JP2006148365A - Noise reduction method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the problem in which image processing speed is low because a filter with a smoothing effect is applied to each pixel for pixel values after inverse orthogonal transformation in order to reduce mosquito noise. <P>SOLUTION: A noise level setting section 32 sets a noise level on the basis of all of comparison results comprising: the magnitude comparison between a quantization step and a threshold value T1; the magnitude comparison between a coefficient value of a DC component of a DCT coefficient obtained by inverse quantization at an inverse quantization section 22 and a threshold value T2; the magnitude comparison between a total sum of the absolute values of AC components of the DCT coefficients and that of a prescribed peripheral block; and the magnitude comparison between the total sum of the absolute values of the AC components of the DCT coefficients and that of a corresponding block of a preceding frame, as to a block where a mosquito noise detection section 31 detects the existence of mosquito noise, and controls a parameter for filtering of a mosquito filtering section 33. The mosquito filtering section 33 applies the filtering only to the decoded DCT coefficients of an intra-frame. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

The present invention relates to a noise reduction method, and more particularly, to a noise reduction method for reducing mosquito noise generated by orthogonal transform and lossy encoding of an image signal.

  In recent years, with the spread of the Internet, DVD (Digital Versatile Disk) recorders, and the like, digitally represented images are becoming more and more common. As a technique for efficiently expressing these images with a small code amount, orthogonal transform coding methods represented by JPEG (Joint Photographic Experts Group), MPEG (Moving Picture Experts Group), and the like are widely used. In this encoding method, orthogonal transformation and quantization processing are performed in order to reduce the amount of data using the characteristics of the image signal. The coefficient value after orthogonal transformation reduces the amount of data by limiting the high-frequency component in particular by quantization and increases the coding efficiency. On the other hand, as the quantization step increases, the noise increases in the reproduced image, and the image quality increases. Deterioration becomes remarkable.

  Among the above-mentioned noises, there is known noise that is called mosquito noise especially around the edge of an image, and it looks like mosquito clings around the edge. This is because the information around the edge is lost due to the restriction of the high frequency component particularly in the quantization processing, and the image quality is deteriorated.

  Thus, several noise reduction methods for improving the image quality by filtering the mosquito noise have been proposed (see, for example, Patent Documents 1 and 2). The conventional noise reduction method described in Patent Document 1 is a device that divides image data into a plurality of blocks and performs encoding for each block. Based on the image data for each block, a difference between adjacent pixels for each block is described. A configuration comprising: a determination unit that obtains a frequency distribution of values and determines whether the image data includes an edge based on the frequency distribution; and a unit that selects a block from which noise is removed based on a determination result by the determination unit To reduce mosquito noise.

  The conventional noise reduction method described in Patent Document 2 detects mosquito noise by detecting the absolute sum of motion vectors and transform coefficients in a decoder or the presence of transform coefficients in a specific range in orthogonal transform coding. The mosquito noise is reduced by a configuration in which each pixel is subjected to filtering of a type corresponding to the mosquito noise.

Japanese Patent No. 3011828 JP 2001-204029 A

  However, in each of the conventional noise reduction methods described in Patent Documents 1 and 2 above, in order to reduce mosquito noise, a filter having a smoothing effect is applied to each pixel value after inverse orthogonal transformation for each pixel. Therefore, the problem is that the operation is slow.

  The present invention has been made in view of the above points, and an object thereof is to provide a noise reduction method capable of reducing mosquito noise by high-speed operation by filtering coefficient values before inverse orthogonal transform.

  In order to achieve the above-mentioned object, the present invention achieves the above-described object, by performing intra-image coding, in which image signals are subjected to orthogonal transform, quantization, and variable-length coding for each predetermined image region in sequence only within one image, and a plurality of images. Mosquito noise reduction that reduces mosquito noise in an encoded image signal that has been intra-coded when decoding an encoded image signal obtained by performing either of inter-picture predictive encoding performed sequentially between A method comprising: a first step of sequentially performing variable length decoding and inverse quantization on a supplied encoded image signal to decode orthogonal transform coefficients; A second step of detecting mosquito noise in image area units using the orthogonal transform coefficient decoded in the first step when the signal is supplied, and an image in which mosquito noise is detected in the second step In the area The first comparison result obtained by comparing the quantization step at the time of intra-picture coding and the first threshold, the coefficient value of the DC component of the decoded orthogonal transform coefficient, and the second threshold The sum of the absolute values of the AC components of the orthogonal transform coefficients decoded between the second comparison result obtained by the magnitude comparison and the image region and the peripheral image region in which the mosquito noise around the image region is detected. The third comparison result obtained by comparing the size of the image and the sum of the absolute values of the AC components of the orthogonal transform coefficients decoded between the image region and the corresponding image region of the previous image are compared in size. A third step of setting a noise level based on at least one of the fourth comparison results obtained, and a decoded orthogonal transform of the image region in which mosquito noise is detected in the second step Set in the third step for the coefficient A fourth step of performing filtering to reduce the coefficient value of the orthogonal transform coefficient by varying at least one parameter of the filter strength, the range to be filtered, and the threshold value of the orthogonal transform coefficient in accordance with the received noise level; It is characterized by including.

  In the present invention, a noise level corresponding to the level of the detected mosquito noise is set, and at least one parameter of the filter strength, the range to be filtered, and the threshold of the orthogonal transform coefficient is varied according to the noise level. Since the filtering for reducing the coefficient value of the orthogonal transform coefficient is performed, the change in adjacent pixels of the decoded image signal obtained by performing the inverse orthogonal transform on the filtered orthogonal transform coefficient can be reduced.

  Further, according to the present invention, when decoding an encoded image signal obtained by performing either intra-picture encoding or inter-picture predictive encoding, only the encoded image signal that has been intra-encoded is filtered. Since it is a target, the pixel of the decoded image signal after inverse orthogonal transform can be operated at a higher speed than the case of filtering every pixel over the entire frame.

  According to the present invention, at least one parameter of the filter strength, the range to be filtered, and the threshold value of the orthogonal transform coefficient is varied according to the noise level of the mosquito noise in the image area where the mosquito noise is detected, and the orthogonality is obtained. By performing filtering to reduce the coefficient value of the transform coefficient, it is possible to appropriately reduce the change in adjacent pixels of the decoded image signal obtained by performing inverse orthogonal transform on the filtered orthogonal transform coefficient, and to code the intra-coded code. Since only the digitized image signal is targeted for filtering, image quality deterioration due to excessive filtering can be suppressed, and mosquito noise can be reduced quickly and effectively.

  Next, the best mode for carrying out the invention will be described with reference to the drawings. Hereinafter, as an example, a case where mosquito noise is reduced when decoding a code after being encoded by MPEG2 will be described. First, a known MPEG2 encoder unit and decoder unit will be described.

  FIG. 1 is a block diagram showing an example of an MPEG2 encoder unit. In FIG. 1, a subtractor 11, an ME unit 12 that performs motion estimation (ME), a DCT unit 13 that performs discrete cosine transform (DCT), a quantization unit 14, an inverse quantization unit 15, an inverse DCT unit 16, a variable length code And the adder 18.

  The encoder unit 10 receives an image signal as an input signal a, and performs an intra-picture encoding or an inter-picture prediction encoding on the input signal a for each image of a frame or a field. The code b is output. Here, for convenience of explanation, it is assumed that the encoder unit 10 performs intraframe encoding or interframe predictive encoding in units of frames. A frame that obtains an intra-frame coded image by intra-frame coding is called an intra frame, and a frame that obtains an inter-frame predictive coded image by inter-frame predictive coding is called an inter frame.

In an intra frame, the encoder unit 10 supplies the input signal a to the DCT unit 13 via the subtractor 11 as it is, and is a discrete type which is a kind of orthogonal transform in block units of a two-dimensional pixel group having a predetermined number of pixels. A cosine transform (DCT) is performed to obtain a transform coefficient (DCT coefficient). Here, when the DCT unit 13 applies DCT in units of blocks of 8 × 8 pixels to the pixels x _{i, j} at the positions i, j in the horizontal and vertical directions of the input signal, the horizontal and vertical frequencies represented by the following equations: DCT coefficients Xi, j at the respective positions u and v are obtained.

図３は８×８画素のブロック単位のＤＣＴ係数を示したものであり、一つの正方形が一つのＤＣＴ係数を表わす。ＤＣＴ係数は、図３中、右へ向かうほど高い水平周波数成分の係数となり、同図中、下へ向かうほど高い垂直周波数の係数となる。ここで、最低周波数成分のＤＣＴ係数ｄ_００は直流成分であるのでＤＣ係数、この直流成分以外は交流成分であるのでＡＣ係数である。

FIG. 3 shows a DCT coefficient in a block unit of 8 × 8 pixels, and one square represents one DCT coefficient. The DCT coefficient becomes a higher horizontal frequency component coefficient toward the right in FIG. 3, and becomes a higher vertical frequency coefficient toward the lower side in FIG. Here, the DCT coefficient d ₀₀ of the lowest frequency component is a DC coefficient because it is a direct current component, and an AC coefficient because other than this direct current component is an alternating current component.

  The encoder unit 10 supplies the above DCT coefficients to the quantization unit 14, performs quantization in the quantization step to reduce the amount of information, and then converts the signal output from the quantization unit 14 to the variable length encoding unit 17. The intra-frame image independently encoded with only the image in the frame is output as the output code b.

  The variable length code by the variable length coding unit 17 performs differential pulse code modulation (DPCM), which is a kind of predictive coding, on the DC (direct current) component of the output signal of the quantization unit 14. For AC (alternating current) components, a zigzag scan is performed from low to high, and a zero run length and effective coefficient value are considered as one event, and codes with a short code length are assigned in descending order of appearance probability. Perform Huffman coding.

  The variable length encoded signal (output code b) output from the variable length encoding unit 17 is temporarily stored in a buffer (not shown) and output at a predetermined transfer rate, while a predetermined unit (for example, a macroblock unit). It is known that the quantization step of the quantization unit 14 is variably controlled so that the generated code amount for each becomes the target code amount, and since it is well known, detailed description thereof is omitted.

  In the above intra frame, the variable length encoded signal output from the variable length encoding unit 17 is supplied to the inverse quantization unit 15, where it is inversely quantized and then further inverse DCTed by the inverse DCT unit 16. Then, a primary reference image (reproduced image) for performing motion estimation with the input signal a is stored in the memory in the ME unit 12 via the adder 18. Here, the inverse DCT by the inverse DCT unit 16 can be expressed by the following equation.

また、エンコーダ部１０は、インターフレームでは、ＭＥ部１２において入力信号ａと内部のバッファから読み出した上記の参照画像とを用いて公知の方法で動き推定信号を生成して減算器１１に供給し、ここで入力信号ａとの差分信号を得る。その後、エンコーダ部１０は、この差分信号をＤＣＴ部１３でＤＣＴを施して変換係数を得て、量子化部１４で量子化し、その量子化部１４の出力信号を可変長符号化部１７に供給して可変長符号化を行うことでフレーム間予測符号化されたインターフレーム画像を出力符号ｂとして出力する。

Further, in the inter frame, the encoder unit 10 generates a motion estimation signal by a known method using the input signal a and the reference image read from the internal buffer in the ME unit 12 and supplies the motion estimation signal to the subtractor 11. Here, a differential signal from the input signal a is obtained. Thereafter, the encoder unit 10 performs DCT on the difference signal in the DCT unit 13 to obtain a transform coefficient, quantizes the quantized unit 14, and supplies the output signal of the quantizing unit 14 to the variable length coding unit 17. Then, by performing variable length coding, an inter-frame image that has been inter-frame prediction coded is output as an output code b.

  Although not shown, the ME unit 12 receives a motion vector and a picture type (a P picture that is an inter-frame forward prediction image in an inter frame or a B picture that is a bidirectional prediction image, and an intra-frame encoded image in an intra frame. (I picture) is output, and these are also output via the variable length encoding unit 17.

  Also in the inter frame, the variable length encoded signal output from the variable length encoding unit 17 is supplied to the inverse quantization unit 15, where it is inversely quantized and then further inverse DCT by the inverse DCT unit 16. Then, it is temporarily stored as a reference image (reproduced image) for reference for performing motion estimation with the input signal a in the memory in the ME unit 12 via the adder 18.

  Next, the decoder unit will be described. FIG. 2 is a block diagram showing an example of a decoder unit of MPEG2. In FIG. 2, a decoder unit 20 includes a variable length decoding unit 21, an inverse quantization unit 22, an inverse DCT unit 23, an adder 24 that outputs an output signal e, information d such as a motion vector and a picture type, and an output signal e. MC (motion compensation) unit 25 is input.

  The decoder unit 20 generates an MPEG2 bit stream generated by the encoder unit 10 and output as an output code b in which time-series synthesis of an intra frame image and an inter frame image is performed via a transmission path (not shown). Then, after the variable length decoding unit 21 performs variable length decoding on the input code c, the inverse quantization unit 22 performs inverse quantization to obtain a transform coefficient. The transform coefficient is subjected to inverse DCT in the inverse DCT unit 23.

  In the intra frame, the restored image signal is output from the inverse DCT unit 23 and output as an output signal e. In the inter frame, decoding is performed by performing MC (motion compensation) in the MC unit 25 using the output signal e and the information d such as the motion vector and the picture type obtained by the variable length decoding unit 21. Made. The output signal of the MC unit 25 is added to the signal output from the inverse DCT unit 23 by the adder 24, whereby an inter-frame decoded image signal is generated and output as the output signal e.

  Next, an embodiment of the noise reduction method of the present invention will be described. FIG. 4 is a block diagram of a decoder unit for realizing an embodiment of the noise reduction method according to the present invention. In the figure, the same components as those in FIG. The decoding unit shown in FIG. 4 is provided with a noise reduction circuit 30 including a mosquito noise detection unit 31, a noise level setting unit 32, and a mosquito noise filtering unit 33 in the decoding unit 20 having the conventional configuration shown in FIG. There are features.

  The mosquito noise detection unit 31 detects mosquito noise on a block basis by a known method based on the picture type and quantization step information f decoded by the variable length decoding unit 21 and the DCT coefficient after inverse quantization. The obtained noise detection signal g is output to the noise level setting unit 32. The noise level setting unit 32 generates and outputs a noise level signal h from the noise detection signal g, the picture type and quantization step information f, and the distribution of DCT coefficients after inverse quantization. The mosquito noise filtering unit 33 performs filtering of the intensity according to the noise level signal h on the DCT coefficient after inverse quantization, and outputs the obtained DCT coefficient to the inverse DCT unit 23.

  Here, in the noise reduction method of the present embodiment, the noise reduction operation by the noise reduction circuit 30 is performed only for intra frames, and the noise reduction operation by the noise reduction circuit 30 is not performed for inter frames. Frames for noise reduction filtering can be processed at high speed by using only intra frames, and filtering for inter frames can be performed more than twice for reference frames and inter frames. Since filtering is performed, filtering for noise reduction is performed only on intra frames, thereby suppressing image quality deterioration due to excessive filtering.

  Therefore, the operation of the decoding unit in FIG. 4 when the input code c is an intra frame will be described next. The input code c of the intra frame is supplied to the variable length decoding unit 21 and subjected to variable length decoding, and then inverse quantized by the inverse quantization unit 22 to decode the DCT coefficient. Here, in the present embodiment, the DCT coefficients decoded by the inverse quantization unit 22 are respectively supplied to the mosquito noise detection unit 31, the noise level setting unit 32, and the mosquito noise filtering unit 33 in the case of an intra frame. It is not supplied to the inverse DCT unit 23.

  On the other hand, the mosquito noise detection unit 31, for example, based on the picture type and quantization step information f obtained by the variable length decoding unit 21 and the DCT coefficient information output from the inverse quantization unit 22, for example, The mosquito noise is detected on a block basis by a known mosquito noise detection method described in Patent Document 2 and the like, and a noise detection signal g indicating whether the mosquito noise is detected is output.

  Here, mosquito noise is generated due to the addition of quantization error caused by quantizing the AC component of the DCT coefficient. Therefore, when the original image signal is only a DC component, a quantization error occurs. Almost never occurs at difficult high rates. In addition, when the quantization error for the AC component of the DCT coefficient is very large, since the block in which the DCT coefficient exists is almost only the DC component, it does not occur even at a very low rate. When mosquito noise is detected at other times, if the DCT coefficient is reduced, the change in adjacent pixels after inverse DCT is reduced, so that mosquito noise can be reduced.

  Therefore, paying attention to the above points, the noise level setting unit 32 makes the mosquito noise filtering unit 33 perform appropriate filtering according to the noise level of the mosquito noise with respect to the block in which the mosquito noise is detected. For example, a noise level signal for variably controlling the filter strength, the range to be filtered, the threshold value of the DCT coefficient, and the like is generated according to the flowchart shown in FIG.

  That is, the noise level setting unit 32 first determines whether or not the noise detection signal g indicates mosquito noise detection (step S1), and sets the noise level to “1” when mosquito noise detection is indicated. When the mosquito noise detection is not indicated (step S2), the noise level is set to “0” (step S3), and the process is terminated.

  When the noise level is set to “1”, the noise level setting unit 32 subsequently increases the mosquito level as the quantization step increases based on the picture type and the quantization step information f output from the variable length decoding unit 21. Using the increase in the intensity of noise, the quantization step of the block in which mosquito noise is detected is compared in magnitude with the first threshold T1 set in advance (step S4). When the quantization step is large, the previously set noise level is raised above “1” (step S5), and when the quantization step is equal to or less than the threshold T1, the previously set noise level is lowered below “1” (step S6). ).

  The noise level rise value or fall value is changed linearly according to the difference with respect to the threshold value of the parameter to be compared (quantization step in the above case) or non-linearly according to a preset function. Alternatively, it may be a fixed value set in advance according to whether it is large or small (the same applies hereinafter).

  Subsequently, the noise level setting unit 32, from the distribution of DCT coefficients after dequantization supplied from the dequantization unit 22, has a very bright DC component and noise is not noticeable if the DC component is very large. It is compared whether or not the coefficient value (absolute value) is larger than a predetermined second threshold value T2 (step S7), and when a comparison result is obtained in which the coefficient value (absolute value) of the DC component is larger than the threshold value T2. The noise level value is lowered below the current value (step S8). If a comparison result is obtained in which the coefficient value (absolute value) of the DC component is equal to or less than the threshold value T2, the process proceeds to step S9 without doing anything.

  Subsequently, the noise level setting unit 32 has a sum of absolute values of AC components of the DCT coefficients after the inverse quantization of the comparison target block, in each of the peripheral blocks in which the noise level is not 0 among the peripheral blocks of the block. It is compared whether or not the sum of the absolute values of the AC components is larger (step S9), and when a large comparison result is obtained, it is determined that the block has a more complex pattern and a higher intensity of mosquito noise than the surrounding blocks. Then, the value of the noise level is raised above the current value (step S10). When a comparison result is obtained in which the sum of the absolute values of the AC components of the DCT coefficients of the comparison target block is equal to or less than the sum of the absolute values of the AC components of the DCT coefficients of the neighboring blocks, the process proceeds to step S11 without doing anything.

  Further, when the sum of the absolute values of the AC components of the DCT coefficients after the inverse quantization of the comparison target block is not the noise level 0 in the block at the same position in the previous frame of the block, the noise level setting unit 32 It is compared whether or not it is larger than the sum of the absolute values of the AC components of the DCT coefficients of the blocks at the same position in the previous frame (step S11), and when a large comparison result is obtained, the comparison target block is at the same position in the previous frame. It is determined that the picture is more complex than the block and the intensity of the mosquito noise is high, and the noise level is increased from the current value (step S12). When a comparison result is obtained in which the sum of the absolute values of the AC components of the DCT coefficients of the comparison target block is equal to or less than the sum of the absolute values of the AC components of the DCT coefficients of the block at the same position in the previous frame, the process ends ( Step S13).

  In this way, the noise level setting unit 32 generates a noise level signal having a noise level value of 0 or greater than 0, and supplies the noise level signal to the mosquito noise filtering unit 33 as a control signal for the filtering parameter. Here, the filtering parameters include a filter strength, a filtering target range, a threshold value of an absolute value of a DCT coefficient to be filtered, and the like. When the noise level is 0, filtering by the mosquito noise filtering unit 33 is not performed, and the filtering unit 33 outputs the input DCT coefficient after inverse quantization as it is.

  When the noise level is “1”, the mosquito noise filtering unit 33 includes the high-frequency component (for example, the DCT coefficient in the block unit of 8 × 8 pixels shown in FIG. 6) among the AC components of the DCT coefficient. Filtering is performed to halve the coefficient value (range shown by diagonal lines). However, the coefficient value to be filtered is set to “less than 32 in absolute value” in consideration of the influence on the image quality. Note that the maximum value of the coefficient value of the AC component of the 8-bit DCT coefficient is “2040”.

  When the noise level is exactly “1”, the filtering by the mosquito noise filtering unit 33 described above assumes that the filtering target range is G, and the DCT coefficient Xu, v in the expression (3) indicating the inverse DCT is represented by X By replacing it with 'u, v, However, in the following equation, [x] is a Gaussian symbol indicating the maximum integer not exceeding x.

また、ノイズレベルが「１」より大きいノイズレベル信号ｈがモスキートノイズフィルタリング部３３に入力された場合は、モスキートノイズフィルタリング部３３は、逆量子化部２２から入力されるＤＣＴ係数に対して、上記のノイズレベルが「１」のときに比べて、フィルタ強度を強くしたり、フィルタリング対象範囲を広げたり、フィルタリング対象となるＤＣＴ係数の係数値（絶対値）の閾値を高くしたフィルタリングを行う。

When a noise level signal h having a noise level greater than “1” is input to the mosquito noise filtering unit 33, the mosquito noise filtering unit 33 applies the above-described DCT coefficients input from the inverse quantization unit 22 to the above-described DCT coefficients. As compared with the case where the noise level of “1” is “1”, filtering is performed by increasing the filter strength, expanding the filtering target range, or increasing the threshold value of the coefficient value (absolute value) of the DCT coefficient to be filtered.

  Conversely, when a noise level signal h having a noise level smaller than “1” is input to the mosquito noise filtering unit 33, the mosquito noise filtering unit 33 applies the DCT coefficient input from the inverse quantization unit 22 to the DCT coefficient. As compared with the case where the noise level is “1”, filtering is performed by reducing the filter strength, narrowing the filtering target range, or lowering the threshold value of the coefficient value (absolute value) of the DCT coefficient to be filtered.

  In this way, the mosquito noise filtering unit 33 appropriately reduces the mosquito noise by causing the DCT coefficient to be filtered in such a manner that the DCT coefficient decreases in absolute value in comparison with the noise level. The filtered DCT coefficients extracted from the mosquito noise filtering unit 33 are supplied to the inverse DCT unit 23 shown in FIG. 4 and subjected to inverse DCT, whereby the original image signal is decoded and the adder 24 Is output as an output signal i, while being temporarily stored in the memory in the MC unit 25 for use as a reference image at the time of interframe decoding.

  The above description has been made on the case where the input code c is an intra frame. However, when the input code c is an inter frame, for example, a switch means (not shown) causes the inverse DCT unit 23 to receive the same as in the conventional case. Only the DCT coefficient output from the inverse quantization unit 22 is input, and the DCT coefficient output from the mosquito noise filtering unit 33 is not input to the inverse DCT unit 23 (that is, the noise reduction circuit 30 is substantially The output signal of the inverse DCT unit 23 and the motion output from the variable length decoding unit 21, in which noise is reduced in the intra frame and temporarily stored in the memory in the MC unit 25. Decoding is performed by performing MC (motion compensation) in the MC unit 25 using information d such as a vector and a picture type, and the MC unit 2 Output signals of the inverse DCT portion 23 of the can by being added in the adder 24, the decoded image signal is output as an output signal i.

  Thus, according to the present embodiment, the higher the noise level, the stronger the filtering strength, the wider the filtering target range, or the higher the coefficient value (absolute value) threshold of the DCT coefficient to be filtered. Since the filtering is performed to reduce the DCT coefficient in accordance with the noise level, the change in adjacent pixels after inverse DCT is reduced, and mosquito noise can be reduced adaptively.

  Furthermore, according to the present embodiment, since only the intra frame of the AC component of the DCT coefficient is subjected to filtering, the pixel after inverse DCT is operated at a higher speed than when filtering every pixel over the entire frame. It becomes possible. As described above, according to the present embodiment, mosquito noise can be effectively reduced.

  Note that the present invention is not limited to the above-described embodiment. For example, in the noise level setting unit 32, the quantization step and the threshold T1 are compared in magnitude, the coefficient value of the DC component of the DCT coefficient, and the threshold T2. All of the magnitude comparison of the DCT coefficient, the sum of the absolute values of the AC components of the DCT coefficients, and the magnitude comparison of that of the predetermined peripheral block, and the magnitude comparison of the sum of the absolute values of the DC components of the DCT coefficients and the corresponding blocks of the previous frame Based on the comparison results, the noise level that controls the filtering parameters for the block in which mosquito noise was detected was set. However, the noise level based on the comparison result of one or more of these comparison results was set. A level may be set.

  In the above embodiment, MPEG2 is used as the encoding method for the mosquito noise reduction target. However, in the present invention, the data encoded by any encoding method is used for the mosquito noise reduction target. Is also applicable. In the above embodiment, when the noise level is “1”, the coefficient value of the DCT coefficient is halved during filtering. However, this is not limited to this if the coefficient value is reduced. Absent. Further, the coefficient value (absolute value) may be made smaller as the frequency component of the DCT coefficient is higher.

  Further, the filtering target range of the mosquito noise filtering unit 33 is not limited to the range shown in FIG. 6 as long as it is an AC component other than the DC component of the DCT coefficient. Further, the threshold “ Of course, “32” is an example, and the present invention is not limited to this.

  The present invention also includes a program for realizing the operation function of the noise reduction circuit 30 by a computer. This program may be read from a recording medium and loaded into a computer, or may be transmitted via a communication network and loaded into a computer.

It is a block diagram which shows an example of the encoder part of MPEG2. It is a block diagram which shows an example of the decoder part of MPEG2. It is a figure which shows the DCT coefficient of the block unit of 8x8 pixel. It is a block diagram of the decoder part which implement | achieves one embodiment of the method of this invention. It is a flowchart for operation | movement description of the noise level setting part in FIG. It is a figure explaining an example of a filtering object range among DCT coefficients of a block unit of 8x8 pixels.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Encoder part 11 Subtractor 12 ME part 13 DCT part 14 Quantization part 15, 22 Inverse quantization part 16, 23 Inverse DCT part 17 Variable length encoding part 18, 24 Adder 20 Decoder part 21 Variable length decoding part 25 MC unit 30 Noise reduction circuit 31 Mosquito noise detection unit 32 Noise level setting unit 33 Mosquito noise filtering unit

Claims (1)

Intra-image coding in which image signals are orthogonally transformed, quantized, and variable-length coding for each predetermined image area sequentially within only one image, and inter-picture predictive coding in which images are sequentially performed between a plurality of images. A mosquito noise reduction method for reducing mosquito noise of an encoded image signal that has been intra-encoded when decoding an encoded image signal obtained by performing any of the following:
A first step of sequentially performing variable length decoding and inverse quantization on the supplied encoded image signal to decode orthogonal transform coefficients;
A second detection unit that detects the mosquito noise in units of the image area using the orthogonal transform coefficient decoded in the first step when the encoded image signal that has been intra-encoded is supplied; Steps,
In the image region where the mosquito noise is detected in the second step, a first comparison result obtained by comparing the quantization step at the time of the intra-coding and a first threshold value, The second comparison result obtained by comparing the coefficient value of the DC component of the decoded orthogonal transform coefficient with the second threshold value, and the periphery in which the mosquito noise around the image region and its periphery is detected A third comparison result obtained by comparing the sum of the absolute values of the AC components of the orthogonal transform coefficients decoded with respect to the image area, and a corresponding image area one image before the image area; A noise level is set based on at least one comparison result among the fourth comparison results obtained by comparing the sum of absolute values of the AC components of the orthogonal transform coefficients decoded between And the steps
The orthogonal transform coefficient decoded in the image area in which the mosquito noise is detected in the second step is subjected to filter strength and filtering according to the noise level set in the third step. And a fourth step of performing filtering to reduce at least one parameter of the range and the threshold value of the orthogonal transform coefficient to reduce the coefficient value of the orthogonal transform coefficient.

Images