JP2005175584A - Image processing system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve a problem that color shift cannot fully be corrected when applying a phase correction filtering uniformly in the case that the color shift amount is different for every color and also for every pixel. <P>SOLUTION: The image processing system includes: a pixel shift amount storage means for storing a pixel shift amount depending on pixel positions; and a phase correction filter for correcting a color shift by each pixel by reading the pixel shift amount. Further, the image processing system is configured to include also a sharpness correction filter for receiving a frequency characteristic of the phase correction filter to correct the sharpness thereby correcting the sharpness of the image blurred by the phase correction filter. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to an image processing apparatus and means having an apparatus for reading a color document, such as an image scanner, a copying machine, and a FAX.

In color reading devices used in copiers, image scanners, etc., the position of each color shifts due to lens aberrations, etc., resulting in pixel shifts for each color, which causes image degradation as color shifts. There was. Color misregistration is a phenomenon in which when a black vertical line is read, a color shift occurs due to pixel misregistration in each color. In order to correct the color shift, a phase correction filter that shifts the pixel position may be used for each pixel to correct the pixel shift (for example, Patent Document 1).
Japanese Patent Laid-Open No. 11-127447

  However, when the amount of color shift differs not only for each color but for each pixel, there is a problem that the color shift cannot be corrected by applying a uniform phase correction filter. Furthermore, the phase correction filter is generally realized by using a simple filter coefficient, and there is a problem that sharpness is lowered.

  According to the present invention, the pixel shift amount storage means for storing the pixel shift amount depending on the pixel position is provided, and the phase correction filter for reading out the pixel shift amount and correcting the color shift for each pixel is provided. In addition, the image processing apparatus includes a sharpness correction filter that receives the frequency characteristics of the phase correction filter and corrects the sharpness, thereby correcting the sharpness of an image blurred by the phase correction filter. Solve.

As explained above,
It has a pixel shift amount storage means for storing a pixel shift amount depending on the pixel position, and has a phase correction filter for reading out the pixel shift amount and correcting the color shift for each pixel. In addition, the image processing apparatus includes a sharpness correction filter that receives the frequency characteristics of the phase correction filter and corrects the sharpness, thereby correcting the sharpness of an image blurred by the phase correction filter. Better image quality can be achieved by eliminating differences in pixel shift and sharpness.

(First embodiment)
Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 2 shows an image reading apparatus for explaining the present embodiment.

  The scanner unit 200 that is an image input device illuminates an image on paper as a document and scans a CCD line sensor (not shown) to convert it into an electrical signal as raster image data. When the original is set on an original platen glass (not shown) provided in the scanner unit 200 and an apparatus user gives an instruction to read the original paper through an operation unit (not shown), the scanner 200 reads the original as described above. Convert to electrical signal. Further, the original paper is set on the tray 202 of the original feeder 201, and the user of the apparatus gives an instruction to read through an operation unit (not shown). The scanner 200 feeds the original paper one by one and reads the original image. An operation may be performed.

  FIG. 4 shows a platen glass and a three-line image reading sensor. The apparatus user sets an image to be read on the platen glass so that the original is directed downward, is illuminated by illumination means (not shown), and the reflected light is imaged using a lens (not shown) to form a three-line image. It is converted into an electrical signal by a reading sensor, and digital image data is generated by an A / D converter (not shown). The three-line image reading sensor is a photoelectric conversion device such as a CCD device. The pixel array is arranged on a line in the main scanning direction, and each of the three lines has a different color separation filter. For example, each line reads R, G, and B, respectively. Each line image reading sensor generates image data as an electrical signal in the main scanning direction, and scans the three-line image reading sensor in a direction (sub-scanning direction) perpendicular to the main scanning direction. Is generated. In the present embodiment, description will be made assuming that the resolution of the image data is 600 dpi and the luminance level resolution per pixel of each color is 256 gradations, but of course this is not limited to this resolution and luminance level resolution. Not too long.

  Here, the distribution of luminance levels when color misregistration occurs is shown in FIG. When the 3-line image reading sensor has R, G, and B color separation filters, G is 0.3 pixels on the right, B is 0.3 pixels on the left, and R is a diagram when there is no pixel shift. . If the RGB readings are exactly the same, it will be a black and white reading, but in FIG. 5 there is a shift for each color, so on the left side, blue is bright and green is dark, and on the right side is green. Because it is dark and blue becomes bright, it will appear colored. When the distribution of pixel shift amount of each color for each pixel is known due to lens chromatic aberration, etc., the pixel shift is eliminated by phase correction, and the sharpness correction filter corrects the sharpness drop by the phase correction. This is the purpose of this embodiment.

  FIG. 1 is a block diagram that best represents the features of this embodiment.

  101 is an image input unit, 102 is a phase correction filter that corrects a pixel shift amount for each color, and 103 is a phase correction filter coefficient correction that receives a pixel shift amount for each color according to a pixel position and generates a filter coefficient. Means 104, a pixel position information data input unit, 105 stores a pixel shift amount for each color, a pixel shift amount storage unit 106, sharpness correction for correcting blurring by the phase correction filter A filter 107 is a sharpness correction filter coefficient generation unit that generates a sharpness correction filter coefficient according to the amount of blurring by the phase correction filter.

  Continue explaining in more detail. 101. An image data read from a document through a three-line reading sensor from an image input unit and converted into a digital image signal by A / D conversion is input for each color.

  The 102 phase correction filter corrects pixel shift with the following configuration.

IMG1 '= (A * IMG0 + B * IMG1 + C * IMG2) / 32 (Formula 1)
IMG1 is a pixel of interest. IMG0 is the left pixel of the target pixel. IMG2 is the right pixel of the target pixel. A, B, and C are coefficients of the phase correction filter, which are generated and supplied by the 103 phase correction filter coefficient generation means described later. IMG1 ′ becomes the image data after the phase correction filter.

  Here, in this embodiment, it is assumed that a normalized coefficient is supplied so that A + B + C = 32. The normalization number is preferably a power of 2 if hardware processing is used for real-time processing, but need not be 32 in particular, and needless to say, it may not be a power of 2. Increasing the normalized number enables more accurate phase correction. However, if the normalized number is too large, the calculation scale becomes large and the result does not change so much. 32). When the pixel position is shifted to the left from the ideal pixel position, the filter coefficient is supplied so as to shift the pixel position to the right, and A = 0. Conversely, when the pixel position is shifted to the left, C = 0. Further, in this embodiment, the configuration of the phase correction filter is a filter of 3 pixels × 1 line, but it goes without saying that the filter size can be increased to provide a more accurate phase correction filter.

  The 103 phase correction filter coefficient generation unit receives the 104 pixel position information, reads the pixel shift amount corresponding to the pixel position information from the 105 pixel shift amount storage unit, and generates a phase correction filter coefficient.

  In this embodiment, it is assumed that the pixel shift amount for each pixel is stored in advance in the 105 pixel shift amount storage means. That is, since each of R, G, and B has a pixel shift amount for each pixel, in this embodiment, the number of pixels in the main scanning direction × 3 pixel shift values are stored. Further, the number of bits of the pixel shift amount to be stored may be determined regardless of the normalization number of the phase correction filter described above, but in this embodiment, 0 to 31 according to the normalization number. And If the pixel position is shifted to the left with respect to the ideal pixel position, a negative value is set as −31 to 0 and a signed 6-bit signal is stored. The stored pixel shift amount is 1/32 pixel unit. For example, if the 104 pixel position information is the 3000th pixel in the R main scan, the pixel shift amount stored in the 3000th main scan pixel is read out. Assuming that -19 is stored, 19/32 = about 0.6 pixel shift is present, and since it is negative, this indicates that there is a shift of about 0.6 pixels to the left. Actually, based on the measurement data, etc., it is detected that the 3000th pixel of R is shifted to the left by about 0.6 pixels, and the pixel shift amount is recalculated to the value to be stored in the 3000th pixel of R. This is stored in advance.

  The 103 phase correction filter coefficient generation means reads the pixel shift amount as described above, and generates a phase correction filter. Assume that a value of −19 is read out as in the above example. Since the shift is about 0.6 pixels to the left, the phase can be corrected by shifting the target pixel by 0.6 pixels to the right. In order to shift to the right, A = 0 in Equation 1. Since it is desired to shift to the right of 19/32 pixels, the coefficient of the phase correction filter based on the linear interpolation calculation is calculated as B = 32-19 = 13 C = 19.

  Conversely, when a value of 19 is read, it is shifted to the right by about 0.6 pixels, and C = 0 in equation 1, B = 32-19, and A = 19.

  The calculation of the 103 phase correction filter has a normalized number of 32 and is a phase correction filter based on linear interpolation calculation. If they are equal and minus, the absolute value of the pixel shift amount is equal to the coefficient C. B is obtained by subtracting the absolute value of the pixel shift amount from 32, and it can be seen that the coefficient calculation is simplified. Of course, the phase correction filter is not limited to this, and the pixel shift amount is not limited to this. It is also possible to configure so that the pixel shift amount does not coincide with the normalized number. What is important is that it has a mechanism for generating a coefficient in accordance with the pixel shift amount of the target pixel.

  In addition, the 105 pixel deviation amount storage means has been described in the present embodiment so as to memorize the pixel deviation amounts of all colors and all pixels, but the deviation of two other colors is stored on the basis of one certain color. It is possible to save a storage area for one color. In this case, the 103 phase correction filter can also reduce the calculation processing for one color.

Further, it is not always necessary to store the pixel shift amounts of all the pixels. The pixel shift amount of the representative point is stored, and the pixel shift amount of the target pixel position is calculated by the 103 phase correction filter coefficient generation unit. You may comprise so that it may estimate from the pixel deviation | shift amount of a nearby representative point. For example, with a pixel shift of R, the pixel shift amount of the 0th pixel is shifted by -19 = approximately 0.6 pixels to the left, and the pixel shift amount of the 3500th pixel is shifted by -8 = 0.25 pixels to the left. Thus, assuming that the pixel shift amount of the 7000th pixel is shifted to the right by 8 = 0.25 pixel, these three points (actually, 3 points × each color) are stored.
The pixel shift amount of the 3000th pixel is calculated as follows.

(-19 * (35003000) + (-8) * 3000) / 3500) = 10 (Formula 2)
Note that it is rounded to the nearest whole number.

  Similarly, the 0th to 3500th pixels are obtained by linear interpolation calculation of two pixel deviation amounts of 0 and 3500 pixels, and the 3500th to 7000th pixels are obtained by linear interpolation calculation of pixel deviation amounts of 3500 and 7000 pixels. It can be calculated. Because it is linear interpolation, the pixel shift amount other than the pixel that stores the pixel shift amount will cause an error from the actual pixel shift amount, but it is very difficult to measure all the pixel shift amounts of each pixel. Actual measurement is performed at intervals of several pixels at most, and by increasing the number of pixels of the pixel shift amount to be stored, the error can be reduced and implementation without any problem is possible.

  The 106 sharpness correction filter corrects the pixel shift with the following configuration based on the sharpness correction filter coefficient generated by the 107 sharpness correction filter coefficient generation unit.

IMG1 '' = (D * IMG0 '+ E * IMG1' + D * IMG2 ') / 128 (Formula 3)
IMG1 'is the pixel of interest. IMG0 'is the left pixel of the target pixel. IMG2 'is the right pixel of the pixel of interest. Since [′] is attached to each, it is understood that image data in which pixel shift is corrected by the phase correction filter in Expression 1 is input. D and E are coefficients of the sharpness correction filter, and are generated and supplied by 107 sharpness correction filter coefficient generation means described later. IMG1 ”is the image data after the sharpness correction filter. Also, the filter coefficient normalized so that 2 * D + E = 128 is set. However, if it is large, more accurate correction is possible, but if it is 128, there is no practical problem, and the same number D is assigned to the left and right coefficients of the target pixel. This is because a symmetrical coefficient is required when only the sharpness is corrected without shifting the phase.

  107 Sharpness correction filter coefficient generation means aims to generate a filter coefficient for sharpening an image blurred by the phase correction filter, and from 103 phase correction filter coefficient generation means, phase correction applied to the target pixel. A signal representing the characteristics of the filter is received. In the present embodiment, the signal representing the characteristic of the phase correction filter represents the characteristic with B in the coefficient equation 1 of the phase correction filter, and therefore the B signal is received. The frequency response of the phase correction filter has the same characteristics with a 0.5 pixel shift as a boundary. Therefore, if the coefficient B in Equation 1 of the phase correction filter is a, B has the same frequency response as when 32a. Therefore, it can be understood that a value of 0 to 16 should be obtained, and it may be carried out as such, but in this embodiment, a value of 0 to 31 is received.

  FIG. 7 shows a graph of the frequency response of the phase correction filter when B = 20 0.375 pixel shift. As can be seen from the Fourier transform, the same is true for B = 32-20 = 12. Here, focus on the power of 150 dpi. There is no problem if it can be made exactly the same, but there are cases where the targeted frequency response cannot be achieved with a limited filter size, and it is realistic to pay attention to a certain representative frequency and design it to guarantee the power of that frequency. In addition, there are few problems if it is controlled at a frequency targeted for practical use. From the graph, it can be seen that the power is 0.73, that is, the amplitude of the sin wave with a period of 150 dpi is attenuated by 0.73 times. The frequency response of the filter for correcting this may be 150 dpi and designed so as to be close to the reciprocal of 1.73 (referred to as a power target value for convenience) of 0.73.

  FIG. 8 shows the frequency response of the sharpness correction filter when D = F = -23 E = 174. FIG. 9 shows the frequency response when a filter as designed at 1.36 at 150 dpi has been designed, and both the phase correction filter and the sharpness correction filter are applied. It can be confirmed that the power has returned to 1 at 150 dpi, almost as intended (sub-scanning power is omitted in this figure).

  For example, the sharpness correction filter coefficient is selected as follows.

The frequency response of the phase correction filter when B = 4,8,12,16. The power of 150 dpi is 0.88, 0.79, 0.73, 0.71, and the target value of 150 dpi power of the sharpness correction filter, which is the reciprocal of it. Are 1.13, 1.26, 1.37, and 1.4, respectively. Create four sets of frequency response filter coefficients (sets of D and E in Equation 3) that are the target values of this power, and set them as COEF1 to COEF4.
The value of B is
When 0 to 2,30,31, D = 0, E = 128, no filtering
COEF1 for 3-6, 26-29
7-10, 22-25, COEF2
For 11-14, 18-21, COEF3
At 15-17, COEF4
It is possible to correct the image blurred by the phase correction filter so that the sharpness becomes the same. Although only the coefficient of COEF3 is shown above, it is a design matter and not important, so the other coefficients are omitted here.

(Second embodiment)
FIG. 3 is a block diagram that best represents the features of this embodiment.

  101 is an image input unit, 102 is a phase correction filter that corrects a pixel shift amount for each color, and 103 is a phase correction filter coefficient correction that receives a pixel shift amount for each color according to a pixel position and generates a filter coefficient. Means 104, a pixel position information data input unit, 105 stores a pixel shift amount for each color, a pixel shift amount storage unit 106, sharpness correction for correcting blurring by the phase correction filter A filter 107 is a sharpness correction filter coefficient generation unit that generates a sharpness correction filter coefficient according to an amount of blurring by the phase correction filter, and 301 is pixel position information input that is input to the sharpness correction filter coefficient generation unit. Means 302 is a sharpness storage means. The sharpness is generally stored for each color and each pixel based on the measured sharpness data. Although 302 is described in a different block from 105, it is needless to say that the storage medium may be the same or the storage location may be different.

  The difference from the first embodiment will be described in more detail in the operation of this embodiment. The 107 sharpness correction filter coefficient generation means not only receives the frequency response characteristics of the phase correction filter coefficients but also receives sharpness information for each pixel and each color from the 302 sharpness storage means, and the sharpness is predetermined. Calculate or select a sharpness correction filter so that the target value is obtained. By doing so, it is possible to realize an image with less pixel shift (color shift) and sharpness depending on the location, which absorbs the blurring by the phase correction filter and the difference between each color and each pixel.

  It is assumed that a signal indicating the sharpness of 150 dpi is stored in the 301 sharpness storage unit. Power is 0.4, 0.5, 0.6, 0.7, or 0.8, and five signals of 0, 1, 2, 3, and 4 are assigned to each color and each pixel. It is done.

  On the other hand, in the first embodiment, the 107 sharpness correction filter coefficient means receives the coefficient B of Equation 1 as a signal representing the characteristics of the frequency response of the phase correction filter coefficient. Therefore, it assigns to the signal of 0-4.

That is, the value of B is
0 to 2, 30, 31
1 for 3-6, 26-29
7-10, 22-25, 2
3 for 11-14, 18-21
For 15-17, it is 4.

  The sharpness target is 0.6. In the first embodiment, the value is set to 1. This is because only the characteristics of the phase correction filter are considered, and in this embodiment, the target is set in accordance with the actual measured sharpness. Is set.

  FIG. 10 shows a table of target values of the frequency response of the sharpness filter that becomes 0.6 at 150 dpi from the signal from the sharpness storage means and the characteristic signal of the phase correction filter. The sharpness filter coefficients for 25 are designed so as to achieve this target value, and 107 sharpness filter coefficient generation means selects this coefficient from the signal from the sharpness storage means and the characteristic signal of the phase correction filter. You just have to do it.

  As shown in FIG. 11, when the frequency response of the sharpness filter is quantized in increments of 0.25, the six filters from 0.75 to 2 can be substituted.

It is a block diagram of the image processing means used for description of a 1st Example. 1 is an external view of an image reading apparatus. It is a block diagram of the image processing means used for description of the 2nd Example. It is a figure for demonstrating the reading operation | movement of an image reading apparatus. It is a figure for demonstrating a color shift and a pixel shift. It is a figure for demonstrating operation | movement of a phase correction filter process. It is a graph for demonstrating the example of the frequency response of a phase correction filter. It is a graph for demonstrating the example of the frequency response of a sharpness filter. It is a figure for demonstrating the example of the frequency response at the time of implementing a phase correction filter + sharpness correction filter. It is a table | surface which shows the target value of the power of the sharpness correction filter used in the 2nd Example. It is a table | surface which can quantify the target value of the power of the sharpness correction filter used in the 2nd Example, and can reduce the number of filter coefficients.

Claims (5)

A color image reading device that reads a document image as a digital signal for each pixel;
A pixel shift amount storage means for storing a pixel shift amount for each color;
Reading out a pixel shift amount from the pixel shift amount storage means, and a phase correction filter coefficient generation means;
Phase correction filter means for receiving a coefficient from the phase correction filter coefficient generation means and performing phase correction for each pixel;
A sharpness correction filter coefficient generation means for receiving a phase correction filter characteristic signal informing the frequency characteristic of the filter of the phase correction filter coefficient from the phase correction filter coefficient generation means and generating a sharpness correction filter coefficient;
Image processing characterized by having image processing means comprising sharpness correction filter means for receiving sharpness correction filter coefficients for each pixel from the sharpness correction filter coefficient generation means and performing sharpness correction system. In claim 1,
A sharpness storage means for storing a sharpness signal in each pixel is provided,
The sharpness correction filter coefficient generation means generates a sharpness correction filter coefficient in accordance with the sharpness signal read from the sharpness storage means and the phase correction filter characteristic signal from the phase correction filter generation means. An image processing system.   3. The image processing system according to claim 1, wherein the pixel shift amount storage means stores pixel shifts of all pixels.   In Claims 1 and 2, the pixel shift amount storage means stores the pixel shift amounts of at least three pixels in the left and right ends and the center in the main scanning direction, and the pixels that do not store the pixel shift amount are: An image processing system which calculates based on a pixel shift amount of a stored pixel. In Claims 1 and 2, the pixel shift amount storage means stores the pixel shift amounts of at least three pixels in the left and right ends and the center in the main scanning direction, and the pixels that do not store the pixel shift amount are: Calculation is performed based on the pixel shift amount of the stored pixel. In this case, an image processing system is characterized by reading out pixel shift amounts of two pixels sandwiching a pixel of interest for which a pixel shift amount is to be calculated, and calculating the pixel shift amount by linear interpolation based on the number of pixels.

Images