However, by restoring the high-frequency component from the LL component that is considered to contain the least noise by the above-described contour correction unit 17, it is possible to reconstruct a more accurate multi-tone image. Also, in the above-described contour correction means 17, the result of the differential filter processing is divided by 4, but this value can be set arbitrarily, and processing is performed on many images in advance, and an appropriate value is set. You should keep it. This correction is similar to a method generally performed in edge reproduction using a Laplacian filter.

Next, the image re-conversion means 18 will be described. The image reconversion unit 18 performs the octave reconstitution processing based on the tree structure of the wavelet transform shown in FIG. A two-divided filter bank Fi (i = 0, 1) is used as a filter bank, F0 is a low-pass component reconstruction filter, and F1 is a high-pass component reconstruction filter. At this time, a multi-tone image can be formed from the subband image by the two-dimensional inverse discrete wavelet transform processing shown in FIG. FIG. 14 is a diagram showing an octave reconstruction process based on a tree configuration of a two-dimensional inverse discrete wavelet transform for actually converting a sub-band image into a multi-tone image by the image re-conversion means 19. First, the vertical component reconstruction low-pass filter 5 for the LL component and the HL component of the subband image, respectively.
1. The L component in the horizontal direction is formed using the vertical component reconstruction high-pass filter 52. Similarly, the horizontal H component is reconstructed for the LH component and the HH component of the subband image using the vertical component reconstruction low-pass filter 53 and the vertical component reconstruction high-pass filter 54, respectively. Further, by using the horizontal component reconstruction low-pass filter 55 for the horizontal L component and the horizontal component reconstruction high-pass filter 56 for the horizontal H component, a multi-tone image can be reconstructed ( See FIG. 6). In the present embodiment, the flow of processing by one-octave division is shown. However, the number of octave division processing is not actually limited to one, and an appropriate division is performed in consideration of processing time and processing accuracy. The number of times may be set.

The image processing for converting a halftone image into a multi-tone image in the image processing apparatus 1 described above may be realized by a program for causing a computer to execute this image processing. This program is recorded on a computer-readable recording medium. Therefore,
A program for performing the above-described image processing can be provided in a portable manner.

As the recording medium, for example, a memory for performing processing by a microcomputer, for example, an R
The OM itself may be a program medium, or a recording medium provided with a program reading device as an external storage device and readable by inserting a recording medium into the program reading device. In any case, the stored program may be configured to be accessed and executed by a microprocessor, or in any case, the program may be read out,
The read program may be downloaded to a program storage area of a microcomputer, and the program may be executed. It is assumed that this download program is stored in the main unit in advance.

Here, the program medium is a recording medium configured to be separable from the main body, such as a tape-based recording medium such as a magnetic tape or a cassette tape, or a magnetic disk such as a floppy (registered trademark) disk or a hard disk. And CDROM, MO, MD, DVD, DVDRA
M, etc., a disc-type recording medium such as an optical disc, a card-type recording medium such as an IC card including a memory card and an optical card, a mask ROM, an EPROM (Erasable Progr.
ammable Read Only Memory), EEPROM (Electric
It may be a recording medium that fixedly holds a program, including a semiconductor memory such as an ally erasable programmable read only memory (ROM) or a flash ROM.

Further, in the present embodiment, since the system configuration is such that it can be connected to a communication network including the Internet, it may be a medium that carries the program fluidly so that the program is downloaded from the communication network. When downloading a program from a communication network in this way, the downloaded program is stored in the main body in advance, or
It may be installed from another storage medium.

As described above, according to the present invention, a halftone image can be accurately restored to a multi-tone image.
For example, a personal computer can handle a halftone image as a compressed image. When a halftone image is treated as a compressed image, the compressed image is a halftone low gradation image. By displaying the gradation image as it is, the JPEG (The Jo
int Picture Experts Group) Does not restore images as long as images do. Further, it is possible to know the rough contents of the image file. Also, instead of handling low gradation images as compressed data, MH (Modified H
uffman), MR (Modified Read), JBIG (The Joi
nt Bi-Level Image Experts Group)
Data compressed by the value image compression technique can be handled as compressed data. Therefore, when an image is downloaded on the Internet or the like, the amount of information to be downloaded can be reduced, and communication costs can be reduced.

[0071]

According to the present invention, since a multi-tone image can be accurately restored from a half-tone image, the multi-tone image is converted into a half-tone image by half-tone processing, and the half-tone image is converted into compressed data. Can handle.

Further, according to the present invention, the frequency spectrum of the divided sub-band image is obtained, and high frequency component noise is removed from the frequency spectrum according to the band of the sub-band image. Can be.

Further, according to the present invention, a block composed of the target pixel and its neighboring pixels is extracted, and the absolute value of the difference in density between the target pixel and its neighboring pixels is obtained. This absolute value is determined in advance. When the number of neighboring pixels that are larger than the threshold value is equal to or more than a predetermined number, the density of the pixel of interest is replaced with the average value of the density of the neighborhood where the absolute value of the density difference is larger than a predetermined threshold. Noise can be removed while preserving edges.

According to the present invention, the histogram of the reduced image is used as a feature of the reduced image of multiple gradations.
Reconstruction of low frequency components can be easily performed.

[Brief description of the drawings]

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

FIG. 2A is a halftone image input to the image processing apparatus 1 of FIG. 1, and FIG. 2B is a multi-tone image output from the image processing apparatus 1 of FIG. .

FIG. 3 is a flowchart illustrating a processing operation of the image processing apparatus 1 according to an embodiment of the present invention.

FIG. 4 is a diagram showing an octave division process and a reconstruction process based on a tree structure of a wavelet transform.

FIG. 5 is a diagram showing an octave division process based on a tree structure of a two-dimensional discrete wavelet transform for actually dividing an input halftone image into subband images.

FIG. 6 is a diagram illustrating a result of performing an octave division process on an input halftone image using a tree structure of a wavelet transform.

FIG. 7 is a diagram illustrating a frequency spectrum of a subband image.

FIG. 8 is a diagram schematically illustrating a frequency spectrum region of a high frequency component in a horizontal direction.

FIG. 9 is a diagram schematically illustrating a frequency spectrum region of a vertical high frequency component.

FIG. 10 is a diagram schematically illustrating a frequency spectrum region of a high frequency component in the horizontal and vertical directions.

FIG. 11 is a diagram showing a flowchart of a histogram conversion process.

FIGS. 12A, 12B, and 12C are diagrams illustrating an example of a histogram conversion process.

13A is a diagram illustrating an example of a horizontal direction differential filter, and FIG. 13B is a diagram illustrating an example of a vertical direction differential filter.

FIG. 14 is a diagram illustrating processing for reconstructing an image from a subband image by two-dimensional inverse discrete wavelet processing.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Image processing apparatus 12 Feature detection means 13 Image conversion means 14 High frequency component noise removal means 15 Low frequency component reconstruction means 16 Low frequency component noise removal means 17 Edge correction means 18 Image reconversion means

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5B057 CA07 CB08 CD05 CE02 CE03 CE11 DA17 DC23 5C077 LL19 MP01 NN02 PP20 PP49 PQ19

Claims (12)

[Claims] 1. An image processing apparatus for converting an inputted halftone image into a multi-tone image, a feature detecting means for generating a multi-tone reduced image from the halftone image, and detecting a feature of the reduced image. Image conversion means for dividing the halftone image into a plurality of sub-band images of different frequencies; and using the characteristics of the reduced image detected by the characteristic detection means to reduce the sub-band image divided by the image conversion means. An image processing apparatus comprising: low-frequency component reconstruction means for reconstructing frequency components; and image re-conversion means for forming a multi-tone image from a sub-band image. 2. The image processing apparatus according to claim 1, further comprising a low frequency component noise removing unit that removes noise of a low frequency component in the subband image divided by the image converting unit. . 3. A high frequency component noise removing unit for removing noise of a high frequency component in the subband image divided by the image converting unit; and a divided subband converted by the image converting unit. The image processing apparatus according to claim 1, further comprising: a contour correction unit configured to correct a contour by correcting a high-frequency component of the sub-band image based on a low-frequency component of the image. 4. The low-frequency component noise removing unit extracts a block including a pixel of interest and its neighboring pixels,
3. The image processing apparatus according to claim 2, wherein the noise is removed by converting the density of the target pixel so as to approach the average value of the density of the neighboring pixels. 5. The high frequency component noise removing means obtains a frequency spectrum of the divided subband image, and removes high frequency component noise from the frequency spectrum according to the band of the subband image. The image processing apparatus according to claim 3, wherein: 6. The image processing apparatus according to claim 1, wherein said characteristic detecting means obtains a density histogram from a reduced image of multiple gradations. 7. An image processing method for converting an input halftone image into a multi-tone image, a feature detecting step of generating a multi-tone reduced image from the halftone image and detecting a feature of the reduced image. An image conversion step of dividing the halftone image into a plurality of sub-band images having different frequencies; and using the characteristics of the reduced image detected in the characteristic detection step to reduce the sub-band image divided in the image conversion step. An image processing method, comprising: a low-frequency component reconstruction step for reconstructing a frequency component; and an image re-conversion step for constructing a multi-tone image from a sub-band image. 8. The image processing method according to claim 7, further comprising a low frequency component noise removing step of removing low frequency component noise in the divided subband images. 9. A high-frequency component noise removing step of removing noise of a high-frequency component of the divided sub-band image, and a sub-band based on the low-frequency component of the divided sub-band image. 9. The image processing method according to claim 7, further comprising the step of: correcting a contour by correcting a high-frequency component of the band image. 10. A feature detecting step of generating a multi-tone reduced image from a half-tone image and detecting a feature of the reduced image when converting the input half-tone image into a multi-tone image; An image conversion step of dividing the tone image into a plurality of sub-band images of different frequencies; and using the characteristics of the reduced image detected by the characteristic detection step, the low-band frequency of the sub-band image divided in the image conversion step Computer-readable recording in which a program for causing a computer to execute an image processing method including a low-frequency component reconstruction process for reconstructing components and an image conversion process for constructing a multi-tone image from a subband image is recorded. Medium. 11. A program for causing a computer to execute an image processing method including a low-frequency component noise removing step of removing low-frequency component noise in a divided sub-band image. Computer readable recording medium. 12. A high-frequency component noise removing step of removing noise of a high-frequency component of the divided sub-band image; and a sub-band based on the low-frequency component of the divided sub-band image. 12. The computer-readable recording medium according to claim 10, wherein a program for causing a computer to execute an image processing method including: a contour correcting step of correcting a contour by correcting a high-frequency component of a band image.