JP4972122B2 - Image processing device - Google Patents

Description

The present invention relates to an image processing equipment that provides both edge enhancement and noise reduction of an image signal.

  In an image processing apparatus incorporated in a digital multi-function peripheral, a printer, or the like, an image to be read includes an image of only characters, an image of only a photograph or a picture, or an image in which characters, a photograph, and a picture are mixed. In order to improve the reproducibility of the read image, the image processing apparatus performs an image correction process. One of the processes for correcting the read image is a filter process. Filter processing increases the amount of edges for character images to sharpen characters, reduces background noise for photo and picture images, smoothes halftone dots, and reduces moire Effect is obtained.

  Conventionally, a pixel block including a plurality of pixels adjacent to each other along a predetermined direction including a target pixel of image data is extracted, and an edge amount in the pixel block is detected using a primary differential filter. Next, the detected edge amount is corrected so as to increase the edge amount of the pixel adjacent to the pixel block. Next, a technique is known in which the mixing ratio between the image data from the edge enhancement filter and the image data from the smoothing filter is changed based on the corrected edge amount of the pixel (see Patent Document 1). As a result, halftone image data and photographic image data can be smoothed and smoothly reproduced while maintaining sharp edges.

JP-A-5-307603

  However, the technique disclosed in Patent Document 1 detects an edge amount from an image using a first-order differential filter, binarizes the detected edge amount with a predetermined threshold, and determines whether edge pixels are continuous. To do. The threshold value for binarization must be given in advance by the designer.

  In addition, in the case of an image having a certain period such as a ladder pattern that is a pattern in which black and white lines are arranged, it is very similar to a halftone dot pattern, and thus smoothing processing is performed by misrecognizing a halftone dot. Therefore, the edge of the line of the ladder pattern may become dull.

The present invention has been made in view of such points, and does not require a specific threshold value or complicated processing, and can achieve both effects of edge enhancement and noise reduction adapted to image characteristics. the image of characters and lines, and an object thereof is to provide an image processing equipment which can be compensated to a sharper image by edge enhancement.

An image processing apparatus according to the present invention is an image processing apparatus that determines the characteristics of an image such as a character, line, photograph, or picture in an image signal. A pixel extraction unit for extracting the pixel block, a luminance calculation unit for calculating a luminance value of each pixel in the pixel block extracted by the pixel extraction unit, and a luminance value of each pixel calculated by the luminance calculation unit A pixel block in the predetermined range extracted by the pixel extraction unit using a standard deviation calculation unit for obtaining a standard deviation value from a difference in luminance value of the target pixel and the standard deviation value calculated by the standard deviation calculation unit for, was performed and a filter processing unit that performs filter processing using the shape obtained by inverting the Gaussian distribution element of the bilateral filter filters, the filtering by the filter processing unit An output unit for outputting the pixel block, and when the standard deviation value calculated by the standard deviation calculation unit is large, the filter processing unit performs edge enhancement processing on the pixel block, and the standard deviation When the value is small, noise reduction processing is performed on the image block.

According to the image processing equipment according to the present invention, without requiring a certain threshold and complex processing, it is possible to achieve both the effect of edge enhancement and noise reduction adapted to the characteristics of the image, in particular characters or lines In the image of (1), the edge can be emphasized and corrected to a clearer image.

1 is a block diagram showing a schematic configuration of an image forming apparatus according to an embodiment of the present invention. The block diagram which shows schematic structure of the image adjustment part in embodiment same as the above. The figure which shows an example which performs pixel extraction from the image in embodiment same as the above. The flowchart which shows the operation | movement which calculates the standard deviation value of the extracted image range in embodiment same as the above. The figure which shows the waveform of the Gaussian distribution of the bilateral filter in embodiment same as the above. The figure which shows the waveform of the inverted Gaussian distribution of the bilateral filter in embodiment same as the above. The figure which shows the waveform which multiplied the Mexican hat filter and Gaussian distribution in embodiment same as the above. The flowchart which shows operation | movement of the 3rd filter processing example in embodiment same as the above.

  Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

  FIG. 1 is a diagram illustrating a system configuration example of an image forming apparatus 1 according to an embodiment of the present invention, for example, a digital MFP (Digital Multi-Function Peripheral). The image forming apparatus 1 functions as a digital color multifunction peripheral. For example, a copy function, a printer function, a scanner function, a FAX function, a storage function, and the like are realized.

  The image forming apparatus 1 includes a reading unit 11, an image processing unit 12, a page memory unit 13, and an image recording unit 14, and a copy function is realized by these configurations. The reading unit 11 optically reads a document with, for example, a three-line CCD sensor, and converts it into red (R), green (G), and blue (B) color digital image signals.

  The image processing unit 12 includes a scanner image processing unit 21, a printer image processing unit 22, and an image adjustment unit 20.

  The scanner system image processing unit 21 performs various image processing such as shading correction for correcting non-uniformity of the signal level in the main scanning direction. Further, when operating as a copy function, the page memory unit 13 converts the primary colors of R, G, and B into, for example, cyan (C), magenta (M), yellow (Y), and black (K) color signals. On the other hand, when it operates as a scanner function, it outputs to the page memory unit 13 as three primary colors of R, G, and B.

  The printer system image processing unit 22 performs image processing for printing, for example, γ correction processing and gradation processing on the image signal output from the page memory unit 13, and outputs the image signal to the image recording unit 14.

  The image adjustment unit 20 performs noise reduction processing and edge enhancement processing on the image signal input from the reading unit 11 or the page memory unit 13. Details will be described later.

  The page memory unit 13 temporarily stores the image signal output by the scanner system image processing unit 21 in units of pages, for example, and outputs the image signal to the image processing unit 12 again.

  The image recording unit 14 is a component that prints an image on a recording sheet by, for example, an electrophotographic method, and includes an exposure device, a photosensitive drum, a developing device, and the like (all not shown).

  The image forming apparatus 1 includes a storage unit 15, a printer controller unit 16, a FAX controller unit 17, an electronic data creation unit 18, and an external I / F unit 19. These components are used when the image forming apparatus 1 is operated as a storage function, a printer function, a FAX function, a scanner function, and the like.

  The storage unit 15 is configured by, for example, an HDD (Hard Disk Drive), and stores an image signal read by the reading unit 11 and an image signal input from an external personal computer or the like. The image signal stored in the storage unit 15 can be read out as necessary, printed by the image recording unit 14, or output to the outside as scan data.

  When the image forming apparatus 1 operates as a scanner function, the electronic data creation unit 18 converts the image signal into a predetermined data format and generates scan data. The generated scan data is output to the outside via the external I / F unit 19.

  The FAX controller unit 17 operates the image forming apparatus 1 as a FAX function. The image signal read by the reading unit 11 is converted into a data format for FAX and output to an external telephone line via the external I / F unit 19. In addition, FAX data input from a telephone line can be converted into an image signal, and printing can be performed by the image recording unit 14.

  The printer controller unit 16 performs various controls and various processes necessary for the print function when the image forming apparatus 1 operates as a printer function. The control unit 10 performs control related to the entire image forming apparatus 1.

  FIG. 2 shows a configuration inside the image adjustment unit 20. The image adjustment unit 20 includes a pixel extraction unit 201, a luminance calculation unit 202, a standard deviation calculation unit 203, and a filter processing unit 204. The pixel extraction unit 201 extracts a predetermined range of pixel blocks from an image with an arbitrary target pixel as the center. The luminance calculation unit 202 calculates the luminance value of each pixel. The standard deviation calculation unit 203 calculates the standard deviation value of the difference in luminance value between the target pixel and other pixels in the pixel block within the predetermined range extracted by the pixel extraction unit 201. The filter processing unit 204 performs filter processing on the pixel block extracted by the pixel extraction unit 201 using the standard deviation value obtained by the standard deviation calculation unit 203. A pixel block having a large standard deviation value is regarded as a character, and a filter process for emphasizing a high frequency component is performed to emphasize an edge. A pixel block having a small standard deviation value is regarded as a photograph or a picture, and is subjected to a filter process that intensifies low frequency components (suppresses high frequency components) in order to reduce noise.

  FIG. 3 shows an example of extracting pixels from an image, and FIG. 4 shows a flowchart thereof. When the reading unit 11 scans and reads an image (step S101), the pixel extraction unit 201 has a pixel in a range of 5 × 5 around an arbitrary target pixel (gray shaded portion in the example of FIG. 3). A block is extracted (step S103). The luminance calculation unit 202 calculates the luminance value of each pixel in the pixel block extracted by the pixel extraction unit 201 (step S105). The standard deviation calculation unit 203 calculates a standard deviation value of a difference in luminance value between the target pixel and each of the other pixels from the calculated luminance value of each pixel (step S107). The filter processing unit 204 performs filter processing that reflects the characteristics of the image based on the standard deviation value calculated for the image of the pixel block, and performs appropriate processing on the image. An example of filter processing will be described below.

<First Filter Processing Example>
A modified bilateral filter will be described as a first example of a filter used in the filter processing unit 204.

First, the bilateral filter is expressed by the following equation (1).

In equation (1), let the pixel of interest be I (i, j), and I (i + m, j + n) be any neighboring pixel of the pixel of interest. Here, m and n are pixel displacement amounts from neighboring pixels, where m is a displacement amount in the scanning direction (i direction) and n is a displacement amount in the sub-scanning direction (j direction). Further, σ ₁ is a standard deviation value of the spatial filter, and σ ₂ is a standard deviation value of a difference in luminance value in the extracted pixel block. The numerator second term in the equation (1) indicates a spatial filter, and the numerator third term indicates a Gaussian distribution. The denominator is an expression for normalization.

  In the bilateral filter, the waveform of the Gaussian distribution that controls the spatial filter changes depending on the difference between the luminance value of the target pixel and the luminance value of the neighboring pixel. FIG. 5 shows a waveform of a Gaussian distribution in the bilateral filter. In FIG. 5, the shape is convex upward. The Gaussian distribution draws a gentle waveform as the difference in luminance value increases, and draws a steeper waveform as the difference in luminance value decreases. In FIG. 5, the horizontal axis is the space axis, but the same tendency occurs when the horizontal axis is replaced with the frequency axis.

  When a low-pass filter is selected as the spatial filter (numerator second term in equation (1)), and this Gaussian distribution is combined with the low-pass filter that is a spatial filter and multiplied, the filter application range of the low-pass filter changes. Therefore, when a waveform with a steep Gaussian distribution is drawn, the effect of the low-pass filter is strong, and when a waveform with a gentle Gaussian distribution is drawn, the effect of the low-pass filter is weak. Thus, the magnitude of the effect of the low-pass filter can be controlled by the Gaussian distribution.

In order to apply the effect of this low-pass filter provided by the bilateral filter's Gaussian distribution to the effect of a high-pass filter that emphasizes high-frequency components, the bilateral filter's Gaussian distribution is inverted (the bilateral filter's Gaussian distribution is inverted). ) Is shown in equation (2), and its shape is shown in FIG.

  The formula (2) is a form in which the portion indicating the Gaussian distribution of the third term of the numerator of the formula (1) is subtracted from 1, and as shown in FIG. It has a shape. The Gaussian distribution draws a steeper waveform as the difference in luminance value increases, and draws a gentle waveform as the difference in luminance value decreases. In FIG. 6, the horizontal axis is the space axis, but the same tendency occurs when the frequency axis is replaced.

  When a high-pass filter is selected as the spatial filter (numerator second term in equation (1)) and this inverted Gaussian distribution is combined with the high-pass filter, which is a spatial filter, the adaptive range of the filter is reversed, and the larger the difference, A filter is applied. Therefore, when a waveform having a steep Gaussian distribution is drawn, the effect of the high-pass filter is strong. When a waveform having a gentle Gaussian distribution is drawn, the effect of the high-pass filter is weak.

  Therefore, by using the “bilateral filter Gaussian distribution inversion type” in equation (2), the edge enhancement effect, which is the effect of the high-pass filter, is enhanced for characters and lines with a large difference in luminance value, and a clearer image. Is output. On the other hand, the edge emphasis effect is weakened and a smooth image is output for a photograph or the like having a small difference in luminance value.

<Second filter processing example>
As a second example of a filter used in the filter processing unit 204, an example using a Mexican hat filter will be described. The expression (3) is obtained by multiplying the Gaussian distribution of the bilateral filter of the expression (1) by a Mexican hat filter obtained by normalizing the difference in luminance value with the standard deviation value σ ₁ of the spatial filter.

Here, a is a control coefficient, which is proportional to the standard deviation value σ ₂ of the luminance value. FIG. 7 shows a waveform of an expression obtained by multiplying the Mexican hat filter (the numerator third term in the expression (3)) and the Gaussian distribution (the numerator fourth term in the expression (3)) depending on the control coefficient a. When the difference in luminance value is 0, the term of the Mexican hat filter is 0 and only the Gaussian distribution is obtained. Even when the change in luminance value is small compared to the standard deviation value σ ₁ of the spatial filter, the elements of the Mexican hat filter are small, so that the waveform is close to a Gaussian only waveform, and the effect of the low-pass filter is enhanced. When the difference in luminance value is larger than the standard deviation value σ ₁ of the spatial filter, the element of the Mexican hat filter becomes strong.

  The control coefficient a indicates the width of the waveform in FIG. When the control coefficient a is small, the waveform width becomes large, approaching a Gaussian distribution, and the effects of the high-pass filter and the low-pass filter are weak. When the control coefficient a is large, the waveform becomes a Mexican hat type, and the effect of the high-pass filter and the effect of the low-pass filter become stronger.

  Therefore, by using the Mexican hat filter of equation (3), both the high-pass filter effect and the low-pass filter effect can be obtained. Characters and lines are clearly enhanced by edge enhancement, and photographs and pictures are reduced by noise. It is output smoothly.

<Third filter processing example>
As a third example of the filter used in the filter processing unit 204, switching between the low-pass filter effect and the high-pass filter effect based on the standard deviation value will be described with reference to the flowchart of FIG.

The standard deviation calculation unit 203 normalizes the standard deviation value obtained in step S107 in FIG. 4 to obtain the control coefficient n. This control coefficient n is used as a weight in the spatial filter (low-pass filter and high-pass filter). The filter processing unit 204 gives the filter coefficient of the equation (4) using the control coefficient n (step S201).

  Here, HPF is a term indicating a high-pass filter, and LPF is a term indicating a low-pass filter. From the equation (4), when the control coefficient n is large, that is, when the standard deviation value is large, it is considered that the extracted image has characters and lines, and the filter processing ((4 ) HPF term in the formula. Conversely, when the control coefficient n is small, that is, the standard deviation value is small, it is assumed that the extracted image has a picture or picture, and a filter process (Equation (4)) that emphasizes low frequency components to reduce noise. LPF)).

  Next, the filter processing unit 204 multiplies the 5 × 5 pixels extracted in step S103 by Filter [x] [y] calculated by the equation (4) (step S203). At this time, coordinates correspond to the pixel I [x] [y] to be multiplied by Filter [x] [y].

Next, as shown in the following equation (5), the filter processing unit 204 calculates the sum of the results of multiplication by each pixel, and sums up Filter [x] [y] (displayed as SUM_Filter in equation (5)). Divide by (step S205). Let the output image obtained in step S205 be g [x] [y].

  The above operation is performed on the entire image. Specifically, when the processing in a certain image range is completed, the image range to be extracted next is shifted by one pixel in either the scanning direction or the sub-scanning direction, and the standard deviation value of the luminance value is obtained. When the processing for the entire image is completed (“Yes” in step S207), the image adjustment unit 20 outputs an image to the image recording unit 14 (step S209).

  Thereby, the weight of the filter is changed by the control coefficient n based on the standard deviation value from the target pixel in the extracted image range, and the effect of the low-pass filter and the effect of the high-pass filter can be switched depending on the image. Both enhancement and noise reduction effects can be realized simultaneously.

  Note that the specific examples of the low-pass filter and the high-pass filter in the equation (4) may be any filter as long as the effect is obtained. Further, in the equation (4), when only the edge enhancement is to be performed, it is only necessary to set n = 1 and remove the LPF term. On the other hand, if only noise reduction is desired, n = 0 can be used to remove the HPF term.

  As described above, according to the present embodiment, the standard deviation value of the difference in luminance value from the target pixel in an arbitrary image range is used for the filter process, and the edge enhancement effect for characters and the like and the noise for photos and the like Both reduction effects can be obtained. Further, it is effective in reducing moire, and characters can be emphasized and output while suppressing moire.

  In particular, in a character or line image, the edge can be emphasized to obtain a clearer image. Specifically, a bilateral filter Gaussian inversion type, a Mexican hat filter, a low-pass filter, and a high-pass filter having a filter coefficient weighted by standard deviation are effective.

  As mentioned above, although embodiment of this invention was explained in full detail with reference to drawings, the specific structure is not restricted to this embodiment, The design change etc. of the range which does not deviate from the summary of this invention are included.

  DESCRIPTION OF SYMBOLS 1 ... Image forming apparatus, 10 ... Control part, 11 ... Reading part, 12 ... Image processing part, 13 ... Page memory part, 14 ... Image forming part, 15 ... Storage part, 16 ... Printer controller part, 17 ... FAX controller part , 18 ... Electronic data creation unit, 20 ... Image adjustment unit, 21 ... Scanner system image processing unit, 22 ... Printer system image processing unit, 201 ... Pixel extraction unit, 202 ... Luminance calculation unit, 203 ... Standard deviation calculation unit, 204 ... Filter processing unit.

Claims (3)

In an image processing apparatus that determines the characteristics of an image such as characters, lines, photographs, and pictures in an image signal,
A pixel extraction unit that extracts an arbitrary pixel of interest and a predetermined range of pixel blocks centered on the pixel of interest from the image signal;
A luminance calculation unit that calculates the luminance value of each pixel in the pixel block extracted by the pixel extraction unit;
A standard deviation calculating unit for obtaining a standard deviation value from a difference between the luminance value of each pixel calculated by the luminance calculating unit and the luminance value of the target pixel;
Using the standard deviation value calculated by the standard deviation calculation unit, the pixel block in the predetermined range extracted by the pixel extraction unit is subjected to filter processing using a filter in which a Gaussian distribution element of a bilateral filter is inverted. A filter processing unit to perform;
An output unit that outputs the pixel block that has been filtered by the filter processing unit;
Have
The filter processing unit performs edge enhancement processing on the pixel block when the standard deviation value calculated by the standard deviation calculation unit is large, and performs noise reduction processing on the image block when the standard deviation value is small. Do,
An image processing apparatus. In an image processing apparatus that determines the characteristics of an image such as characters, lines, photographs, and pictures in an image signal,
A pixel extraction unit that extracts an arbitrary pixel of interest and a predetermined range of pixel blocks centered on the pixel of interest from the image signal;
A luminance calculation unit that calculates the luminance value of each pixel in the pixel block extracted by the pixel extraction unit;
A standard deviation calculating unit for obtaining a standard deviation value from a difference between the luminance value of each pixel calculated by the luminance calculating unit and the luminance value of the target pixel;
Using the standard deviation value calculated by the standard deviation calculation unit, for the pixel block in the predetermined range extracted by the pixel extraction unit, a bilateral filter multiplied by a Mexican hat filter element was used. A filter processing unit for performing filter processing;
An output unit that outputs the pixel block that has been filtered by the filter processing unit;
Have
The filter processing unit performs edge enhancement processing on the pixel block when the standard deviation value calculated by the standard deviation calculation unit is large, and performs noise reduction processing on the image block when the standard deviation value is small. Do,
An image processing apparatus. In an image processing apparatus that determines the characteristics of an image such as characters, lines, photographs, and pictures in an image signal,
A pixel extraction unit that extracts an arbitrary pixel of interest and a predetermined range of pixel blocks centered on the pixel of interest from the image signal;
A luminance calculation unit that calculates the luminance value of each pixel in the pixel block extracted by the pixel extraction unit;
A standard deviation calculation unit that obtains a standard deviation value from the difference between the luminance value of each pixel calculated by the luminance calculation unit and the luminance value of the target pixel, and obtains a normalized standard deviation obtained by normalizing the standard deviation as σ ₀ When,
The high-pass filter is weighted by using the σ ₀ calculated by the standard deviation calculation unit, and the low-pass filter is weighted by using (1-σ ₀ ), and the weighted high-pass filter and the low-pass filter A filter processing unit that performs filter processing on pixel blocks in the predetermined range extracted by the pixel extraction unit using filter coefficients synthesized by a combination of
An output unit that outputs the pixel block that has been filtered by the filter processing unit;
Have
The filter processing unit performs edge enhancement processing on the pixel block when the standard deviation value calculated by the standard deviation calculation unit is large, and performs noise reduction processing on the image block when the standard deviation value is small. Do,
An image processing apparatus.