JP2003153262A - Method and apparatus for reducing mosquito distortion - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method for reducing mosquito distortion included in a video signal that can be realized with a simple configuration at a low cost independently of a decoder. SOLUTION: This invention provides the method for reducing mosquito distortion characterized in that a filter coefficient for the video signal is varied by using a lookup table on the basis of a maximum value of a contour within a prescribed range at a noted point of the video signal and flatness around the noted point.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for improving the image quality generated when a lossy-encoded image is decoded, and more particularly to a method for reducing the mosquito distortion and improving the image quality, and an apparatus using the above method. It is about.

[0002]

2. Description of the Related Art When an image is transmitted or recorded as a digital signal, a compression encoding method such as MPEG (Mo
Ving Picture Coding Exper
MP advocated by the Group and internationally standardized
EG-1, MPEG-2, MPEG-4 and the like are widely known.

These perform high-efficiency encoding of moving image information. In this encoding, an image is divided into square pixel blocks of a predetermined size, and each is orthogonally transformed. The obtained orthogonal transform coefficient is then quantized and entropy coded. When decoding, image information can be obtained by performing conversion in the reverse order of the encoding.

DCT (Discret) is used for the orthogonal transformation.
e Cosine Transform) is used, and the size of the pixel block in the conversion unit is 8 pixels × 8 pixels in the case of a luminance signal.

Therefore, the orthogonal transform coefficients obtained after the orthogonal transform are also 8 × 8 = 64, and these 64 coefficients are quantized. Quantization of these coefficients increases the compression rate by roughening the coefficients corresponding to spatial frequencies that are normally higher than human visual characteristics and image characteristics.

By the way, it is known that an image obtained by decoding the moving image information encoded in this way has a particular quantization distortion due to a quantization error.

Of such quantization distortion, the quantization error of the low-frequency component causes discontinuity such that an offset is applied in the luminance direction on a block-by-block basis in which orthogonal transformation is performed, and this is so-called block distortion.

Further, the quantization error of the high frequency component causes so-called mosquito distortion which appears as ringing near a steep edge. This ringing occurs when an edge having a strong high-frequency component exists in a block, and a harmonic component necessary for converging the step waveform is rounded by a quantization error, and as a result, it does not converge in the block. Therefore, ringing occurs only in a block where an edge exists.

It is known that mosquito distortion is particularly noticeable when there are steep edges in a flat image such as subtitles.

Various techniques have been proposed for the technique for reducing mosquito distortion as described above. For example, the deringing filter is introduced in the section of Annex F Post Filter of the Recommendation of MPEG-4 as a technique to be performed at the time of decoding. ing. However, many of the proposed methods including the above method use the decoding information obtained at the time of decoding the block boundary position, the quantization level, the motion vector and the like.

According to these methods, in the case of a device such as an image display device that does not have a decoding device, or even if the device has a decoding device but is divided into circuit blocks, decoding information cannot be obtained, so that mosquito distortion occurs. Can not be reduced. An example of a method for reducing mosquito distortion that does not use decoding information is disclosed in Japanese Patent Laid-Open No. 9-51532.

The above conventional example will be described with reference to FIG. FIG. 8 is a block diagram of a conventional mosquito distortion reduction device. In FIG. 8, 1 is an input terminal into which a video signal is input, and 52 is an input image signal with vertical m being n × horizontal n pixels (m> 1, n> 1)
It is a filter size image extraction circuit that sequentially cuts out for each filter pixel size of an odd value of.

Reference numeral 53 denotes a central pixel of the filter pixel size cut out by the filter pixel 52 or its vicinity includes prime elements,
It is a large area pixel data extraction means for extracting pixel data of a plurality of pixels (the upper limit number of pixels is all pixels of one image frame or one image field) larger than the number of pixels of the filter pixel size.

Reference numeral 54 includes a central pixel of the filter pixel size cut out by the filter size image extraction circuit 92, and is smaller than the number of pixels extracted by the large area pixel data extraction circuit 93. It is a small area pixel data extraction circuit for extraction.

A timing clock generator 55 supplies a clock and a timing signal to other circuits. 56,
Reference numerals 57 and 58 are memories for holding respective input image signals. Reference numerals 59, 60 and 63 are delay circuits for adjusting the delay amounts so that the respective input signals are synchronized with other circuits. Reference numeral 61 denotes a pixel information extraction circuit that extracts, as pixel information, statistics about large area pixel data, such as summation, average, deviation, correlation, or statistics about differences and secondary differences.

Reference numeral 62 denotes a pixel information extraction circuit which similarly extracts a statistic amount from the small area pixel data. Reference numeral 64 denotes a threshold value calculation circuit for determining a smoothing threshold value from the large area pixel information and the small area pixel information, which is usually a DSP for performing complicated calculation.
Etc. are used.

Further, the threshold value is varied as a continuous value to prevent the removal effect from unnaturally varying. 65
Is an ε-filter for smoothing an input video signal using the smoothing threshold as a threshold.

The ε-filter is a filter that performs the following calculation for the pixel p (x, y) at the spatial coordinate (x, y) point.

[0019]

[Equation 1]

M represents the number of pixels used for smoothing, and S (k,
l) is a scale factor. As shown in the above formula, the pixel data p (x, y) at a certain pixel position is smoothed by using the pixel data p (x + k, y + 1) in its vicinity to reduce distortion, but the central pixel data to be smoothed. If the absolute value of the difference between the pixel data and the neighboring pixel data is larger than the smoothing threshold value, the neighboring pixel data is not used for smoothing.

As described above, in the conventional technique, the smoothing threshold value of the pixel to be smoothed is calculated from the pixel information of the large area and the pixel information of the small area, and the base of the size of the distortion removal is based on the large image area. At the same time as the setting, the local distortion removal correction can be performed by the information of the small image area.

[0022]

A complicated operation is required to calculate the smoothing threshold value as described above. For example, conventionally, there is a problem that a circuit scale becomes large because it is necessary to use a DSP or the like and a large number of multiplication circuits and memories are required, and it is used in a display device or the like which is required to have low cost even when actually applied to a product. Was difficult.

The present invention has been made in view of the above-mentioned conventional problems, and it is an inexpensive structure that does not use the decoding information used at the time of decoding and does not use a frame memory or the like, and has a small circuit scale and an optimal ringing component. An object of the present invention is to provide a mosquito distortion reducing method and a mosquito distortion reducing apparatus that reduce the mosquito distortion.

It is another object of the present invention to provide a mosquito distortion reduction method and a mosquito distortion reduction apparatus that do not give an unnatural feeling due to the mosquito noise removal effect unnaturally changing during moving images.

[0025]

In order to solve the above problems, a mosquito distortion reducing method according to the present invention reduces a mosquito distortion generated by filtering a mosquito distortion generated when a compression-coded video signal is decoded. A reduction method, wherein a maximum value of a contour within a predetermined range at a point of interest of the video signal and a flatness in the vicinity of the point of interest are used as a filter coefficient for the filtering using a lookup table. It is a thing.

According to the above invention, the case where the point of interest of the video signal is flat and there is a large contour in the vicinity can be detected without complicated calculation by using the 2D look-up table, so that the circuit scale is small and the condition can be satisfied by the look-up table. It is possible to provide a mosquito distortion reduction method in which unnaturalness does not occur even in a moving image because the filter coefficient is switched step by step according to.

[0027]

(First Embodiment) FIG. 1 is a block diagram of a mosquito distortion reducing apparatus according to a first embodiment of the present invention. In FIG. 1, reference numeral 1 is an input terminal to which a decoded video signal is input as, for example, an 8-bit digital signal, and 2 is an output terminal.

Reference numeral 3 is a filter means for switching the filter characteristics of the video signal according to a given filter coefficient,
Reference numeral 4 is a maximum contour detecting means for obtaining the maximum value of the contour in a predetermined range from the point of interest of the video signal.

Reference numeral 5 is a flatness detecting means for obtaining the flatness near the point of interest of the video signal, 6 is a first quantizing means for making the value of the maximum contour detecting means 5 coarser, and 7 is a flatness detecting means 5 It is a second quantizing means for making the value of R rougher.

Reference numeral 8 is a 2D lookup table for performing table conversion based on the two values from the first quantizing means 6 and the second quantizing means 7 to find the coefficient of the filter means 3.

The operation of the mosquito distortion reducing device configured as described above will be described with reference to FIG. FIG. 2 is a schematic diagram in which the luminance of the video signal at the point of interest is displayed in the horizontal direction. The flatness at the point of interest in the figure is obtained by the flatness detecting means 5, and the maximum contour detecting means 4 obtains the maximum contour value in the figure.

FIG. 2 shows the mosquito noise that is ringingly generated near the contour of the image, and thus the contour maximum value is large and the flatness of the point of interest (in this figure, the smaller the value, the flatter the image becomes. ) If small, consider ringing.

Next, the operation of requantizing the detected value by the first quantizing means and the second quantizing means will be described with reference to FIG. FIG. 3 shows an example of input / output characteristics of the first quantizing means and the second quantizing means. The horizontal axis shows the respective inputs from the flatness detecting means 5 and the maximum contour detecting means 4, and the vertical axis shows the output values after quantization.

Numeral 9 is the characteristic of the second quantizing means, and 10
Is a characteristic of the first quantizing means. In this way, the circuit scale is reduced by performing quantization before being input to the 2D lookup table 8.

FIG. 4 shows an example of the 2D lookup table 8. In the figure, the larger the number, the larger the smoothing characteristic of the filter. As in this example, the filter coefficient is switched according to the maximum contour value and the flatness.

As described above, in the first embodiment of the present invention, the case where the point of interest of the video signal is flat and there is a large contour in the vicinity can be detected by using the 2D lookup table without complicated calculation. Therefore, the circuit scale is small, and since the filter coefficient is switched step by step by the look-up table, unnaturalness does not occur even in a moving image, and it is possible to effectively reduce mosquito distortion with a simple circuit. .

In the first embodiment of the present invention, only the horizontal direction is taken into consideration, but it goes without saying that the same effect can be obtained even if it is applied in the vertical direction. Further, although the quantization has 32 steps, it goes without saying that the number of steps may be larger or smaller than this.

(Embodiment 2) In the above-mentioned embodiment, when the flatness of the point of interest is so low that it cannot be ignored compared with the maximum value of the contour (when the amplitude is large), the first and second quantizations are performed. If the size of the 2D look-up table is reduced by roughly quantizing by means, malfunctions increase.

Particularly, it is difficult to distinguish the mosquito noise generated in the small-amplitude contour from the original texture, and simply L
There is a problem that the circuit scale increases remarkably when the accuracy of the UT is increased. The second embodiment of the present invention aims to solve the above problems.

FIG. 5 is a block diagram of a mosquito distortion reducing apparatus according to the second embodiment of the present invention. The blocks having the same functions as those in FIG. 1 are designated by the same reference numerals, and the description thereof will be omitted.

In FIG. 5, 11 is the maximum contour value detecting means 4.
Comparing means for comparing the magnitude relationship between the output of the above and the output of the flatness detecting means 5 and outputting the result. Reference numeral 12 is a switching means for switching between the output of the 2D lookup table and the value of the coefficient value holding means 13 according to the result of the comparison means 11, and 13 is a coefficient holding means for holding a coefficient for which the filter means 3 is turned off. .

The operation of the second embodiment configured as described above will be described with reference to FIG. FIG. 6 shows the first
6 is a schematic diagram showing a situation in which a problem occurs in the embodiment of FIG. 6, and FIG. 6 is an example of input / output characteristics of the first quantizing means and the second quantizing means, as in FIG. The horizontal axis shows the respective inputs from the flatness detecting means 5 and the maximum contour detecting means 4, and the vertical axis shows the output values after quantization.

Numeral 9 is the characteristic of the second quantizing means and 10
Is a characteristic of the first quantizing means. As shown in the figure, when the output of the first quantizing means is 1 and the output of the second quantizing means is 3, filtering is performed even if the amplitude of the point of interest is larger than the contour.

Therefore, the comparison means 11 detects the magnitude relationship in advance before quantization, and when the amplitude at the point of interest is large, the filter coefficient is forcibly used without using the value from the 2D lookup table 8. The switching operation is performed by the switching means 12 so that the value becomes.

As described above, according to the present embodiment, the LUT for comparing the maximum value of the contour with the flatness of the flatness of the target point before quantization and switching whether to use the LUT or not depending on the result. The mosquito noise can be discriminated with high accuracy regardless of the accuracy, and the advantageous effect that a mosquito noise reduction device with less malfunction and less malfunction can be obtained.

(Embodiment 3) In the above-described embodiment, the maximum contour detecting means detects the maximum value while holding the contour detection result, so that there is a problem that the circuit scale is large. The third embodiment of the present invention aims to solve the above problems.

FIG. 7 is a block diagram of the maximum contour detecting means in the third embodiment of the present invention. Reference numeral 41 is a contour detecting means for detecting a contour of the input video signal, 42 is a quantizing means for further roughly quantizing the detected contour value, and 43 is a predetermined range of quantized contours and holds the maximum value therebetween. It is a maximum value detecting means.

The operation of the third embodiment configured as described above will be described. The contour detection means 41 can be obtained by the absolute value of the difference between two pixels, for example. In this case, 8-bit data is obtained with an 8-bit video signal.

Next, when the quantizing means 42 quantizes into, for example, 16 steps, that is, 4 bits, the maximum value detecting means 4
With 3, the memory size required to hold the value is half. Also, the number of bits of the comparison circuit required to obtain the maximum value is reduced.

Therefore, by giving the output of the maximum contour detecting means 4 to the 2D lookup table 8 as it is, the same operation as that of the first embodiment can be realized with a smaller circuit scale.

As described above, in the present embodiment, the quantization is performed before the maximum value is detected, so that the advantageous effect that the circuit scale can be reduced can be obtained.

(Embodiment 4) In the above embodiment, 2
There is a problem that the D lookup table is large.
The fourth embodiment of the present invention aims to solve the above problems. The operation of this embodiment will be described with reference to FIG.

In FIG. 4, most of the coefficients in the lower left part are 0. Therefore, the table can be made smaller by not having the table data of this portion from the beginning. By using the circuit of the second embodiment, it is possible to implement without adding any circuit.

As described above, in the present embodiment, the 2D look-up table 8 can be made small without adding any circuit, so that the advantageous effect that the circuit scale can be further reduced can be obtained.

[0055]

As described above, according to the present invention, there is provided a mosquito distortion reducing method which has a small circuit scale, does not cause unnaturalness in moving images, and can effectively reduce mosquito distortion with a simple circuit. It becomes possible to provide.

[Brief description of drawings]

FIG. 1 is a block diagram of a mosquito distortion reduction device according to a first embodiment.

FIG. 2 is a schematic diagram in which the luminance of a video signal at a point of interest is displayed in the horizontal direction.

FIG. 3 is a diagram showing an example of input / output characteristics of a first quantizing means and a second quantizing means.

FIG. 4 is a diagram showing an example of a 2D look-up table 8.

FIG. 5 is a block diagram of a mosquito distortion reduction device according to a second embodiment of the present invention.

FIG. 6 is a schematic diagram showing how a problem occurs in the first embodiment.

FIG. 7 is a block diagram of maximum contour detecting means according to a third embodiment of the present invention.

FIG. 8 is a block diagram of a conventional mosquito distortion reduction device.

[Explanation of symbols]

1 Video input terminal 2 Video output terminal 3 filter means 4 Maximum contour detection means 5 Flatness detection means 6 First quantization means 7 Second quantizing means 8 2D lookup table 9 Characteristics of second quantizing means 10 Characteristics of first quantizing means 11 Comparison means 12 Switching means 13 Coefficient value holding means 41 outline detection means 42 Quantization means 43 Maximum value detection means 52 Filter size image extraction circuit 53 Large area pixel data extraction circuit 54 Small area pixel data extraction circuit 55 Timing clock generator 56, 57, 58 memory 59, 60, 63 Delay circuit 61 and 62 pixel information extraction circuit 64 threshold calculation circuit 65 ε-filter

Continued front page    F term (reference) 5C059 KK03 KK04 MA00 MA23 MC11                       MC38 ME01 PP04 PP22 PP23                       TA69 TB08 TC33 TC34 TC43                       TD02 TD08 TD12 UA02 UA05                       UA11                 5J064 AA01 BB07 BC01 BC11 BC16                       BC25

Claims (7)

[Claims] 1. A mosquito distortion reducing method for reducing a mosquito distortion generated when a compression-encoded video signal is decoded by filtering, wherein a contour within a predetermined range at a point of interest of the video signal is reduced. A mosquito distortion reduction method characterized in that a filter coefficient of the filtering is obtained using a look-up table from the maximum value and the flatness near the point of interest. 2. A mosquito distortion reducing apparatus comprising filter means for filtering while changing characteristics at a point of interest according to a given coefficient, the mosquito distortion occurring when a compression-encoded video signal is decoded, A maximum contour value detecting means for detecting a maximum value of the contour of the video signal within a predetermined range from the point of interest; a first quantizing means for quantizing the maximum value of the contour with coarser accuracy; Flatness detecting means for detecting the flatness at the point of interest, second quantizing means for quantizing the flatness with coarser precision, output of the first quantizing means, and the second quantizing means. 2D look-up table which outputs the filter coefficient by referring to two of the outputs of the above as a table, and a mosquito distortion reducing apparatus. 3. The mosquito according to claim 1, wherein the output of the 2D look-up table is used only when the result of comparing the flatness result and the maximum value of the contour satisfies a predetermined condition. Distortion reduction method. 4. A comparison means for comparing the output of the maximum contour value detection means and the output of the flatness detection means with each other, and the output of the 2D lookup table or a predetermined value using the result of the comparison means. The mosquito distortion reducing apparatus according to claim 2, further comprising a switching unit that switches between coefficient values. 5. The mosquito distortion reduction according to claim 1, wherein the maximum contour detecting means detects the maximum value within a predetermined range from the point of interest after performing quantization after the contour is detected. Method. 6. The maximum contour value detecting means, a contour detecting means for detecting a contour of a video signal, a quantizing means for quantizing the contour with coarser precision, and an output of the quantizing means for a predetermined pixel. 2. A maximum value detecting means for holding the range for the maximum value and obtaining the maximum value between the ranges.
The described mosquito distortion reducing device. 7. The 2D look-up table eliminates predetermined components.
The described mosquito distortion reducing device.