JP5325164B2 - Coding block size and position estimation method and program - Google Patents

Description

  The present invention relates to a coding block size and position estimation method and a program. In particular, the present invention relates to a coding block size indicating a horizontal size and a vertical size of a rectangular block (coding block) which is a processing unit of discrete cosine transform of video coding from a video signal, and a boundary between the coding blocks. The present invention relates to a technique for estimating a coding block position indicating a line position.

  With the development of information technology, services using digital video media (IPTV services, video conferencing systems, etc.) and products (Blu-ray, DVD, etc.) are becoming popular. In general, information compression (video coding) is performed on these video media to reduce the amount of data. However, with video coding, the video media deteriorates (block-like distortion (block noise), blur, flicker, etc.). ), And the quality perceived by humans for video media (QoE: Quality of Experience (QoE)) decreases.

  In order to appropriately grasp that the video quality has deteriorated as described above, a video objective quality evaluation technique capable of evaluating the user experience quality for the video media is essential.

  In the video objective quality evaluation technology, a video signal (that is, pixel signal), packet header information, bit stream information (that is, encoded information such as a motion vector), or a combination thereof is input, and the user experience quality is evaluated. Technology. In particular, it is important to evaluate the quality using the received video in order to consider quality degradation due to video coding, the network, and the receiving terminal.

  Among video objective quality evaluation techniques using video signals, there are techniques for evaluating the user experience quality based on information on the encoding block size and the encoding block position (for example, Non-Patent Documents 1, 2, and 3).

  However, the techniques disclosed in Non-Patent Documents 1, 2, and 3 are adapted to determine the encoding block size and encoding block position in advance when a spatial shift of the video signal occurs due to video media format conversion or packet loss degradation. Difficult to grasp. In such a case, a technique for estimating the encoding block size and the encoding block position is required.

  As described above, video quality is degraded by video coding. For this reason, a block noise removal filter that reduces block noise generated in a video signal that has been encoded and improves video quality has been developed.

  The block noise removal technique is a technique for reducing block noise by estimating and subtracting block noise from the pixel space or frequency space of the pixel signal. If the filter strength is increased too much, not only block noise but also information on the definition of the original video will be reduced, and the video media will be deteriorated. Therefore, a filter strength optimization technique is also important.

  Some optimization of the block noise removal filter and the filter strength is based on information on the encoding block size and the encoding block position (for example, Non-Patent Documents 4 and 5).

  However, in these techniques, it is difficult to grasp the coding block size and the coding block position in advance when a video signal spatial shift occurs due to video media format conversion or packet loss degradation. In such a case, a technique for estimating the encoding block size and the encoding block position is required.

  As described above, in the video objective quality evaluation technique and the block noise removal filter, there is a problem that the quality estimation accuracy is deteriorated because the coding block size and the coding block position are not properly known, and the block noise cannot be removed. is there.

  In order to solve these problems, the following solutions have been proposed. Non-Patent Document 6 describes a method of applying an edge extraction filter to a video signal and estimating a coding block position using a cross-correlation between the generated edge video signal and a grid image as shown in FIG. ing. Non-Patent Document 7 describes a method for estimating a coding block size and a coding block position using a block noise amount extraction filter.

Kono, Yamagishi, Okamoto, "Proposal of NR-type media layer model for IPTV service," IEICE Technical Report, CQ2009-76, Jan. 2010. O. Sugimoto1, S. Naito, S. Sakazawa, and A. Koike, "Objective perceptual video quality measurement method based on hybrid no reference framework," IEEE International Conference on Image Processing (ICIP), pp.2237-2240, Nov. 2009. S. Qiu, H. Rui, and L. Zhang, "No-reference perceptual quality assessment for streaming video based on simple end-to-end network measures, proceedings of the international conference on networking and services (ICNS) 2006. S. Liu and A. C. Bovik, "Efficient DCT-domain blind measurement and reduction of blocking artifacts," IEEE Transactions on Circuits and Systems for Video Technology, vol. 12, no. 12, pp. 1139-1149, 2002. Sugiyama, Sagara, Yokoyama, "Distortion amount objective evaluation method for coding block distortion reduction," IEICE Transactions A, Vol.J91-A, No.6, pp.643-646, 2008. R. Dosselmann, X. D. Yang, "A Prototype No-Reference Video Quality System," Proceedings of the Fourth Canadian Conference on Computer and Robot Vision, p.411-417, May 2007 Okawa, Sagara, Sugiyama, "Detection of coding block position from format-converted images," ITEJ Technical Report, vol.33, No.10, pp.33-36, Feb. 2010.

  However, Non-Patent Document 6 is a coding block position estimation technique that assumes only a spatial shift of a video signal, and it is necessary to grasp the coding block size in advance. Therefore, when the video signal is enlarged or reduced by format conversion (for example, conversion from 1440x1080 to 1920x1080) etc., the encoding block size is also enlarged or reduced, so that the encoding block size cannot be known. is there.

  Non-Patent Document 7 is a technique in which a plurality of coding block sizes and coding block positions are designated in advance, and the coding block size and coding block position are estimated from them. However, as shown in FIG. 2, in addition to the problem that the types of video formats are very large, when the video format is converted, the encoding block size and the encoding block position cannot be designated in advance. There is a problem.

  Even if the information about the encoding block size and the encoding block position is unknown due to a spatial shift of the video signal due to, for example, format conversion of the video signal or degradation of packet loss, the present invention The purpose is to estimate the size and coding block position.

In order to solve the above problems, the coding block size and position estimation method of the present invention are:
From the video signal, a coding block size indicating the size in the horizontal direction and the vertical direction of the coding block, which is a processing unit of discrete cosine transform of video coding, and a coding block position indicating the position of the boundary line of the coding block; A coding block size and position estimation method in an apparatus for estimating
An edge extraction step in which an edge extraction unit applies an edge extraction filter to each frame of the video signal to extract an edge video signal from the video signal;
A frequency extraction unit applies a Fourier transform to each frame of the edge video signal, and arranges a one-dimensional array of amplitude values corresponding to the array number, where the horizontal frequency of the obtained edge video signal is an array number as a one-dimensional array, it averaged in every frame for each of the elements, to extract frequency information of the horizontal direction showing the amplitude value corresponding to the horizontal frequency of the edge image signal, Fourier for each frame of the edge image signal Applying the transformation, a one-dimensional array of amplitude values corresponding to the array number with the vertical frequency of the obtained edge video signal as the array number is averaged over all frames for each element of the array, A frequency extraction step of extracting vertical frequency information indicating an amplitude value corresponding to a vertical frequency of the edge video signal;
A horizontal encoding block size extraction unit, wherein the horizontal encoding block size extraction unit extracts a horizontal encoding block size based on a comparison of horizontal frequency amplitude values ;
Is vertical coding block size extracting unit, and the vertical direction coding block size extracting a vertical encoding block size based on a comparison of the amplitude value of the frequency in the vertical direction,
An encoding block position extraction unit that extracts an encoding block position based on the encoding block size in the horizontal direction, the encoding block size in the vertical direction, and the edge video signal;
It is characterized by having.

The program of the present invention is
From the video signal, a coding block size indicating the size in the horizontal direction and the vertical direction of the coding block, which is a processing unit of discrete cosine transform of video coding, and a coding block position indicating the position of the boundary line of the coding block; A device for estimating
Applying an edge extraction filter for each frame of the video signal, and extracting an edge video signal from the video signal;
A Fourier transform is applied to each frame of the edge video signal, and a one-dimensional array of amplitude values corresponding to the array number with the horizontal frequency of the obtained edge video signal as the array number is assigned to each element of the array. as a one-dimensional array by averaging the frames of the extracted frequency information in the horizontal direction showing the amplitude value corresponding to the horizontal frequency of the edge image signal, applying a Fourier transform to each frame of the edge image signal, A one-dimensional array of amplitude values corresponding to the array number, where the vertical frequency of the obtained edge video signal is an array number, is set as a one-dimensional array that is averaged over all frames for each element of the array. A frequency extracting means for extracting vertical frequency information indicating an amplitude value corresponding to a vertical frequency;
Horizontal coding block size extracting means for extracting a horizontal encoding block size based on a comparison of the amplitude values of the horizontal frequency,
Vertical-direction encoded block size extraction means for extracting a vertical-direction encoded block size based on comparison of amplitude values of vertical frequency, and the horizontal-direction encoded block size and the vertical-direction encoded block size; Coding block position extracting means for extracting a coding block position based on the edge video signal;
It is made to function as.

  According to the present invention, it is possible to estimate the encoding block size and the encoding block position even when the information regarding the encoding block size and the encoding block position is unknown.

Grid image example Example of video format The figure which shows horizontal direction frequency information and vertical direction frequency information Diagram showing two-dimensional frequency information The figure which shows the relationship between a reference pixel position and an encoding block boundary line Block diagram of coding block size and position estimation device Encoding block size and position estimation method flowchart

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

  In the embodiment of the present invention, the coding block size and the coding block position are estimated based on edge distortion (block noise) appearing at the coding block position by video coding. More specifically, an edge generated by block noise has a constant frequency (encoded block frequency) in the horizontal direction and the vertical direction. Therefore, the amplitude value of the coding block frequency becomes higher than the amplitude value of the neighboring frequency. In the embodiment of the present invention, the encoding block size and the encoding block position are estimated based on the amplitude value of the encoding block frequency. The encoding block size and position estimation method will be described below.

  First, an edge extraction filter (for example, Canny edge detector, Sobel filter, Laplacian filter, Prewitt filter, etc.) is applied to each frame of the input video signal to generate an edge video signal.

  In order to extract the size of a coding block from an edge video signal including edges generated by encoding, horizontal frequency information (horizontal frequency information) and vertical frequency information (vertical frequency information) of the edge video signal are extracted. Get. As shown in FIG. 3, the horizontal direction frequency information and the vertical direction frequency information can be represented by a one-dimensional array of amplitude values corresponding to the array numbers, with the array numbers in the horizontal direction and the vertical direction as frequencies. Hereinafter, a method for calculating two types of horizontal frequency information and vertical frequency information will be described.

  In the first calculation method, a two-dimensional frequency extraction filter such as Fourier transform is applied to each frame of the edge video signal to extract horizontal and vertical frequency information (two-dimensional frequency information). As shown in FIG. 4, the two-dimensional frequency information is a two-dimensional array of amplitude values corresponding to the array numbers with the array numbers in the horizontal direction and the vertical direction as frequencies. The one-dimensional array obtained by averaging the elements of the one-dimensional array whose array number in the horizontal direction of the two-dimensional frequency information is 0 is used as the horizontal frequency information. The one-dimensional array obtained by averaging the elements of the one-dimensional array whose vertical array number is 0 in the two-dimensional frequency information is used as the vertical frequency information.

  In the second calculation method, horizontal frequency information is extracted by applying a one-dimensional frequency extraction filter such as Fourier transform to each horizontal pixel column of each frame of the edge video signal. This frequency information is a one-dimensional array of amplitude values corresponding to the array number with the array number in the horizontal direction as a frequency. A one-dimensional array in which each element of the extracted frequency information is averaged in all horizontal pixel columns of all frames is set as horizontal frequency information. In addition, a one-dimensional frequency extraction filter such as Fourier transform is applied to each vertical pixel column of each frame of the edge video signal to extract vertical frequency information. This frequency information is a one-dimensional array of amplitude values corresponding to the array number, with the array number in the vertical direction as a frequency. A one-dimensional array obtained by averaging the elements of the extracted frequency information in all the vertical pixel columns of all the frames is defined as the vertical frequency information.

  Since block noise is generated at the boundary line of the encoded block, an edge generated by the block noise has a constant frequency (encoded block frequency) in the horizontal direction and the vertical direction. Therefore, the amplitude value of the coding block frequency becomes higher than the amplitude value of the neighboring frequency. A value obtained by subtracting an average value or an intermediate value of the amplitude values of neighboring SEQ ID NOs from the amplitude value of a certain SEQ ID NO is defined as the peak intensity of the SEQ ID NO. It is defined as a peak frequency, and the peak frequency of the horizontal frequency information is defined as a horizontal encoding block frequency. Although the horizontal encoding block frequency may be extracted within the range of all array numbers, generally, the amplitude value of the low frequency component of the video signal becomes extremely high regardless of the presence or absence of occurrence of block noise. In order not to accidentally extract the peak frequency due to the nature of the original video signal as the horizontal encoding block frequency, set a threshold of the array number, and set the horizontal encoding block frequency from the range of the array number higher than the threshold. It may be extracted. That is, the horizontal encoding block frequency may be extracted within a predetermined array number range. The predetermined sequence number may be a range of all sequence numbers or a range of sequence numbers designated in advance.

  The array element number extracted by the same method from the vertical direction frequency information is set as the vertical direction encoding block frequency. The value obtained by dividing the total number of pixels in the horizontal direction by the horizontal direction coding block frequency is the horizontal direction coding block size, which is the horizontal size of the coding block, and the total number of pixels in the vertical direction is represented by the vertical direction coding block frequency. The divided value becomes the vertical encoding block size which is the vertical size of the encoding block.

  Next, as shown in FIG. 5, when a certain pixel position in the frame of the edge video signal is set as the reference pixel position, a group of vertical lines separated from the reference pixel position in the horizontal direction by a multiple of the horizontal encoded block size. A group of horizontal lines separated from the reference pixel position in the vertical direction by a pixel that is a multiple of the vertical direction encoded block size is defined as an encoded block boundary line. A block noise amount extraction filter (for example, Non-Patent Documents 1, 2, and 3) is applied to the edge video signal within a predetermined range in the frame, and the obtained block noise amount is the maximum at the reference pixel position. The position of the encoding block boundary line is set as the encoding block position. The predetermined range in the frame may be the entire range in the frame or a specified range in the frame.

<Embodiment: Description of Coding Block Size and Position Estimation Device>
Next, a coding block size and position estimation device that realizes an embodiment of the present invention will be described. FIG. 5 is a block diagram of an encoding block size and position estimation apparatus 10 that is an example of an embodiment of the present invention. The coding block size and position estimation apparatus 10 includes an edge extraction unit 100, a coding block size extraction unit 110, and a coding block position extraction unit 120. The encoding block size extraction unit 110 includes a frequency extraction unit 111, a horizontal direction encoding block size extraction unit 112, and a vertical direction encoding block size extraction unit 113.

  The edge extraction unit 100 captures an output received video signal from a display (or a set top or the like) or reads a video file. In the video signal, there are pixel values representing luminance signals and color difference signals. An edge extraction filter is applied to the luminance signal in the captured video signal to generate an edge video signal, which is input to the frequency extraction unit 111 and the encoded block position extraction unit 120. In this embodiment, the Canny Edge Detector is used as the edge extraction filter, but a Sobel filter, a Laplacian filter, a Prewitt filter, or the like may be used. In this embodiment, an edge video signal is extracted using a luminance signal, but a color difference signal or both a luminance signal and a color difference signal may be used.

  The frequency extraction unit 111 applies a frequency extraction filter to the input edge video signal to extract horizontal frequency information and vertical frequency information, and the horizontal encoding block size extraction unit 112 and the vertical encoding block, respectively. Input to the size extraction unit 113. In the present embodiment, discrete Fourier transform is used as the frequency extraction filter. As shown in FIG. 3, the horizontal frequency information and the vertical frequency information are one-dimensional arrays of amplitude values corresponding to the array numbers with the array numbers in the horizontal and vertical directions as frequencies. Hereinafter, a method for calculating two types of horizontal frequency information and vertical frequency information will be described.

  In the first calculation method, a two-dimensional discrete Fourier transform is applied to each frame of the edge video signal as follows.

ここでF(u,v,f)は、図４に示すように、エッジ映像信号のf番目のフレームの水平方向の周波数がu、垂直方向の周波数がvのときの振幅値、Hは垂直方向の総画素数、Wは水平方向の総画素数、y(w,h,f)は、エッジ映像信号のf番目のフレームの画素位置(w,h)の輝度値、exp()は底がネイピア数eである指数関数を示す。F(u,v,f)は複素数を含むため、次式のように絶対値を算出し、実数に変換する。

Here, F (u, v, f) is an amplitude value when the horizontal frequency of the f-th frame of the edge video signal is u and the vertical frequency is v, as shown in FIG. 4, and H is vertical. The total number of pixels in the direction, W is the total number of pixels in the horizontal direction, y (w, h, f) is the luminance value of the pixel position (w, h) of the f-th frame of the edge video signal, and exp () is the bottom Indicates an exponential function with Napier number e. Since F (u, v, f) includes a complex number, an absolute value is calculated and converted to a real number as in the following equation.

abs()は、絶対値を出力する関数を示す。F'(u,v,f)の垂直方向の周波数vが0である１次元配列を次式のように、全てのフレームにおいて平均した水平方向符号化ブロック周波数Fw(u)を算出する。

abs () indicates a function that outputs an absolute value. The horizontal coding block frequency Fw (u) obtained by averaging the one-dimensional array in which the frequency v in the vertical direction of F ′ (u, v, f) is 0 as shown in the following equation is calculated.

ここでNはフレームの総数を示す。同様に、F'(u,v,f)の水平方向の周波数uが0である１次元配列を次式のように、全てのフレームにおいて平均した垂直方向符号化ブロック周波数Fh(v)を算出する。

Here, N indicates the total number of frames. Similarly, the vertical coding block frequency Fh (v) is calculated by averaging the one-dimensional array in which the horizontal frequency u of F ′ (u, v, f) is 0 as shown in the following equation. To do.

第２の算出方法は、エッジ映像信号の各フレームの各水平画素列に対し、１次元の離散フーリエ変換を次式のように適用する。

In the second calculation method, a one-dimensional discrete Fourier transform is applied to each horizontal pixel column of each frame of the edge video signal as follows.

ここでFwtmp(u,h,f)は、エッジ映像信号のf番目のフレームのh行目の水平方向の周波数がuのときの振幅値を示す。F(u,v,f)は複素数を含むため、次式のように絶対値を算出し、実数に変換する。

Here, Fwtmp (u, h, f) indicates an amplitude value when the horizontal frequency in the h-th row of the f-th frame of the edge video signal is u. Since F (u, v, f) includes a complex number, an absolute value is calculated and converted to a real number as in the following equation.

Fwtmp'(u,h,f)を次式のように、全てのフレームの全ての水平画素列において平均した水平方向符号化ブロック周波数Fw(u)を算出する。

A horizontal encoding block frequency Fw (u) obtained by averaging Fwtmp ′ (u, h, f) in all horizontal pixel columns of all frames is calculated as follows.

同様に、垂直方向に対しても次式のように垂直方向符号化ブロック周波数Fw(u)を算出する。

Similarly, also in the vertical direction, the vertical direction encoded block frequency Fw (u) is calculated as in the following equation.

ここでFhtmp(w,v,f)は、エッジ映像信号のf番目のフレームのw列目の垂直方向の周波数がvのときの振幅値を示す。

Here, Fhtmp (w, v, f) represents an amplitude value when the vertical frequency of the w-th column of the f-th frame of the edge video signal is v.

  The horizontal encoded block size extracting unit 112 extracts the horizontal encoded block size from the input horizontal frequency information and inputs the extracted horizontal encoded block size to the encoded block position extracting unit 120. A method of extracting the horizontal encoding block size will be described in detail.

  From the horizontal frequency information Fw (u), Dw (u) is calculated as in the following equation.

或いは、次式のようにDw(u)を算出してもよい。

Alternatively, Dw (u) may be calculated as in the following equation.

ここでm（m≧1）は定数であり、予め設定される。本実施形態では、mは5とする。また、median()は、入力された集合の中間値を出力する関数を示す。このとき、uの範囲は、mからW-m-1になるが、一般に低周波数成分の振幅値が極端に高くなっており、適切にブロックノイズによる振幅のピークを抽出できない場合がある。そのため、予め閾値Tを設定し、uの範囲をTからW-m-1にしてもよい。Dw(u)が最大になるuを水平符号化ブロック周波数umaxとする。水平符号化ブロックサイズCwは水平符号化ブロック周波数umaxから次式のように算出する。

Here, m (m ≧ 1) is a constant and is set in advance. In this embodiment, m is 5. Median () indicates a function that outputs an intermediate value of the input set. At this time, the range of u is from m to Wm-1, but generally the amplitude value of the low frequency component is extremely high, and there is a case where the peak of amplitude due to block noise cannot be extracted appropriately. Therefore, the threshold value T may be set in advance, and the range of u may be changed from T to Wm-1. Let u be the horizontal coding block frequency umax where Dw (u) is maximized. The horizontal coding block size Cw is calculated from the horizontal coding block frequency umax as follows.

Dw(umax)の値が極端に小さい場合、ブロックノイズが発生しておらず、符号化ブロックの周波数を適切に抽出できていない可能性がある。そのため、閾値を設定し、Dw(umax)がその閾値以下のときは、ブロックノイズが発生していないとし、符号化ブロックサイズ及び符号化ブロック位置を出力せずに、ブロックノイズが発生していないという情報を出力する。本実施形態は、映像品質の定量化や改善のための技術であり、映像品質に影響を与える主要因であるブロックノイズを定量化又は低減させるための情報として。符号化ブロックサイズ及び符号化ブロック位置を推定している。従って、映像品質の定量化においては、ブロックノイズが発生していない場合は、ブロックノイズ量を下限値に設定するなどの処理を行えばよい。また、ブロックノイズ除去フィルタにおいても、ブロックノイズが発生していない映像に対しては、ブロックノイズを除去する必要がないため、フィルタを適用しないなどの処理を行えばよい。

When the value of Dw (umax) is extremely small, there is a possibility that block noise is not generated and the frequency of the encoded block cannot be extracted properly. Therefore, when a threshold value is set and Dw (umax) is less than or equal to the threshold value, block noise is not generated, and no block noise is generated without outputting the encoded block size and encoded block position. Is output. This embodiment is a technique for quantifying and improving video quality, and as information for quantifying or reducing block noise, which is a main factor affecting video quality. The encoding block size and the encoding block position are estimated. Therefore, in the quantification of video quality, when block noise is not generated, processing such as setting the block noise amount to a lower limit value may be performed. Also in the block noise removal filter, since it is not necessary to remove block noise for an image in which block noise is not generated, processing such as not applying the filter may be performed.

  The vertical direction encoded block size extracting unit 113 extracts the vertical direction encoded block size from the input vertical direction frequency information, and inputs the extracted vertical direction encoded block size to the encoded block position extracting unit 120. A method for extracting the vertical encoded block size will be described in detail.

  From the vertical frequency information Fh (v), Dh (v) is calculated as in the following equation.

或いは、次式のようにDh(v)を算出してもよい。

Alternatively, Dh (v) may be calculated as in the following equation.

このとき、vの範囲は、mからH-m-1になるが、一般に低周波数成分の振幅値が極端に高くなっており、適切にブロックノイズによる振幅のピークを抽出できない場合がある。そのため、予め閾値Tを設定し、vの範囲をTからH-m-1にしてもよい。Dh(v)が最大になるvを垂直符号化ブロック周波数vmaxとする。垂直符号化ブロックサイズChは垂直符号化ブロック周波数vmaxから次式のように算出する。

At this time, the range of v is from m to Hm-1, but generally the amplitude value of the low frequency component is extremely high, and there is a case where the peak of amplitude due to block noise cannot be extracted appropriately. Therefore, the threshold value T may be set in advance, and the range of v may be changed from T to Hm-1. Let v at which Dh (v) is maximized be the vertical coding block frequency vmax. The vertical coding block size Ch is calculated from the vertical coding block frequency vmax as follows.

Dh(vmax)の値が極端に小さい場合、ブロックノイズが発生しておらず、符号化ブロックの周波数を適切に抽出できていない可能性がある。そのため、閾値を設定し、Dh(vmax)がその閾値以下のときは、ブロックノイズが発生していないとし、符号化ブロックサイズ及び符号化ブロック位置を出力せずに、ブロックノイズが発生していないという情報を出力する。本実施形態は、映像品質の定量化や改善のための技術であり、映像品質に影響を与える主要因であるブロックノイズを定量化又は低減させるための情報として。符号化ブロックサイズ及び符号化ブロック位置を推定している。従って、映像品質の定量化においては、ブロックノイズが発生していない場合は、ブロックノイズ量を下限値に設定するなどの処理を行えばよい。また、ブロックノイズ除去フィルタにおいても、ブロックノイズが発生していない映像に対しては、ブロックノイズを除去する必要がないため、フィルタを適用しないなどの処理を行えばよい。

When the value of Dh (vmax) is extremely small, there is a possibility that block noise is not generated and the frequency of the coding block cannot be extracted appropriately. Therefore, when a threshold is set and Dh (vmax) is less than or equal to the threshold, no block noise is generated, and no block noise is generated without outputting the encoded block size and encoded block position. Is output. This embodiment is a technique for quantifying and improving video quality, and as information for quantifying or reducing block noise, which is a main factor affecting video quality. The encoding block size and the encoding block position are estimated. Therefore, in the quantification of video quality, when block noise is not generated, processing such as setting the block noise amount to a lower limit value may be performed. Also in the block noise removal filter, since it is not necessary to remove block noise for an image in which block noise is not generated, processing such as not applying the filter may be performed.

  The encoded block position extraction unit 120 extracts an encoded block position from the input edge video signal, the horizontal direction encoded block size, and the vertical direction encoded block size. A method for extracting a coding block position will be described in detail.

  First, an arbitrary pixel in each frame of the edge video signal is temporarily set as a reference pixel position. As shown in FIG. 5, a horizontal line group vertically separated from the reference pixel position by a multiple of the horizontal encoded block size Ch and a vertical line separated from the reference pixel position by a multiple of the vertical encoded block size Cw in the horizontal direction. Let the group be the encoding block boundary. When the reference pixel position is moved in parallel in the horizontal direction and the vertical direction, the reference pixel position at which the block noise amount obtained by applying the block noise extraction filter to the edge video signal is maximized is set as the encoding block position. In this embodiment, the block noise extraction filter described in Non-Patent Document 1 is used. As the block noise extraction filter, those described in Non-Patent Document 2 and Non-Patent Document 3 may be used.

<Embodiment: Description of Coding Block Size and Position Estimation Method>
Next, a coding block size and position estimation method according to the embodiment of the present invention will be described. FIG. 6 is a flowchart of a coding block size and position estimation method according to an embodiment of the present invention.

  In step S10, the edge extraction unit 100 captures the video signal, extracts the edge video signal from the captured video signal, and inputs the edge video signal to the frequency extraction unit 111 and the encoded block position extraction unit 120.

  In step S11, the frequency extraction unit 111 extracts the horizontal direction frequency information and the vertical direction frequency information from the input edge video signal, and the horizontal direction encoding block size extraction unit 112 and the vertical direction encoding block size extraction unit 113, respectively. To enter.

  In step S <b> 12, the horizontal direction encoded block size extracting unit 112 extracts the horizontal direction encoded block size from the input horizontal direction frequency information and inputs the extracted horizontal direction encoded block size to the encoded block position extracting unit 120.

  In step S <b> 13, the vertical coding block size extraction unit 113 extracts the vertical coding block size from the input vertical frequency information, and inputs the vertical coding block size to the coding block position extraction unit 120.

  In step S14, the coding block position extraction unit 120 extracts a coding block position from the input edge video signal, horizontal frequency information, and vertical frequency information.

  Therefore, as the final output, from the horizontal direction encoding block size extraction unit 112, the horizontal direction encoding block size, from the vertical direction encoding block size extraction unit 113, from the vertical direction encoding block size, the encoding block position extraction unit From 120, the encoded block position can be obtained. In addition to the above three pieces of information, the final output may be converted from the horizontal encoding block size, the vertical block size, and the encoding block position to all pixel position information on the boundary line of the encoding block. Good.

<Effect of embodiment>
As described above, according to the embodiment of the present invention, even when the coding block size is unknown due to a spatial shift of the video signal due to the format conversion of the video signal or packet loss degradation, the coding block size and It becomes possible to grasp the encoding block position.

  As a result, the user experience quality can be appropriately estimated even for a video in which a spatial shift of the video signal occurs due to the format conversion of the video signal or the packet loss degradation. Therefore, it is possible to provide high-quality services and improve the market value of services by implementing quality design and quality management in accordance with user experience quality for services using media signals such as IPTV services.

  As described above, the present invention can be applied to a video in which a spatial shift of the video signal due to video signal format conversion or packet loss degradation occurs, so the coding size and the coding block position are considered. Thus, a block noise removal filter can be applied, and block noise of the video signal can be reduced.

  Therefore, by displaying the media signal output to the monitor with high quality, the user satisfaction of the user can be increased, and the market value of the service and the product is improved.

  For convenience of explanation, the coding block size and position estimation apparatus according to the embodiment of the present invention is described using a functional block diagram, but the coding block size and position estimation apparatus of the present invention is hardware, It may be realized by software or a combination thereof. For example, each functional unit of the encoding block size and position estimation device may be realized by software and may be realized in a computer. Moreover, each component of the embodiment may be used in combination as necessary.

  Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments, and various modifications and applications are possible within the scope of the claims.

  The present invention can be applied to a video quality objective evaluation technique for estimating a user experience quality from a physical quantity obtained from a video signal. Further, the present invention can be applied to a block noise removal filter technique that reduces block noise generated by video coding.

DESCRIPTION OF SYMBOLS 10 Encoding block size and position estimation apparatus 100 Edge extraction part 110 Encoding block size extraction part 111 Frequency extraction part 112 Horizontal direction encoding block size extraction part 113 Vertical direction encoding block size extraction part 120 Encoding block position extraction part

Claims (5)

From the video signal, a coding block size indicating the size in the horizontal direction and the vertical direction of the coding block, which is a processing unit of discrete cosine transform of video coding, and a coding block position indicating the position of the boundary line of the coding block; A coding block size and position estimation method in an apparatus for estimating
An edge extraction step in which an edge extraction unit applies an edge extraction filter to each frame of the video signal to extract an edge video signal from the video signal;
A frequency extraction unit applies a Fourier transform to each frame of the edge video signal, and arranges a one-dimensional array of amplitude values corresponding to the array number, where the horizontal frequency of the obtained edge video signal is an array number as a one-dimensional array, it averaged in every frame for each of the elements, to extract frequency information of the horizontal direction showing the amplitude value corresponding to the horizontal frequency of the edge image signal, Fourier for each frame of the edge image signal Applying the transformation, a one-dimensional array of amplitude values corresponding to the array number with the vertical frequency of the obtained edge video signal as the array number is averaged over all frames for each element of the array, A frequency extraction step of extracting vertical frequency information indicating an amplitude value corresponding to a vertical frequency of the edge video signal;
A horizontal encoding block size extraction unit, wherein the horizontal encoding block size extraction unit extracts a horizontal encoding block size based on a comparison of horizontal frequency amplitude values ;
Is vertical coding block size extracting unit, and the vertical direction coding block size extracting a vertical encoding block size based on a comparison of the amplitude value of the frequency in the vertical direction,
An encoding block position extraction unit that extracts an encoding block position based on the encoding block size in the horizontal direction, the encoding block size in the vertical direction, and the edge video signal;
A coding block size and position estimation method comprising: The horizontal encoding block size extraction step includes:
A predetermined array of frequency information in the horizontal direction when a value obtained by subtracting an average value or an intermediate value of amplitude values of adjacent array element numbers from an amplitude value of an array element number is the peak intensity of the array element number Within the range of numbers, extract the peak frequency indicating the array number that maximizes the peak intensity, and divide the total number of pixels in the horizontal direction by the peak frequency to extract the encoded block size in the horizontal direction. The encoding block size and position estimation method according to claim 1 . The vertical encoding block size extraction step includes:
A predetermined array of frequency information in the vertical direction when a value obtained by subtracting an average value or an intermediate value of amplitude values of adjacent array element numbers from an amplitude value of an array element number is the peak intensity of the array element number Within the range of numbers, extract the peak frequency indicating the array number that maximizes the peak intensity, and divide the total number of pixels in the vertical direction by the peak frequency to extract the encoded block size in the vertical direction The encoding block size and position estimation method according to claim 1 or 2 . The encoded block position extracting step includes:
A vertical line group having a pixel position in a frame of the edge video signal as a reference pixel position, and a number of pixels that is a multiple of the encoded block size in the horizontal direction in the horizontal direction from the reference pixel position, and the reference pixel position When a horizontal line group that is a number of pixels that is a multiple of the encoded block size in the vertical direction from the vertical direction is defined as an encoded block boundary line, the amount of block noise in the edge video signal is within a predetermined range within the frame. by block noise amount obtained by applying the block noise amount extraction filter for extracting extracts the reference pixel position having the largest, it extracts the coded block position, any one of claims 1 to 3 The encoding block size and position estimation method described in 1. From the video signal, a coding block size indicating the size in the horizontal direction and the vertical direction of the coding block, which is a processing unit of discrete cosine transform of video coding, and a coding block position indicating the position of the boundary line of the coding block; A device for estimating
Applying an edge extraction filter for each frame of the video signal, and extracting an edge video signal from the video signal;
A Fourier transform is applied to each frame of the edge video signal, and a one-dimensional array of amplitude values corresponding to the array number with the horizontal frequency of the obtained edge video signal as the array number is assigned to each element of the array. as a one-dimensional array by averaging the frames of the extracted frequency information in the horizontal direction showing the amplitude value corresponding to the horizontal frequency of the edge image signal, applying a Fourier transform to each frame of the edge image signal, A one-dimensional array of amplitude values corresponding to the array number, where the vertical frequency of the obtained edge video signal is an array number, is set as a one-dimensional array that is averaged over all frames for each element of the array. A frequency extracting means for extracting vertical frequency information indicating an amplitude value corresponding to a vertical frequency;
Horizontal coding block size extracting means for extracting a horizontal encoding block size based on a comparison of the amplitude values of the horizontal frequency,
Vertical-direction encoded block size extraction means for extracting a vertical-direction encoded block size based on comparison of amplitude values of vertical frequency, and the horizontal-direction encoded block size and the vertical-direction encoded block size; Coding block position extracting means for extracting a coding block position based on the edge video signal;
Program to function as.

Images