JP2011022762A - Image signal processing device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image with texture, without generating block noise or unnecessary stripe when adding random noise to an image signal. <P>SOLUTION: An image signal processing device includes an orthogonal transformation unit 6 which performs an orthogonal transformation process to predetermined surrounding pixels of each pixel of an image signal; an additional noise level determination unit 7, which determines additional noise level, based on the amplitude of spatial frequency component resulting from the orthogonal transformation unit 6; and random noise addition units 8, 9, 10 which add random noise to the image signal, in accordance with the determined additional noise level by the additional noise level determining unit 7. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to an image signal processing apparatus that gives a texture to an image.

  If the brightness of the object surface in the image is flat due to human visual characteristics, the object is perceived as an image having no texture and lacking reality. In view of this, a technique has been proposed in which a noise component is added to the object surface to give a texture to make the image more realistic. For example, Patent Document 1 describes an image processing apparatus that gives a texture to an image having gradation by adding random noise for giving fluctuation to the image. In this method, an image is divided into block areas, and random noise (fluctuation) is added to each frequency component calculated by orthogonal transform in units of blocks, and random noise is added to the image by inverse orthogonal transform. It is the composition to do.

Japanese Unexamined Patent Publication No. 7-57077

  The technique described in Patent Document 1 adds random noise (fluctuation) to frequency components in units of each block area, and stirs the fluctuations in the entire image in the block by inverse orthogonal transformation. In this case, since the processing is performed in units of blocks, there is a problem that a so-called block noise is generated due to a difference in noise level between adjacent blocks. In addition, since fluctuation is added to the frequency component, there is a problem that the added region of fluctuation in the image is localized, and stripes and spots are likely to occur.

  An object of the present invention is to provide an image signal processing apparatus that gives a texture to an image without generating block noise or unnecessary stripes when adding random noise to the image signal.

  The image signal processing apparatus according to the present invention includes an orthogonal transform unit that performs orthogonal transform processing on surrounding pixels of a predetermined region surrounding each pixel of the image signal, and an amplitude of a spatial frequency component converted by the orthogonal transform unit. A noise addition level determination unit that determines a noise addition level based on the noise addition level; and a random noise addition unit that adds random noise to the image signal according to the noise addition level determined by the noise addition level determination unit.

  Here, the noise addition level determination unit calculates the sum of each amplitude by multiplying the amplitude of each spatial frequency component converted by the orthogonal transform unit by a weighting factor, and determines the noise addition level based on the calculated value. The noise addition level determining unit determines a noise addition level by performing level correction in a substantially S-curve shape on the amplitude of the spatial frequency component converted by the orthogonal transform unit.

  According to the present invention, it is possible to give a texture to an image without generating block noise or unnecessary stripes.

1 is a block configuration diagram of an image signal processing apparatus according to a first embodiment. The figure which shows the two-dimensional area | region made into the process target of orthogonal transformation. The figure which shows the spatial frequency component obtained by orthogonal transformation processing. The figure which shows the example of the weighting coefficient by which a spatial frequency component is multiplied. The figure which shows the example of the weighting coefficient by which a spatial frequency component is multiplied. The figure which shows the example of level correction of a noise addition level. The block diagram of the image signal processing apparatus which concerns on a 2nd Example. The block block diagram of the image signal processing apparatus which concerns on a 3rd Example. The block block diagram of the image signal processing apparatus which concerns on a 4th Example.

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

  FIG. 1 is a block diagram of an image signal processing apparatus according to the first embodiment of the present invention. In FIG. 1, 1 is an input terminal, 2 to 5 are line memories, 6 is an orthogonal transformation unit, 7 is a noise addition level determination unit, 8 is a random noise generation unit, 9 is a multiplier, 10 is an adder, and 11 is an output. Terminal. In this embodiment, noise is added to the input image signal for each pixel. In order to determine the amount of noise to be added, the result of orthogonal transformation processing for surrounding pixels in a predetermined area surrounding the processing target pixel is used. To do. Here, the orthogonal transform unit 6 performs an orthogonal transform process on a rectangular area of 5 pixels in the horizontal direction and 5 lines in the vertical direction.

  The image signal 101 input from the input terminal 1 is input to the orthogonal transform unit 6 and also input to the line memory 2 and delayed by one line. The signal 102 delayed by one line is input to the orthogonal transform unit 6 and also input to the line memory 3 and further delayed by one line. Similarly, line memories 4 and 5 are connected to the orthogonal transformation unit 6, so that the image signal 101 of the current line, the image signal 102 delayed by one line, the image signal 103 delayed by two lines, and the three line delay The signal 104 and the signal 105 delayed by four lines, that is, the image signal for five lines in the vertical direction of the image are input.

  The orthogonal transform unit 6 extracts five pixels in the horizontal direction from the input image signals 101 to 105 for five lines. The 25 pixels are subjected to orthogonal transformation processing such as discrete cosine transformation (DCT). The amplitude 106 of each frequency component is obtained by orthogonal transform processing.

  The noise addition level determination unit 7 multiplies the amplitude 106 of each frequency component thus obtained by a weighting factor to select a frequency range to which noise is added, and takes the sum of the amplitudes of the selected range as an evaluation value. Based on this evaluation value, the noise addition level 107 is determined. This noise addition level 107 is given to the multiplier 9.

  The random noise generator 8 generates random noise 108. The multiplier 9 multiplies the random noise 108 by the noise addition level 107 and outputs a noise signal 109. The adder 10 adds the noise signal 109 from the multiplier 9 to the image signal 103 a from the line memory 3 that is the processing target pixel, generates an image signal 110, and outputs the image signal 110 from the output terminal 11.

Next, the operation of each unit will be described in detail.
FIG. 2 is a diagram illustrating a two-dimensional region to be processed by the orthogonal transform unit 6. By performing serial / parallel conversion on the input image signals 101 to 105 for five lines, five pixels are extracted in the horizontal direction. Thus, a two-dimensional area 200 including 5 × 5 = 25 pixels of 5 pixels in the horizontal direction and 5 lines in the vertical direction is set. Among these, the pixel 201 at the center of the region is a processing target pixel (a pixel to which noise is added), and is the central pixel of the pixel column output from the line memory 3.
Here, a 5 × 5 spatial region is targeted as the size of the two-dimensional region 200, but the size is not limited to this, and the spatial region may be set as appropriate, such as 3 × 3 or 7 × 7. .

  FIG. 3 is a diagram illustrating the spatial frequency components obtained by the orthogonal transformation process of the orthogonal transformation unit 6. The pixel value distribution of 25 pixels in FIG. 2 is converted into the values of the spatial frequency components 300 in the horizontal direction and the vertical direction. K00 to K44 represent the amplitude of each frequency component, the amplitude K00 in the upper left corner indicated by reference numeral 301 is a DC component, and the other AC components are taken up to the fourth order in the horizontal and vertical directions, respectively. The extent to which high-frequency components are taken into consideration may be appropriately set according to the spatial frequency range (picture) of an image to which noise (texture) is to be added.

The noise addition level determination unit 7 multiplies the amplitude data Kij of the spatial frequency component of the image signal by the weight coefficient Xij to select a frequency range to which noise is added. Further, the sum of the amplitudes in the selected frequency range is calculated and used as the evaluation value Z of the processing target pixel.
Z = ΣΣKij ・ Xij

4A and 4B are diagrams showing examples of the weighting factor Xij. The weighting factor Xij is described by a value 400 of a 5 × 5 matrix so as to correspond to each frequency component 300 in FIG. For example, the weighting factor “0” indicated by reference numeral 401 in the upper left of FIG. 4A is for the amplitude K00 indicated by reference numeral 301 in FIG.
The example of FIG. 4A is a case where the weighting coefficient in the intermediate frequency region indicated by the broken line 402 is increased. The intermediate frequency region corresponds to, for example, an edge portion of an image pattern, and by selecting this, noise can be preferentially added to the edge portion of the image, and the texture of the edge portion can be improved.

  On the other hand, the example of FIG. 4B is a case where the weighting coefficients in the direct current and low frequency regions indicated by the broken line 403 are increased. The direct current and low frequency regions correspond to the flat portion of the image pattern, and by selecting this, noise is preferentially added to the flat portion of the image. For example, in the case of a flat image such as the sky, a conventional gradation image in which a gradation step is conspicuous can be made inconspicuous by random noise, and the texture can be improved.

Further, the noise addition level determination unit 7 determines the noise addition level based on the evaluation value Z, and performs the next level correction at that time.
FIG. 5 is a diagram illustrating an example of level correction of a noise addition level. The horizontal axis represents the input evaluation value Z, and the vertical axis represents the noise addition level N to be output. Here, the S-shaped curve is approximated by a broken line of three regions. In the region 501 where the evaluation value Z is smaller than the threshold value Z1, the slope of the curve is reduced to suppress noise addition. In a region 502 between the threshold value Z1 and the threshold value Z2, the inclination is increased to increase noise addition. In the region 503 larger than the threshold value Z2, the inclination is reduced and the limit level NL is used.

The limit level NL of the added noise is desirably suppressed to about ± 10 gradations at the maximum when the pixel level is 256 gradations, for example. By performing such correction, the texture can be given more naturally.
It should be noted that the above-described approximation of the broken line of the S-shaped curve is an example, and it is needless to say that it may be divided into two regions (one threshold value) or four regions (three threshold values).
In this embodiment, the noise addition level determination unit 7 performs both selection of the frequency range to which noise is added and level correction of the noise addition level, but either one may be used.

According to the present embodiment, orthogonal transformation is sequentially performed for each pixel, and an addition level of random noise is determined for each pixel according to the result, so that block noise does not occur. Further, since random noise is directly added to the image signal instead of the frequency component, the noise addition position in the image is dispersed, and the occurrence of stripes and spots can be prevented.
Furthermore, there is no need for inverse orthogonal transform processing, and the circuit scale of the image signal processing device can be reduced.

  FIG. 6 is a block diagram of an image signal processing apparatus according to the second embodiment of the present invention. In FIG. 6, 12 is a user input terminal, and 13 is a control unit. The other parts are the same as those in the first embodiment (FIG. 1), and the same reference numerals as those in FIG.

  The user input terminal 12 is a terminal for inputting an instruction signal 112 from the user, and is connected to an operation panel, a controller, etc. (not shown). The control unit 13 sends and controls a control signal 113 to the noise addition level determination unit 7 in accordance with a user instruction from the input terminal 12. Here, the control signal 113 includes the value of the weighting factor by which the amplitude of each frequency component described in FIG. 4 is multiplied, and the value of the S-curve shown in FIG. 5 (threshold values Z1, Z2, curve slope, limit level). NL value). Thereby, the frequency range which gives random noise, and the level correction of random noise can be set. At this time, if the type of display device (CRT, plasma display, liquid crystal display) connected to the output terminal 11 is selected by the instruction signal 112, a random suitable for the characteristics of the connected display device is selected. Noise can be added.

  FIG. 7 is a block diagram of an image signal processing apparatus according to the third embodiment of the present invention. In FIG. 7, reference numeral 14 denotes a luminance detection unit, which is connected to the output 103 of the line memory 3. The other parts are the same as those in the first embodiment (FIG. 1), and the same reference numerals as those in FIG.

  The luminance detection unit 14 detects the luminance level Y of each pixel of the input image signal 103 a and sends information 114 of the detected luminance level Y to the noise addition level determination unit 7. When the luminance level Y is lower than the predetermined value Y1 or higher than the predetermined value Y2, the noise addition level determination unit 7 determines the value of the S-curve shown in FIG. 5 (curve slope, limit level NL). )) To reduce the added noise, or correct the level so that no noise is added. Thereby, it is possible to avoid image quality deterioration due to excessive noise in a dark image portion and a bright image portion where noise is conspicuous.

  FIG. 8 is a block diagram of an image signal processing apparatus according to the fourth embodiment of the present invention. In FIG. 8, 15 is a frame memory, 16 is a motion detection unit, and is connected to the output 103 of the line memory 3. The other parts are the same as those in the first embodiment (FIG. 1), and the same reference numerals as those in FIG.

  The image signal 103 from the line memory 3 is written in the frame memory 15 in units of frames, and is simultaneously supplied to the motion detection unit 16. The motion detection unit 16 compares the image signal 103 of the current frame output from the line memory 3 with the image signal 115 of the past frame read from the frame memory 15, and detects the amount of motion M of the image. Information 116 of the detected motion amount M is sent to the noise addition level determination unit 7. In accordance with the information 116 from the motion detection unit 114, the noise addition level determination unit 7 determines that the value of the S-curve shown in FIG. The level is corrected so as to reduce the additional noise by reducing the limit level NL). As a result, image quality deterioration due to excessive noise can be avoided in an image portion where noise is conspicuous and there is no movement.

While several embodiments according to the present invention have been described above, each embodiment can be used alone or in combination. According to each embodiment, orthogonal transformation is sequentially performed for each pixel, and the addition level of random noise is determined for each pixel according to the result, so that block noise does not occur. Further, since random noise is directly added to the image signal instead of the frequency component, the noise addition position in the image is dispersed, and the occurrence of stripes and spots can be prevented. Therefore, a texture can be given without deteriorating the original image quality.
The image signal processing apparatus of the present invention can be widely applied to video cameras, image display apparatuses, image printing apparatuses, and the like.

1 ... Input terminal,
2-5 ... line memory,
6 ... Orthogonal transformation unit,
7: Noise addition level determination unit,
8 ... Random noise generator,
9 ... multiplier,
10 ... adder,
11: Output terminal,
12 ... User input terminal,
13 ... control unit,
14: Luminance detection unit,
15 ... Frame memory,
16: Motion detection unit.

Claims (7)

In an image signal processing apparatus for adding random noise to an image signal,
An orthogonal transform unit that performs an orthogonal transform process on surrounding pixels in a predetermined area surrounding each pixel of the image signal;
A noise addition level determination unit that determines a noise addition level based on the amplitude of the spatial frequency component transformed by the orthogonal transformation unit;
An image signal processing apparatus comprising: a random noise adding unit that adds random noise to the image signal according to a noise addition level determined by the noise addition level determination unit. The image signal processing apparatus according to claim 1,
The noise addition level determination unit calculates a sum of each amplitude by multiplying the amplitude of each spatial frequency component converted by the orthogonal transform unit by a weighting factor, and determines the noise addition level based on the calculated value. An image signal processing apparatus. The image signal processing apparatus according to claim 2,
The image signal processing apparatus, wherein the noise addition level determination unit selects a spatial frequency range to which the random noise is added, based on the weighting factor. The image signal processing apparatus according to claim 1,
The image processing apparatus according to claim 1, wherein the noise addition level determination unit determines the noise addition level by performing level correction in a substantially S-curve shape on the amplitude of the spatial frequency component converted by the orthogonal transform unit. The image signal processing apparatus according to claim 2 or 4,
An image signal processing apparatus comprising: a control unit configured to set the weighting coefficient or the level correction in response to a user instruction to the noise addition level determination unit. The image signal processing apparatus according to claim 4,
A luminance detection unit for detecting a luminance level of the image signal;
The image signal processing apparatus, wherein the noise addition level determination unit sets the level correction according to a detected luminance level. The image signal processing apparatus according to claim 4,
A frame memory for writing image signals in units of frames;
A motion detection unit that detects a motion amount of an image of a current frame in comparison with an image signal of a past frame read from the frame memory;
The image signal processing apparatus, wherein the noise addition level determination unit sets the level correction according to the detected amount of motion.

Images