JP2002247368A - Image processing apparatus, image processing method, program for executing the method, and recording medium storing the program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the image quality by determining image regions, based on band-divided coefficient signals and applying an optimum process to every attribute. SOLUTION: A filtering means 3 detects the feature quantity of characters or mesh dots using band-divided coefficient signals, and corrects the frequency characteristics suited to the characters or the dots, thereby controlling the sharpness of the characters.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for improving image quality, for example, an image forming apparatus such as a printer, a digital copying machine, a facsimile machine, etc. using an electrophotographic process, and an image processed by judging the attribute of an image area. Applied to processing equipment.

[0002]

2. Description of the Related Art Conventionally, as a method of enhancing an image,
An image processing method has been proposed in which an image is decomposed into signals in a plurality of frequency bands by performing a wavelet transform, a signal in a certain frequency band is multiplied by a predetermined number to enhance the signal, and inverse wavelet transform is performed. -274
No. 614).

As a method related to image area separation, a plurality of band signals are generated from image data, edge information of the image data is extracted by at least one band signal, and an image in which a character area is detected based on the distribution of the edge information. A processing method has been proposed (see JP-A-6-223172).

However, in the former method, since the same control is performed on the character image and the halftone image, it is difficult to realize both the sharpness of the character image and the suppression of the moire of the halftone image. . In the latter method, character determination is performed by detecting the density of a band signal having a value larger than a certain threshold. However, even in a halftone image, a band signal having a large value as described above is used. As a result, the accuracy of character determination is not good, and the accuracy of determination is reduced because a configuration is used in which a low frequency band signal is generated while performing downsampling.

By the way, in order to convert the input image signal into a desired frequency characteristic, the input image signal is divided into a plurality of frequency bands and a plurality of directional components, and the divided signal is corrected. Has been proposed.

According to this method, since the frequency characteristic can be selectively corrected for each frequency band or each direction, fine sharpness control can be performed, and the image signal is once divided into each frequency band and each direction component. If so, the sharpness control can be performed by multiplying the parameters, so that the processing becomes very simple.

However, when the same processing as described above is performed on a real space image signal, the characteristic of the image is detected using a detection filter having frequency characteristics for each band and each direction, and according to the detection result. It is necessary to prepare a plurality of filters for enhancing or smoothing only a specific direction, and to perform a convolution operation with a large number of product-sum operations.

[0008]

As described above, the process of correcting the frequency characteristics of a band-divided signal is a very effective technique. However, when the same processing is performed on an image having different attributes such as a character image and a halftone image, there is a problem that it is difficult to achieve both sharpness of the character image and suppression of moiré of the halftone image. .

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned problems, and an object of the present invention is to determine an attribute of an image area based on a band-divided coefficient signal and to perform an optimum process for each attribute. To provide an image processing apparatus which improves the image quality.

Usually, in order to reduce the amount of data in image compression or the like, downsampling is performed on the obtained coefficient signal. However, when downsampling is performed, the correction accuracy of the frequency characteristic is reduced, There is a problem that attribute determination accuracy is reduced.

Another object of the present invention is to provide an image processing apparatus for determining an image attribute with high accuracy by using a band signal for which downsampling is not performed.

[0012]

According to the present invention, a document in which different image attributes such as a character image and a halftone image are mixed is used.
The optimal frequency characteristic is corrected with high accuracy in all regions, and the image attributes are determined with high accuracy, thereby realizing high image quality.

According to the present invention, the band dividing means has redundancy and guarantees continuity due to the redundancy. In other words, when the band dividing means divides into a plurality of frequency bands and coefficient signals of a plurality of directional components, the band dividing means performs at least one frequency band without thinning out the pixels of the input image. An image signal after processing is output based on a plurality of inverse transform outputs (even, odd) obtained for the input pixels.

According to the present invention, the character portion is determined with high accuracy based on the magnitude and continuity (variance) of the coefficient signal, and the halftone image portion is determined with high accuracy based on the density of inflection points of the coefficient signal. I do. Evaluation of the variance of the coefficient signal is as follows: if HL, upper and lower coefficients, LH
If it is HH, the diagonal coefficient is used. The inflection point is H
Counting is performed in the vertical direction for L, the horizontal direction for LH, and the oblique direction for HH.

According to the present invention, the number of line memories and buffers to be used is kept low by using the coefficient signal of the highest frequency band, thereby realizing a low-cost apparatus and method.

In the present invention, a character feature value and a halftone feature value are extracted with high precision using a band signal for which downsampling is not performed, thereby realizing high image quality.

In the present invention, the sharpness of the low frequency band is controlled with a smaller number of line memories.

[0018]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be specifically described below with reference to the drawings. (Embodiment 1) FIG. 1 shows the configuration of an embodiment of the present invention. An image signal input by an image input unit 1 such as a scanner is input to a filter processing unit 3 after an image magnification is changed by a scaling unit 2. The filtering means 3 converts the image into a predetermined spatial frequency characteristic. The image data output from the filter processing means 3 is converted by the gamma conversion processing means 4 to have a predetermined density characteristic, and further converted into multi-valued or binary image data by the halftone processing means 5, and Image output means 6 such as a printer
Is output to

FIG. 2 shows the configuration of the filter processing means of the present invention. As shown in FIG. 2, the input image signal S is first input to the band dividing means 7 and decomposed into a plurality of image band signals W. In the feature amount calculating means 8, the decomposed image signal W
Is used to determine a feature amount indicating an image attribute such as a character image or a halftone image. Next, the sharpness control unit 9 performs an image enhancement process and a smoothing process according to the calculated feature amount. Finally, it is configured to be converted into a real space image signal by the band synthesizing unit 10 and output.

The band dividing means 7 according to the embodiment of the present invention comprises:
This is realized by a wavelet transform as shown in FIG. First, the input image signal S
Wavelet transform is performed in the x direction by (x) 701 and high-pass filter H (x) 702. Here, the low-pass filter G (x) is a low-frequency component extraction filter for obtaining an average value component as shown in FIG. 4A, and the high-pass filter H (x) is as shown in FIG. This is a high-frequency component extraction filter for obtaining a difference component.

In the present embodiment, description will be made by taking a wavelet basis function (Haar) having the characteristics shown in FIG. 4 as an example. Filter groups 703, 704, 7 are applied to the obtained image signals.
05 and 706, a wavelet transform in the y direction is performed. The one-layer wavelet coefficients are obtained by the above conversion.

In FIG. 3, 1st-LL is a low frequency component of one layer, and is an image signal obtained by averaging 2 × 2 pixels with respect to the original image. Also, 1st-LH is 1
It is a horizontal high-frequency component of the hierarchy, and is an image signal obtained by extracting a horizontal edge signal corresponding to the Nyquist frequency. Similarly, 1st-HL is a vertical high-frequency component of one layer, and a vertical edge signal, and 1st-HH is an image signal obtained by extracting an oblique edge signal.

As shown in FIG. 5, Haa is used as a basis function.
In the wavelet transform using the r function, conversion is performed in units of 2 × 2 pixels. When values of four pixels are a, b, c, and d as shown in FIG. 5, an image of a 2 × 2 pixel block is obtained. The information is converted into four coefficients LL, HL, LH, and HH, and conversion for extracting an average value and an edge component in each direction is performed.

The obtained 1st-LL signal is subjected to a two-layer wavelet transform in the same procedure, and the image signal 2nd converted by the filter groups 707 to 712.
-LL, 2nd-LH, 2nd-HL, 2nd-HH are obtained. 2nd-LL is an image signal obtained by calculating an average value of 4 × 4 pixels, and is an image signal in a lower frequency band than the first layer.
Similarly, 2nd-LH is a component in a lower frequency band than 1st-LH, and is an image signal obtained by extracting a horizontal edge in a frequency band of １ ／ of the Nyquist frequency. Similarly, 2nd-HL is a two-layer vertical high-frequency component, which is a lower-frequency vertical edge signal, and 2nd-HH is an image signal obtained by extracting an oblique edge signal.
As described above, the wavelet coefficient signals W (1st-LL to 1st-HH, 2nd-LL to 2
nd-HH).

FIG. 6 shows the configuration of the characteristic amount calculating means 8.
The feature value calculating means 8 receives the output wavelet coefficient signal W and outputs the character feature value detecting means 11 from the coefficient signal W.
The characteristic E of the character image is obtained by
Thus, a feature A of the halftone dot image is detected, and these feature amounts are corrected by the feature amount correcting means 13 and output.

FIG. 7 shows the structure of the character feature amount detecting means 11. The character feature detection means 11 receives one-layer high-frequency components (1st-HL, 1st-LH, 1st-HH), evaluates the magnitude and continuity of the high-frequency components, and extracts character features. . The edge amount determination means 1101 receives the one-layer vertical high frequency component (1st-HL), compares the attention coefficient HL (i, j) at the center position in FIG. 8 with a predetermined threshold value, and determines the magnitude of the edge amount. Is determined.

In the vertical direction continuity determining means 1102, the vertical high frequency component (1st-HL) of the same level is also used.
Is input, and the variance values of the coefficient of interest HL (i, j) and the upper and lower coefficients HL (i, j−1) and HL (i, j + 1) in FIG. 8 are compared with a predetermined threshold value. The continuity of the edge is determined by evaluating that the value is small. That is, when Expressions (1) and (2) are satisfied, the character determination unit 1109 determines that a large vertical edge component exists continuously in the vertical direction and is a character.

(Equation 1)

(Equation 2)

Similarly, for the horizontal component, the magnitude (1103) of the edge amount and the vertical continuity (1104) are determined from the horizontal high-frequency component (1st-LH) of one layer, and this determination is made. It is determined whether or not it is a character based on the result (11
10). That is, the attention coefficient LH (i, j) at the center position and the left and right coefficients LH (i-1, j), HL in FIG.
When Expressions (3) and (4) are satisfied based on the three coefficients of (i + 1, j), a large horizontal edge component exists continuously in the horizontal direction and is determined to be a character.

(Equation 3)

(Equation 4)

Similarly, as shown in FIGS. 10 and 11, the oblique direction high-frequency components (1st-H
Based on H), the character feature amounts in the diagonally right and diagonally left directions are detected by the following equations (5) to (8).

(Equation 5)

(Equation 6)

(Equation 7)

(Equation 8) Four-way character determination means 1109 obtained as described above
OR (1113) of the determination results of .about.1112 is taken and output as character feature value E.

FIG. 12 shows the structure of the halftone dot feature detection means 12. The halftone feature detecting means 12 receives the high frequency components of the first hierarchy (1st-HL, 1st-LH, 1st-HH), counts the number of inflection points in a predetermined area, and determines the feature of the halftone dot. Things. As shown in FIG.
For the HL, only the inflection point in the vertical direction is counted by the vertical inflection point counting means 1201, and for the 1st-LH, only the inflection point in the horizontal direction is counted by the horizontal inflection point counting means 1202. The HH is configured to count inflection points only in the diagonally right and left directions.

FIG. 13 shows the feature amount detection for the 1st-HL.
This will be described with reference to FIG. First, as shown in FIG. 13, a 5 × 5 size window whose center is the target position is set for the 1st-HL coefficient, and the sign of the coefficient is determined. Next, the sign is inspected in the vertical direction, and when the sign changes from + to-or from-to +, counting is performed. In FIG. 14, the signs are inverted at the arrows, and ten inflection points are detected in the 5 × 5 window. In the same manner, the inflection points in the horizontal and oblique directions are counted,
At 205, the count values in all directions are added and output as the halftone feature amount A.

FIG. 15 is a diagram for explaining a process for a line drawing (vertical line), and FIG. 16 is a diagram for explaining a process for a halftone dot image. A window is set for the vertical line as shown in FIG. 15A, and the 5 × 5 one-layer HL component and L
FIG. 15B shows the coefficient sign of the H component. 1
For the st-HL, the inflection points are counted in the vertical direction, but in the example of FIG. Also 1st-
For LH, the inflection points are counted in the horizontal direction,
Since all LH components are originally zero, there is no inflection point, and the output of the horizontal inflection point counting means 1202 is also zero. Although not shown, the HH component is 0 in the diagonal direction as in the horizontal direction.
So there is no inflection point. As described above, the feature value of a halftone dot is detected as 0 for a line drawing.

As described above, the inflection point in the vertical direction is evaluated for the HL component from which the vertical edge is extracted.
By detecting the inflection point in the same direction as the extracted direction component, erroneous detection in the line drawing part can be prevented, and a highly accurate halftone dot feature can be detected.

On the other hand, in the halftone image, FIG.
When a window is installed as shown in (a), 5 × 5 1
As shown in FIG. 16B, the coefficient codes of the hierarchical HL component and the LH component are a mixture of a + coefficient and a -coefficient.
-If the inflection point of -HL is counted in the vertical direction,
The count value also exists when the -LH coefficient is counted in the horizontal direction, and a large halftone dot feature can be detected.

By the way, the above-mentioned character feature amount detecting means 11
In the halftone dot feature value detection means 12, the detection is performed using the highest frequency band signal (one-layer signal). This makes it possible to reduce the number of line memories and buffers as compared with the case of using coefficient signals of two layers. The two-layer signal is a signal obtained from an area extending over four pixels in the sub-scanning direction, and refers to an area wider than a signal obtained from two pixels on one layer, and the scale becomes large. In the present embodiment, from such a viewpoint, the feature amount is calculated using the coefficient signal in the highest frequency band.

Next, the feature amount correcting means 13 binarizes the halftone feature amount A with a predetermined threshold value, and classifies the halftone region into a halftone region and a non-halftone region. Further, a character area and a non-character area result based on the character feature amount E are combined as shown in FIG.

As shown in FIG. 28, only a character and a non-halftone dot region are used as a character region, and an erroneously detected portion obtained by extracting a character feature value is corrected to a picture region by using a halftone feature value.

On the basis of the control signal C output from the characteristic amount calculating means 13, the sharpness controlling means 9 performs an emphasizing process and a smoothing process on the image signal. FIG. 17 shows an embodiment of the sharpness control means 9. As shown in FIG. 17, according to the result of the feature amount calculating means 8, the high-frequency component (1st-HL, 1
The correction means (902 to 907) performs predetermined correction on the st-LH, 1st-HH) and the high frequency components (2nd-HL, 2nd-LH, 2nd-HH) of the two layers. The correction by these correction means will be described with reference to FIG.

As shown in FIG. 18, the correction means (902-902)
907), the absolute value of the high-frequency component coefficient signal is
When the value is smaller than a predetermined noise removal threshold (Thn), the correction is performed so that the value of the corrected coefficient signal becomes zero. When the absolute value of the high-frequency component coefficient signal is equal to or greater than a predetermined noise removal threshold Thn, the input coefficient signal is multiplied by a predetermined enhancement coefficient α and output.

By performing such control, high-frequency components smaller than the noise elimination threshold Thn are removed as noise, resulting in a smoothed image, and high-frequency components equal to or greater than the noise elimination threshold Thn are α-fold (α>) as signals. 1)
As a result, the difference component increases, and an image subjected to the enhancement processing is obtained.

Note that Thn of the correction means 902 to 907,
α represents different frequency bands (1st, 1st,
2nd), and can be individually set for different direction components (HL, LH, HH). That is, since the magnitude and the degree of enhancement of the high-frequency component to be removed can be controlled for each band and direction, it is possible to perform fine noise removal (smoothing) and enhancement.

Further, in the embodiment of the present invention, different parameter groups (Thn, α) are stored in the character area and the picture area, and these parameter groups are switched according to the result of the characteristic amount calculating means. Make up. That is, the emphasis coefficient α in the character area is larger than that in the picture area, and is set so that sufficient sharpness can be obtained in the character or line drawing part. Conversely, by setting the noise removal threshold Thn in the character area to a value smaller than that in the picture area, a relatively minute change in density can be emphasized, and the sharpness of the character or line drawing portion can be satisfied. .

The band is set so that the degree of enhancement is highest for the high frequency components of the two layers. In a general original, sufficient sharpness can be obtained for a small character image such as a 6-point Mincho font by increasing the emphasis near the image frequency corresponding to 6 lines / mm. Even if excessive emphasis is applied to a higher frequency band, noise components may be emphasized, which is not preferable.

This embodiment is based on an image signal read by a scanner having a resolution of 600 dpi.
The high-frequency component corresponding to this number / mm is a signal of two layers (2nd
-HL, 2nd-LH, 2nd-HH).
The emphasis coefficient of the hierarchy is set to the highest. It is sufficient to appropriately set which layer has the highest enhancement coefficient according to the resolution of the scanner.

In the picture area, a relatively small emphasis coefficient is applied in order to suppress the occurrence of moire in the halftone dot image portion. In particular, a large value is set for the noise removal threshold Thn for one layer in order to remove a halftone dot structure and to improve the graininess of a halftone dot image having a high screen ruling of 150 lines or more.

By setting the parameters as described above, it is possible to control the sharpness of the character area and the picture area (mainly the halftone image area) so as to achieve both image qualities.

In the present embodiment, the extracted feature value is subjected to binary threshold processing, so that the two
Although the example in which the regions are separated and the sharpness control is performed in a binary manner has been described, it is also possible to perform the multi-stage sharpness control using the multi-stage feature amounts. For example, for an area determined to be a character, the size of the character feature is held in multiple stages, and the emphasis coefficient is controlled stepwise from 1 to α based on the size of the character feature. Here, the magnitude of the character feature amount is, for example, 1st-HL, 1st-LH, 1st-H.
The maximum value of the absolute value of H may be used. As described above, by performing the multi-step sharpness control in accordance with the multi-step feature amount, the degree of emphasis becomes smooth, and high-quality image quality is obtained.

In the present embodiment, as a sharpness control method for a frequency band lower than the second layer, the convolution calculator 9 for the second layer low frequency component (2nd-LL) is used.
01. This is to hold a few lines of the 2-layer-LL signal and perform filtering as performed on a real space image. For example, by applying an emphasis filter as shown in FIG. 19, emphasis of a lower frequency band can be performed. Further, with this configuration, the sharpness of the low frequency band can be controlled with a smaller number of line memories. Of course, wavelet transform up to three layers is performed,
There is no problem even if the noise removal threshold value and the enhancement coefficient are set as in the case of the two layers, and fine sharpness control can be performed on the signal obtained by dividing the band, which is preferable in terms of image quality.

Returning to FIG. 2, the data from the sharpness control means 9 is input to the band synthesizing means 10 and inversely converted into a real space image. FIG. 20 shows the configuration of the band combining means 10. The band synthesizing unit 10 performs processing from a higher-layer wavelet coefficient signal. The two-layer coefficient signals (2nd-LL ', 2nd-H) corrected by the sharpness processing means 9
L ′, 2nd−LH ′, 2nd−HH ′) are inversely transformed in the y direction by inverse transform filters H * (y) and G * (y), and further transformed by H * (x) and G * (x). The inverse transform is performed in the x direction. The obtained image signal is the corrected one-layer LL signal (1st-LL '), and the other corrected coefficient signals (1st-HL', 1st-LH ', 1st-H) of one layer.
Along with H ′), similar band synthesis processing is performed. Thus, the real space image signal S ′ after the filtering process is obtained.

By the way, the wavelet transform in the present embodiment has a configuration in which the sub-sampling (pixel thinning-out) performed in the normal compression processing or the like is not performed. As shown in FIG. 21, a block diagram of a wavelet transform for performing sub-sampling is to perform processing by a high-pass filter and a low-pass filter, then perform processing of thinning out one pixel to two pixels (for example, processing of thinning out odd-numbered pixels), Signal. In the inverse transform, up-sampling is performed, and inverse transform is performed by an inverse transform filter.

The wavelet transform of this embodiment is shown in FIG.
As shown in (1), downsampling is not performed during forward conversion. Therefore, as shown in FIG. 23, the image size of the coefficient signal in each layer and in each direction is the same as the input image Org. At the time of inverse conversion, up-sampling is performed for each even and odd pixel group as shown in FIG. An inverse conversion result from the even pixel group and an inverse conversion result from the odd pixel group are obtained for one original image pixel, and these are averaged to obtain image data after inverse conversion.

By the above-described processing without performing sub-sampling, it is possible to calculate an image feature amount with high accuracy, and to perform high-quality filter processing without emphasis / smoothness unevenness. This situation will be described with reference to FIGS. FIGS. 24 to 26 are signals in which the original image is the same, and the shading is repeated at a cycle of four pixels. 24 and 2
FIG. 5 shows an example in which subsampling is performed.
Shows the method of this embodiment in which subsampling is not performed. FIGS. 24 and 25 show an example in which the pixels to be sub-sampled are shifted by one pixel.

In the case of FIG. 24, the pixel pair of d1 and d2, d3
This is an example in which the wavelet coefficient signal performed for the pixel pair of d0 and d4 is thinned out.
The coefficient signal for the pixel pair 2 and d3 remains. For example, for the emphasis filter processing, when the high frequency component is multiplied by twice the emphasis coefficient and inversely transformed, the high frequency components d0 and d1
0 is amplified by a factor of two, and the data difference between d0 'and d1' after the inverse conversion is doubled, indicating that the desired enhancement processing has been performed.

On the other hand, when the sub-sampling is shifted by one pixel as shown in FIG. 25, the high-frequency component obtained by the pixel pair of d1 and d2 is 0, and the high-frequency component is not amplified even when multiplied by the enhancement coefficient. As a result, the result is not different from the original signal at all, and it can be seen that the desired enhancement processing has not been performed. In such a conversion system that performs sub-sampling, there are cases where the frequency characteristics cannot be corrected correctly depending on the sampling position.

The solution to this problem is the wavelet transform without subsampling according to the present embodiment.
As shown in FIG. 6, the result of FIG. 24 and FIG. 25 is averaged, and the frequency characteristic can be corrected without any leakage. In addition, not only the enhancement processing, but also the calculation of the feature amount of the image, the signal is not subjected to sub-sampling, so that the feature amount can be extracted with high accuracy.

In the present embodiment, a Haar type wavelet is used as an example of a wavelet basis function, but other basis functions can be similarly realized. Further, the present invention is not limited to the wavelet transform, and can be similarly realized for a sub-band transform, a Fourier transform, a Hadamard transform, a Laplacian pyramid, and the like that divide an image into a plurality of frequency bands.

In the above-described embodiment, the configuration in which the image attribute is determined with respect to the wavelet transform coefficient has been described. However, as shown in FIG. Alternatively, the filter processing for converting the spatial frequency may be performed in the wavelet coefficient space.

As described above, the present invention may be implemented by dedicated hardware, but may be implemented by software using a general-purpose computer system. When implemented by software, a program for realizing the image processing function (processing such as filter processing, γ conversion, and halftone processing) and the processing procedure of the present invention is recorded on a recording medium or the like. The image processing function of the present invention is implemented when the program is read into the computer system and executed by the CPU. The image data is, for example, document image data read from a scanner or the like, image data prepared in a hard disk or the like in advance, or image data captured via a network. The processing result is output to a printer or written to a hard disk. Alternatively, it is output to an external device (such as a printer) via a network.

[0059]

As described above, according to the present invention, the following effects can be obtained. (1) Since the optimal frequency characteristics are corrected in all regions according to the detected image attributes, high image quality can be realized for a document in which different image attributes such as a character image and a halftone image are mixed. . (2) When dividing an image signal into each band and each direction, downsampling is not performed, and a plurality of (ev)
en, odd), by averaging the post-transform data, there is no missing data (coefficient signal), the frequency characteristic can be corrected with high accuracy, and the image attribute can be determined with high accuracy. it can. (3) High-quality spatial frequency correction can be performed on a character image, a halftone dot image, and a document image in which characters and halftone dots are mixed. (4) Since the feature amount is extracted by using the coefficient signal of the highest frequency band (one layer) and the image attribute is determined, the number of line memories and buffers to be used can be reduced, and the cost is reduced. An apparatus and a method can be realized. (5) Since the characteristic signal of the specific direction and the inflection point of the specific direction are detected when extracting the characteristic amount of the character image and the halftone image, the character and the halftone characteristic amount are extracted with high accuracy. Can be
High image quality can be realized. (7) Since the convolution operation is performed on the low frequency components in the lowest frequency band, the sharpness of the low frequency band can be controlled with a smaller number of line memories.

[Brief description of the drawings]

FIG. 1 shows a configuration of an embodiment of the present invention.

FIG. 2 shows a configuration of a filter processing unit of the present invention.

FIG. 3 shows a configuration of a band dividing unit.

FIG. 4 shows a Haar wavelet.

FIG. 5 is a diagram illustrating a Haar wavelet transform.

FIG. 6 shows a configuration of a feature amount calculating unit.

FIG. 7 shows a configuration of a character feature amount detection unit.

FIG. 8 shows a mask for determining a vertical edge amount and continuity.

FIG. 9 shows a mask for determining a lateral edge amount and continuity.

FIG. 10 shows a mask for determining the edge amount and continuity in the diagonally right direction.

FIG. 11 shows a mask for determining the edge amount and continuity in the diagonally left direction.

FIG. 12 shows a configuration of a halftone dot feature amount detection unit.

FIG. 13 shows an inflection point detection mask (window).

FIG. 14 shows locations where coefficient signs in the inflection point detection mask change.

FIG. 15 is a diagram illustrating detection of an inflection point for a line drawing.

FIG. 16 is a diagram illustrating detection of an inflection point in a halftone image.

FIG. 17 shows a configuration of sharpness control means.

FIG. 18 shows an example of input / output characteristics of a correction unit.

FIG. 19 shows an example of an emphasis filter.

FIG. 20 shows a configuration of a band combining means.

FIG. 21 shows a configuration of a wavelet transform for performing subsampling.

FIG. 22 shows a configuration of a wavelet transform without performing subsampling according to the present invention.

FIG. 23 is a diagram illustrating a wavelet transform without performing subsampling.

FIG. 24 is a diagram illustrating a wavelet transform for performing sub-sampling.

FIG. 25 is a diagram illustrating a wavelet transform for performing sub-sampling.

FIG. 26 is a diagram illustrating wavelet transform according to the present embodiment without performing subsampling.

FIG. 27 shows a configuration of another embodiment of the present invention.

FIG. 28 shows an overall determination result determined based on a character and a dot feature amount.

FIG. 29 shows an example of setting to correction means corresponding to a character and picture area.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 image input means 2 scaling processing means 3 filter processing means 4 γ conversion processing means 5 halftone processing means 6 image output means

 ──────────────────────────────────────────────────続 き Continuing from the front page (72) Inventor Etsuro Morimoto 1-3-6 Nakamagome, Ota-ku, Tokyo F-term in Ricoh Co., Ltd. 5B057 AA11 CA07 CA12 CA16 CB07 CB12 CB16 CC01 CE03 CG09 DA17 DC16 5C077 LL19 MP06 NN04 PP03 PP27 PP49 TT02 TT06

Claims (16)

[Claims] 1. A band dividing unit that divides a predetermined image into a plurality of directional component coefficients for each of a plurality of frequency bands and outputs the divided coefficients, and calculates a predetermined feature amount of the image based on the divided coefficients. Amount calculation means, and a sharpness control means for correcting a frequency characteristic by performing a predetermined correction on each of the coefficients for each band in accordance with the calculated predetermined feature amount,
An image processing apparatus comprising: a band synthesizing unit that inversely converts an output from the sharpness control unit to obtain a processed image. 2. When dividing a predetermined image into a plurality of directional component coefficients for each of a plurality of frequency bands and outputting the divided image, at least one frequency band is output without thinning out pixels of the predetermined image. Band dividing means, a characteristic amount calculating means for calculating a predetermined characteristic amount of the image based on the divided coefficient, and for each of the coefficients for each band according to the calculated predetermined characteristic amount. Sharpness control means for correcting the frequency characteristic by performing a predetermined correction, and band synthesis for obtaining a processed image based on a plurality of inverse transform outputs obtained for one pixel from the sharpness control means. And an image processing apparatus. 3. The image processing apparatus according to claim 1, wherein the characteristic amount calculating unit includes a character characteristic amount detecting unit that detects a characteristic amount of a character image as the predetermined characteristic amount. 4. The image processing apparatus according to claim 1, wherein the feature amount calculating unit includes a halftone feature amount detecting unit that detects a feature amount of a halftone image as the predetermined feature amount. apparatus. 5. The feature amount calculating unit includes a character feature amount detecting unit that detects a feature amount of a character image as the predetermined feature amount, and a halftone dot that detects a feature amount of a halftone image as the predetermined feature amount. And a feature amount correcting unit for correcting a feature amount based on a detection result of the character feature amount detecting unit and a detection result of the halftone dot feature amount detecting unit. Item 3. The image processing apparatus according to item 1 or 2. 6. The image processing apparatus according to claim 1, wherein the characteristic amount calculating unit calculates the predetermined characteristic amount using a coefficient of a highest frequency band. 7. The image processing apparatus according to claim 3, wherein the character feature amount detecting unit detects a feature amount of the character image by evaluating a coefficient size and a coefficient variance. 8. The image processing apparatus according to claim 7, wherein a plurality of coefficients used when evaluating the variance of the coefficients are selected from a group of coefficients in the direction for each direction component. 9. The image processing apparatus according to claim 4, wherein said halftone dot feature amount detecting means detects a halftone dot feature amount based on the number of inflection points of coefficients in a predetermined area. 10. The image processing apparatus according to claim 9, wherein the number of inflection points is a number obtained by counting inflection points in the direction for each direction component. 11. The image processing apparatus according to claim 1, wherein said sharpness control means includes means for performing a convolution operation on at least one frequency band among the divided coefficients. apparatus. 12. The image processing apparatus according to claim 11, wherein the means for performing the convolution operation performs a convolution operation on a low-frequency component in the lowest frequency band. 13. A step of inputting a predetermined image; a step of dividing the image into a plurality of directional component coefficients for each of a plurality of frequency bands; and outputting the divided image. Calculating a characteristic amount, correcting the frequency characteristic by performing a predetermined correction on each of the coefficients for each of the bands according to the calculated predetermined characteristic amount, and outputting the corrected output. An image processing method, comprising the steps of: inverting and restoring an image; and outputting the restored image. 14. A step of inputting a predetermined image and, when the image is divided into a plurality of directional component coefficients for each of a plurality of frequency bands and outputted, at least one frequency band includes pixels of the predetermined image. Performing the output without thinning out, and calculating a predetermined feature amount of the image based on the divided coefficients; and Correcting the frequency characteristic by performing a predetermined correction on the coefficient; restoring an image based on a plurality of inverse transform outputs obtained for one pixel after the correction; Outputting the processed image. 15. A program for causing a computer to execute each step of the image processing method according to claim 13. Description: 16. A computer-readable recording medium on which a program for causing a computer to execute each step of the image processing method according to claim 13 or 14 is recorded.