KR100463508B1 - Gain Control Method for Noise Reduction of Video Encoder - Google Patents

Abstract

The present invention relates to a gain control method for noise removal of a video encoder. In particular, the energy of the noise component and the pure signal component is calculated to perform variable gain control of the filtering path according to the energy amount of the noise component relative to the entire signal component. A gain control method for removing noise of a video encoder.

As shown in Fig. 4, the present invention separates the noise component and the pure signal component of the video signal to be filtered, calculates the respective energy, and variably controls the gain (a) on the filtering path according to the energy ratio. Since the signal is separated into noise components and pure video signal components, filtering gain control considering energy ratios is performed, so that accurate filtering on the actual noise components can be performed and image quality can be improved accordingly.

Description

Gain Control Method for Noise Reduction of Video Encoder

The present invention relates to a gain control method for noise removal of a video encoder. In particular, the energy of the noise component and the pure signal component is calculated to perform variable gain control of the filtering path according to the energy amount of the noise component relative to the entire signal component. A gain control method for removing noise of a video encoder.

Background Art [0002] Noise removal, for example, in an MPEG2 video encoder as a video encoder, relies on gain control on a signal path represented by 1-a as shown in FIG.

FIG. 1 is a video encoder using a conventional noise reduction method, and includes a subtractor 101 for extracting an error signal of an input video signal processed in units of blocks, and 1-a (a) with respect to an output signal of the subtractor 101. FIG. Is a gain variable unit 102 performing a gain control operation of a gain), a gain controller 103 performing the gain variable control, and performing DCT conversion and quantization of an output signal of the gain variable unit 102. An adder 106 that adds and outputs the output of the DCT / quantization unit 104, the IQ / IDCT conversion unit 105 for inversely transforming the DCT and the quantized signal, and the inverse transformed signal and the motion weight information upper portion 108; ), A memory 107 for storing the added signal, and a motion estimation compensator 108 for performing motion estimation and compensation on the output signal of the memory 107. The operation is as follows. .

The input signal g (i, j, k) containing noise (i, j is an image processing unit two-dimensional coordinate on a frame, k is time axis information corresponding to i, j on a time axis frame), and a noise component ( It is regarded as the sum of n (i, j, k)) and the signal component f (i, j, k).

The input signal g (i, j, k) is a pixel f (i-dx, j-dx, corresponding to g (i, j, k) from the image of the previous frame that has already been filtered by the subtractor 101. k-1) (predicted by the motion estimation compensation unit 108) and subtracted into the gain variable unit 102, and the gain controller 103 obtains the gain of the gain variable unit 102 in accordance with the magnitude of the difference value. The noise is reduced by increasing and decreasing (a).

That is, as shown in Fig. 2, the difference between the input signal and the prediction signal d (i, j, k) = g (i, j, k) -f (i-dx, j-dx, k-1) = f ( i, j, k) -f (i-dx, j-dx, k-1) + n (i, j, k) is calculated, and if the motion estimation and compensation are correct, the difference signal d (i, j , k)) will consist mainly of noise component n (i, j, k), so adjust gain (a) to be close to '1' to remove all d (i, j, k), If the compensation is not correct, d (i, j, k) will also contain many signal components, f (i, j, k) -f (i-dx, j-dx, k-1), so n ( Even if i, j, k) cannot be removed, gain (a) is adjusted to '0' so that d (i, j, k) is not removed.

That is, as shown in Fig. 2, the absolute value calculation of the difference signal and the noise value is performed, and then the average value of this value is compared with the reference value Ref to increase or decrease the gain a according to the magnitude.

The gain-controlled signal in which noise is reduced is input to the DCT / quantization unit 104 and converted into frequency information by Discrete Cosine Transform (DCT) and quantization, and the quantized signal is converted into IQ / IDCT conversion unit 105. Inverse quantization and inverse DCT conversion are input into the memory 107 through the adder 106.

The memorized signal is output by the motion estimation compensator 108 after motion estimation and compensation are performed.

A video encoder employing such a conventional noise removing method has a signal component f (i, j, k) -f (i-dx, j-dx, k-1) for the difference signal d (i, j, k). The magnitude of the sum of the difference signals was used as a criterion to determine how much of the difference was included.

Therefore, since the sum of the difference signals includes both the signal component and the noise component without distinguishing them, it is impossible to distinguish whether the signal component is large or the noise component is large. Therefore, when the noise component is larger than the signal component or vice versa, There is a fear that unnecessary filtering is performed, which cannot cope with each other, and removes not only the noise but also the video signal.

The present invention separates the noise component and the pure signal component of the video signal to be filtered, calculates the respective energy, and variably controls the gain (a) on the filtering path according to the energy ratio to improve the noise removal performance and the image quality. A gain control method for noise reduction of a video encoder is provided.

3 is an embodiment of a video encoder to which the noise removing method of the present invention is applied, and includes a subtractor 301 for extracting an error signal of an input video signal processed in units of blocks, and an output signal of the subtractor 301. a gain variable unit 302 which performs a gain control operation of a -a (a is gain), calculates energy of noise components and pure signal components of the input video signal, and performs variable control of the gain according to the energy ratio A gain controller 303, a DCT / quantizer 304 for performing DCT conversion and quantization of the output signal of the gain variable section 302, and an IQ / IDCT converter for inversely converting the DCT and the quantized signal. 305, an adder 306 for adding and outputting the inverse transformed signal and the output of the motion tracking information 308, a memory 307 for storing the added signal, and an output signal of the memory 307 To perform motion estimation and compensation for Jikim consists of an upper estimation information 308. The

A gain control method for noise removal of a video encoder of the present invention applied to a video incubator having the above structure includes a signal separation process for separating noise components and pure signal components of an input video signal, and the separated noise components and pure signals. An energy calculation process for calculating energy of each component and a gain control process for variably controlling the gain on the filtering path according to the calculated energy ratio.

The noise removing operation for video encoding according to the present invention will be described with reference to FIGS. 3 to 6 as follows.

The input signal g (i, j, k) is a pixel f (i-dx, j-dx, corresponding to g (i, j, k) from an image of a previous frame that has already been filtered by the subtractor 301. k-1) (predicted by the motion estimation compensator 308) and subtracted into the gain variable part 302, and gain controller 303 according to the energy ratio of pure signal component and noise component in the difference signal. Decreases the noise by increasing or decreasing the gain a of the gain variable section 302.

That is, as shown in Fig. 4, the difference between the input signal and the prediction signal d (i, j, k) = g (i, j, k) -f (i-dx, j-dx, k-1) = f ( i, j, k) -f (i-dx, j-dx, k-1) + n (i, j, k) is input from the subtractor 301 and filtered to obtain pure signal components and noise components. Isolate.

For example, low-pass filtering results in a signal with reduced noise components, which are considered pure signal components.

FIG. 5 is a diagram showing this. The difference signal d (i, j, k) is obtained by performing a difference value subtraction 503 between the input video signal 501 and the signal 502 predicted by the motion estimation compensator. In addition, low pass filtering is performed to obtain the low pass filtered image signal 504 and the predicted signal 505, and the difference value subtraction 506 is performed, and a 3 * 3 window is set around the pixel to be filtered. Each signal component (e (k) = original signal, e _L (k) is a low pass filtered signal) is obtained from the variance value.

The energy is found by calculating the sum of the squares of the respective energy differences from the average value of the windows.

The energy E {{f (i, j, k) -f (i-dx, j-dx, k-1)} ² } and E ({{n (i, j, k)) for the signal component thus obtained } Gain (a) is controlled corresponding to the ratio (B / (A + B)) of} ² }.

Where A is signal component energy and B is noise component energy.

As mentioned above, since pure video signal components and noise components relatively affect variable gain control on the filtering path according to the energy ratio of each signal component, it is optimal to determine how much noise components are included in the input video signal. Noise filtering can be performed on condition.

The gain-controlled signal in which noise is reduced is input to the DCT / quantization unit 304 and converted into frequency information by Discrete Cosine Transform (DCT) and quantization, and the quantized signal is converted into IQ / IDCT conversion unit 305. Inverse quantization and inverse DCT conversion are then input to the memory 307 through the adder 306.

The memorized signal is output by the motion estimation compensator 308 after motion estimation and compensation are performed.

On the other hand, as described above, the method for separating the image signal component and the noise component from the input signal containing noise, as well as the method of performing low pass filtering in pixel units as described above, as well as the average value of the difference signal block The signal component and the noise component can be extracted in three ways: by taking the unit and by taking the adaptive average value in consideration of the abrupt displacement of adjacent pixels.

The average value is calculated in block units by simply accumulating the difference values for g (i, j, k) and f (i, j, k-1) and dividing them by the block size.

The low-pass filtering method by pixel unit gives a weight (L) according to the positions of i and j for the difference between g (i, j, k) and f (i, j, k-1) and filters them. It is simply obtained by accumulating as much as the width (-W, + W) with adjacent pixels on the basis of the pixel to be used.

The method of obtaining the adaptive weighted average on a pixel basis is also the same as the low pass filtering on a pixel basis. In this case, the amplitude of the noise is considered in consideration of the amplitude of the noise and the displacement of the pixels in the front and rear positions with respect to the pixel to be filtered. Based on the result of accumulating between -W and + W, the difference between the displacement of the and the adjacent pixels is large. If the difference is large, the weight is reduced and if it is small, the weight is increased.

According to the noise removing method and apparatus for video encoding according to the present invention, since the input video signal is separated into a noise component and a pure video signal component to perform filtering gain control considering each energy ratio, Accurate filtering is possible and image quality can be improved accordingly.

1 is a block diagram of an MPEG2 video encoder to which a conventional noise removing method is applied.

2 is a view for explaining a problem in the conventional noise removing method.

3 is a block diagram of a video encoder to which the noise removing method of the present invention is applied.

4 is a view for explaining a noise removing method of the present invention.

5 is a view for explaining an energy calculation method in the noise removing method of the present invention.

Claims (5)

A signal separation process for separating the noise component and the pure signal component of the input image signal, an energy calculation process for calculating the respective energy of the separated noise component and the pure signal component, and a gain in the filtering path according to the calculated energy ratio Gain control method for removing noise of the video encoder, characterized in that consisting of a gain control process for variable control. The gain control method according to claim 1, further comprising adding weights to the extracted signal components in consideration of displacements of the video signals immediately before and after. The gain control method of claim 1 or 2, wherein the signal separation process is performed by a block value of a video signal in units of blocks. The gain control method of claim 1 or 2, wherein the signal separation process is performed by performing low pass filtering on a pixel-by-pixel basis. The method of claim 1 or 2, wherein the signal separation process is obtained by applying an adaptive weight corresponding to the displacement degree in consideration of the displacement degree with respect to the pixels in the front and rear positions based on the pixel to be filtered. Gain Control Method for Noise Reduction of Video Encoder.

Images