JP2006331163A - Image processing device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To enhance quality in both cases where edged places are arranged continuously and distributed discretely. <P>SOLUTION: An edge feature calculation part 101 inputs input image information and outputs a value depending on the distribution of edge information. A noise generation part 102 generates and outputs noise. A noise correction part 103 computes a correction factor from the value calculated by the edge feature calculation part and corrects the noise with the correction factor. Output image information can be computed from the sum of the corrected noise and the input image information to enhance quality in both cases where edged places are arranged continuously and distributed discretely. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a display device such as a TV broadcast receiver or a projector.

  The image in which the high-frequency component is lost has a lowered texture at the portion where the pixel value changes.

  In order to improve the texture of this portion, it is necessary to change each pixel value, and noise is an example of an element that gives this change.

  As a conventional technique for increasing the texture using noise, the difference between pixel values around the target pixel is obtained from the image information before resolution conversion, the size of the noise is determined according to the difference, and the image information after resolution conversion is obtained. There was something to add (see, for example, Patent Document 1). FIG. 21 shows the prior art described in Patent Document 1.

In FIG. 21, a linear interpolation unit 2101 performs linear interpolation on input image information. The dynamic range calculator 2104 outputs a value for correcting the magnitude of the output value of the noise generator 102 from the input image information. The noise generator 102 generates and outputs a random number. The noise correction unit 103 corrects the size of the random number with the output value of the dynamic range calculation unit 2104. The output value of the noise correction unit 103 is added to the image information after linear interpolation to obtain output image information.
Japanese Laid-Open Patent Publication No. 7-44701 (page 3, FIG. 4)

  However, in the conventional configuration, since the value for correcting the magnitude of the noise is calculated from the difference between the pixel values around the target pixel, it is distributed discretely from the case where the locations with the edges are continuously arranged. I cannot distinguish between the cases. Therefore, the noise of the same magnitude is added in these two cases, and there is a problem of lowering the texture in either case.

  The present invention solves the above-described conventional problems, and an object of the present invention is to provide a method for improving the texture in each of these two cases.

  In order to solve the above-described conventional problems, the image processing apparatus according to the present invention calculates an edge feature amount that is a value based on a distribution of edge information from a target pixel and its surrounding pixels in pixels constituting input image information. Input a value calculated by the feature amount calculation unit, a noise generation unit that generates noise having a predetermined magnitude, a noise correction unit that corrects the magnitude of the noise by the edge feature amount, and the noise correction unit An addition unit for adding to image information is provided.

  In this specification, information about an edge such as the size, direction, and inclination of the edge is defined as edge information.

  With this configuration, noise is corrected and added to image information with a value that can distinguish between cases where edges are arranged side by side and cases where they are discretely distributed, so that changes that match the shape of the image are given. The effect of raising the texture can be achieved.

  According to the image processing apparatus of the present invention, the noise magnitude is corrected with a value based on the distribution of the edge information, so that the locations where the edges are located are discretely distributed. Can be distinguished from each other, so that a change according to the shape of the image can be given and the texture can be improved.

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

(Embodiment 1)
FIG. 1 is a block diagram of an image processing apparatus according to Embodiment 1 of the present invention.

  In FIG. 1, an edge feature quantity calculation unit 101 receives input image information and outputs an edge feature quantity that is a value based on the distribution of edge information. The noise generator 102 generates noise and outputs it. The noise correction unit 103 receives an edge feature amount and noise. Then, a correction coefficient is obtained from the edge feature amount, and noise is corrected. The corrected noise is added to the input image information to obtain output image information.

  The operation of the image processing apparatus configured as described above will be described below.

  First, the edge feature quantity calculation unit 101 will be described in the present embodiment.

  FIG. 2 is a block diagram for explaining the operation of the edge feature quantity calculation unit 101.

  In FIG. 2, the calculation area extraction unit 201 extracts pixel values used in the calculation process as one block including the target pixel and its surrounding pixels. For example, 65 × 65 pixels are extracted around the target pixel. This block is extracted for all the pixels constituting the image information.

  The edge direction classification unit 202 obtains the edge direction for each pixel group in the block, and classifies it into each class based on the direction. FIG. 5 is a flowchart for explaining the processing of the edge direction classification unit 202 for one block. First, a pixel group X having the same size as the filter centered on the calculated pixel as shown in FIG. 3 is extracted from the block extracted by the calculation area extraction unit (step S501). In this specification, a pixel at the center position of a pixel group that is currently targeted by processing in a block is defined as a calculated pixel. Filters (A to H) corresponding to the directions of the edges in FIG. 4 are sequentially applied to the pixel group X. Then, the maximum value is obtained from the calculated values of each filter and pixel group as shown in (Equation 1).

  The edge direction corresponding to the filter having the maximum value becomes the edge direction of the calculated pixel (step S502). The class value corresponding to the direction of FIG. 4 is set as the value of the calculated pixel (step S503). When the class value is obtained for all the pixels in the block, the processing of the edge direction classifying unit 202 is terminated (step S504). The result of finishing the process is as shown in FIG. 6, for example.

  The matrix creation unit 203 obtains a run length matrix for the class value obtained by the edge direction classification unit 202. Since the run-length matrix counts the number of the same class that is continuously arranged on a straight line, a run-length matrix is created for each angle of the straight line. Here, the horizontal direction is assumed to be 0 °, and run-length matrices are obtained for four angles of 45 °, 90 °, and 135 °, respectively. FIG. 7 shows an example when counting is performed for an angle of 0 °. The rows are continuous numbers, the columns are class values, and the values in the matrix are frequency numbers.

  The feature amount calculation unit 204 calculates a value paying attention to the number of consecutive locations where there is an edge. First, as shown in (Expression 2), F is calculated such that the larger the position where the edge is, the longer the position is.

Here, θ is an angle when counting consecutive numbers, i is a class value, j is a consecutive number, and P _θ (i, j) is a frequency. Using this function F, an edge feature amount that is a small value when the location of the edge is long and continuous and is a large value when the location is not continuous is obtained from the function of FIG. The edge feature value obtained by this function is used as the output value of the edge feature value calculation unit 101.

  The above is the description of the operation of the edge feature quantity calculation unit.

  In the present embodiment, the noise generation unit 102 will be described.

  The noise generator generates a uniform random number with an average of 0 and a variance of 1, and outputs the value as noise. Note that a Gaussian random number or an exponential random number may be used instead of a uniform random number.

  In the present embodiment, the operation of the noise correction unit 103 will be described.

  FIG. 9 is a block diagram for explaining the operation of the noise correction unit 103.

  The correction coefficient calculation unit 901 multiplies the edge feature value, which is an output of the edge feature value calculation unit 101, by an adjustment coefficient to obtain a correction coefficient for correcting the noise magnitude. This adjustment coefficient is a predetermined value.

  The correction unit 902 multiplies the noise that is the output of the noise generation unit 102 and the correction coefficient, and uses the value as the output of the noise correction unit 103.

  The output value of the noise correction unit 103 is added to the input image information to obtain output image information.

  According to such a configuration, by correcting the magnitude of noise with a value based on the distribution of edge information, there are cases where the locations with the edges are arranged side by side and the locations with the edges are distributed discretely. Since they can be distinguished, it is possible to give a change according to the shape of the image and improve the texture.

(Embodiment 2)
FIG. 10 is a block diagram of the image processing apparatus according to the second embodiment of the present invention. 10, the same components as those in FIG. 1 are denoted by the same reference numerals, and description thereof is omitted.

  In FIG. 10, a high frequency feature quantity calculation unit 1003 receives input image information and outputs a high frequency feature quantity that is a value based on the distribution of high frequency components of each pixel group. The noise correction unit 1004 receives an edge feature value, a high frequency feature value, and noise as inputs. Then, a correction coefficient is obtained from the edge feature quantity and the high frequency feature quantity, and noise is corrected. The corrected noise is set as the output of the noise correction unit 1004.

  Detailed description of the same parts as in the present embodiment and the first embodiment will be omitted. The difference between the present embodiment and the first embodiment is a high-frequency feature amount calculation unit 1003 and a noise correction unit 1004.

  In the present embodiment, the high frequency feature amount calculation unit 1003 will be described.

  FIG. 11 is a block diagram for explaining the operation of the high-frequency feature amount calculation unit 1003.

  In FIG. 11, the calculation area extraction unit 201 is the same as the calculation area extraction unit 201 of FIG. The threshold value determination unit 1102 obtains the magnitude of the high-frequency component in each pixel group of the block selected by the calculation region extraction unit 201, and classifies it into two classes by threshold value determination processing. The feature amount calculation unit 1103 calculates a value focusing on one of the classes classified by the threshold determination unit 1102.

  The threshold determination unit 1102 multiplies the pixel group X having the same size as the filter centered on the calculated pixel as shown in FIG. 3 by the 3 × 3 high-pass filter HPF shown in FIG. The threshold value of (Equation 3) is determined for the output value, and classified into 0 and 1 classes.

  Here, T is a predetermined threshold value. Y is obtained for all the pixel groups in the block, for example, as shown in FIG.

  The feature amount calculation unit 1103 calculates a value focusing on the class determined as Y = 1 by the threshold determination unit 1102 as shown in (Expression 4).

  Here, i is a block row, j is a block column, and C (i, j) is a class value. The high frequency feature value calculated in (Equation 4) is set as the output value of the high frequency feature value calculation unit 1003. Therefore, the higher the high frequency feature amount, the larger the pixel group having a high frequency component equal to or higher than a predetermined threshold value.

  The above is the description of the operation of the high-frequency feature amount calculation unit 1003.

  In this embodiment, the operation of the noise correction unit 1004 will be described.

  FIG. 14 is a block diagram for explaining the operation of the noise correction unit 1004.

  The correction coefficient calculation unit 1401 multiplies the value obtained by multiplying the edge feature value output from the edge feature value calculation unit 101 by the high frequency feature value output from the high frequency feature value calculation unit 1003 by an adjustment coefficient. This value is a correction coefficient that determines the magnitude of noise.

  The correction unit 1402 multiplies the noise output from the noise generation unit 102 and the correction coefficient, and uses the corrected noise as the output of the noise correction unit.

  According to such a configuration, a high-frequency feature quantity calculation unit is further provided, and by correcting the noise magnitude with the high-frequency feature quantity, it is possible to distinguish between cases where the number of locations with edges is large and cases where the number of places is small. Can change the texture.

  Further, the texture can be further improved by correcting the magnitude of noise with two different values that can be distinguished.

(Embodiment 3)
FIG. 15 is a block diagram of an image processing apparatus according to the third embodiment of the present invention. In FIG. 15, the same components as those in FIG.

  In FIG. 15, a total feature amount calculation unit 1504 receives input image information and outputs a total feature amount which is a value based on the sum of frequency components. The variance feature quantity calculation unit 1505 receives input image information and outputs a variance feature quantity that is a value based on the variance of pixel values. A hue feature quantity calculation unit 1506 receives input image information and outputs a hue feature quantity that is a value based on the hue of the pixel value. A noise correction unit 1507 receives an edge feature value, a total feature value, a dispersion feature value, a hue feature value, and noise. Then, a correction coefficient is obtained from the edge feature value, the total feature value, the dispersion feature value, and the hue feature value, and noise is corrected. The corrected noise is set as the output of the noise correction unit 1507.

  Detailed description of the same parts as in the present embodiment and the first embodiment will be omitted. The difference between the present embodiment and the first embodiment is a total feature amount calculation unit 1504, a dispersion feature amount calculation unit 1505, a hue feature amount calculation unit 1506, and a noise correction unit 1507.

  In the present embodiment, the total feature amount calculation unit 1504 will be described.

  FIG. 16 is a block diagram for explaining the operation of the total feature amount calculation unit 1504.

  In FIG. 16, the calculation region extraction unit 201 is the same as the calculation region extraction unit 201 of FIG. The FFT unit 1602 performs FFT on the block selected by the calculation area extraction unit 201. A feature amount calculation unit 1503 calculates the sum of amplitude components from the output value of the FFT unit 1602.

  The FFT unit 1602 performs an FFT on the block selected by the calculation region extraction unit 201 to obtain an amplitude component for each frequency. Note that DFT or DCT may be performed instead of FFT.

  The feature amount calculation unit 1603 calculates the sum of the amplitude components at a frequency equal to or higher than a predetermined threshold as shown in (Expression 5).

  Here, i is a real axis, j is an imaginary axis, and FFT (i, j) is an amplitude. T1 is a real axis threshold, and T2 is an imaginary axis threshold. The total feature amount calculated in (Equation 5) is used as the output value of the total feature amount calculation unit 1504. Therefore, the total feature amount becomes larger as the total value is larger.

  In the present embodiment, the distributed feature amount calculation unit 1505 will be described.

  FIG. 17 is a block diagram for explaining the operation of the distributed feature quantity calculation unit 1505.

  In FIG. 17, the calculation area extraction unit 201 is the same as the calculation area extraction unit 201 of FIG. The luminance conversion unit 1702 obtains a luminance component for each pixel value in the block of the calculation area extraction unit 201. The feature amount calculation unit 1703 calculates the variance of the luminance component from the output of the luminance conversion unit 1702.

  The luminance conversion unit 1702 performs YCrCb conversion and calculates a luminance component for each pixel value in the block of the calculation region extraction unit 201.

  The feature amount calculation unit 1703 obtains the block whose luminance has been converted by the luminance conversion unit 1702 as shown in (Expression 6).

Here, i is the row of the block, j is the column of the block, N is the maximum value of the row, M is the maximum value of the column, X is the luminance component, and X _ave is the average of the luminance component. The variance feature value calculated in (Equation 6) is used as the output value of the variance feature value calculation unit 1505. Therefore, the variance feature amount increases as the variance value increases.

  In the present embodiment, a hue feature amount calculation unit 1506 will be described.

  FIG. 18 is a block diagram for explaining the operation of the hue feature quantity calculation unit 1506.

  In FIG. 18, the color conversion unit 1801 separates the hue from the pixel value of the target pixel. A feature amount calculation unit 1702 obtains a hue feature amount from the hue that is an output value of the color conversion unit 1801.

  A color conversion unit 1801 performs HSV conversion to obtain a hue angle. In place of HSV conversion, YCrCb conversion or L * a * b * conversion may be performed.

  The feature amount calculation unit 1802 obtains a hue feature amount from a function that has a larger value as the hue includes more green components and a smaller value as the more red components are included. The hue feature amount calculated by the function is set as an output value of the hue feature amount calculation unit 1506.

  In the present embodiment, the noise correction unit 1507 will be described.

  FIG. 20 is a block diagram for explaining the operation of the noise correction unit 1507.

  The correction coefficient calculation unit 2001 includes an edge feature value output from the edge feature value calculation unit 101, a total feature value output from the total feature value calculation unit 1504, and a dispersion feature value and hue output from the dispersion feature value calculation unit 1505. The hue feature amount that is the output of the feature amount calculation unit 1506 is multiplied and the value is multiplied by the adjustment coefficient. This value is a correction coefficient that determines the magnitude of noise.

  The correction unit 2002 multiplies the output value of the noise generation unit 102 and the correction coefficient, and uses the value as the output of the noise correction unit 1507.

  According to such a configuration, the sum total feature amount calculation unit is further provided, and by correcting the noise magnitude with a value based on the sum, it is possible to distinguish between the case where the number of places with edges is large and the case where the number is small. Can give a suitable change and raise the texture.

  In addition, a dispersion feature amount calculation unit is further provided, and by correcting the noise magnitude with a value based on dispersion, it is possible to distinguish between a case where the change in the pixel value is large and a case where the change is small. Can be raised.

  A hue feature amount calculation unit is further provided, and by correcting the magnitude of noise even in the hue feature amount, it is possible to narrow down the additional objects by color, and to change the image according to the shape of the image and improve the texture.

  The texture can be further improved by correcting the magnitude of the noise with a plurality of different values that can be distinguished.

  The image processing apparatus according to the present invention is configured to correct the magnitude of noise with a value based on the distribution of edge information and add it to the image information, so that the texture of the image can be improved, and a TV broadcast receiver or projector It can be used for display devices such as.

1 is a block diagram of an image processing apparatus according to Embodiment 1 of the present invention. Block diagram of edge feature quantity calculator Diagram showing processing for each pixel in the block Diagram showing the relationship between edge detection filter, edge direction and class value The flowchart which shows the process of the edge direction classification | category part with respect to 1 block The figure which shows an example of the processing result of an edge direction detection part Diagram showing run-length matrix The figure which shows the function which calculates the edge feature quantity FIG. 3 is a block diagram showing a configuration of a noise correction unit in the first embodiment. Block diagram of an image processing apparatus in Embodiment 2 of the present invention Block diagram of the high-frequency feature quantity calculator Diagram showing an example of a high-pass filter The figure which shows an example of the process result of a threshold value determination part Block diagram showing a configuration of a noise correction unit according to the second embodiment Block diagram of an image processing apparatus in Embodiment 3 of the present invention Block diagram of total feature amount calculation unit Block diagram of distributed feature quantity calculation unit Block diagram of the hue feature quantity calculator The figure which shows the function which calculates the hue feature quantity Block diagram showing a configuration of a noise correction unit according to the third embodiment Prior art block diagram that improves the texture by adding noise

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 Edge feature-value calculation part 102 Noise generation part 103 Noise correction part 1003 High frequency feature-value calculation part 1504 Sum total feature-value calculation part 1505 Dispersion feature-value calculation part 1506 Hue feature-value calculation part

Claims (12)

An edge feature amount calculation unit that calculates an edge feature amount that is a value based on a distribution of edge information from the target pixel and its surrounding pixels in the pixels constituting the input image information;
A noise generating unit for generating noise having a predetermined magnitude;
A noise correction unit that corrects the magnitude of the noise by the edge feature amount;
An addition unit for adding the value calculated by the noise correction unit to the input image information;
An image processing apparatus. The image processing apparatus according to claim 1, wherein the distribution of the edge information is a distribution in an edge direction. 3. The image processing according to claim 2, wherein the edge feature amount is a value that is corrected so that the magnitude of the noise decreases as the locations having the same edge direction around the pixel of interest gather longer. apparatus. The pixel constituting the input image information further includes a high-frequency feature amount calculation unit that calculates a high-frequency feature amount that is a value based on a high-frequency component for each local area from the target pixel and its surrounding pixels,
The image processing apparatus according to claim 1, wherein the noise correction unit corrects the magnitude of the noise based on the high-frequency feature amount. The high-frequency feature amount is a value that is corrected so that the larger the local area having a high-frequency component equal to or higher than a predetermined threshold around the target pixel, the larger the noise level. Image processing device. A pixel that constitutes the input image information further includes a total feature amount calculation unit that calculates a total feature amount that is a value based on the sum of frequency components from the target pixel and its surrounding pixels,
The image processing apparatus according to claim 1, wherein the noise correction unit corrects the magnitude of the noise based on the total feature amount. The image processing apparatus according to claim 6, wherein the total feature amount is a value that is corrected so that the magnitude of the noise increases as the value of the total sum increases. The pixel constituting the input image information further includes a dispersion feature amount calculation unit that calculates a dispersion feature amount that is a value based on the dispersion of the pixel value from the target pixel and its surrounding pixels,
The image processing apparatus according to claim 1, wherein the noise correction unit corrects the magnitude of the noise based on the variance feature amount. The image processing apparatus according to claim 8, wherein the variance feature value is a value that is corrected so that the magnitude of the noise increases as the variance value increases. A hue feature amount calculation unit that calculates a hue feature amount that is a value based on the hue from the pixels constituting the input image information;
The image processing apparatus according to claim 1, wherein the noise correction unit corrects the magnitude of the noise based on the hue feature amount. The image processing apparatus according to claim 10, wherein the hue feature amount is a value that is corrected so that the noise amount increases as the hue includes more green components. The image processing apparatus according to claim 10, wherein the hue feature amount is a value that is corrected so that the noise amount decreases as the hue includes more red components.

Images