JP2007129628A - Encoding error measurement device and encoding error measurement program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an encoding error measurement device which is capable of accurately and easily measuring a deterioration in picture quality due to the compression encoding without using a subject copy even if the compression encoding is carried out together with motion-compensated prediction, and to provide an encoding measurement program. <P>SOLUTION: The encoding error measurement device that measures the encoding error of the compression-coded signals is equipped with a first means of subjecting the motion-reference images decoded from the compression-encoded signals to orthogonal conversion to form frequency distribution and a second means of measuring a difference between the average value of the variance of the frequency distribution of the images as measuring objects and the average value of the variance of the frequency distribution of the motion-reference images and considering the above difference as an encoding error that is lost and not measured in the images as the measuring objects. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a coding error measuring apparatus and a coding error measuring program, and more particularly to a coding error measuring apparatus and a coding error measuring program for measuring a coding error of a compressed coded signal obtained by compression coding a video signal.

  For example, the transmission image deteriorates in image quality due to compression encoding. Therefore, in monitoring facilities such as broadcasting stations and network operators, automatic monitoring technology is used to automatically monitor how much the quality of transmitted images has deteriorated due to compression coding. Some automatic monitoring techniques are based on a comparison between an original image (material image) before compression coding and a received image (decoded image) after compression coding, and some are based only on a received image after compression coding.

  In the automatic monitoring technology based on the comparison between the original image before compression encoding and the received image after compression encoding, the phase of the original image and the received image are matched, and PSNR (Peak Signal-to-To A value called Noise Ratio has been used.

  In addition, the automatic monitoring technology based only on the received image after compression encoding does not require an original image before compression encoding, and there is a method for evaluating deterioration specific to encoding included in the received image after decompression decoding by image analysis. Has been used.

  Patent Documents 1 and 2 describe an example of an automatic monitoring technique based only on a received image after compression encoding. In compression encoding generally used such as JPEG and MPEG, orthogonal transform is performed for each small block, and the amount of information is reduced based on the small block. In such compression encoding, since the received image has light and dark (shading) unevenness, the quality of the received image is measured by enhancing and detecting the boundary line generated at the boundary of the small block.

  However, since the automatic monitoring technology based on the comparison between the original image before compression encoding and the received image after compression encoding requires the original image before compression encoding, how much the image quality of the received image has deteriorated. There was a problem that it was not possible to monitor a video distribution destination that did not have an original picture.

  In addition, since a time delay is generated from when the original image is input by the compression / decompression processing of the compression encoder until the received image is decoded, it is not easy to match the frame synchronization between the original image and the received image. There was a problem.

  In addition, in the automatic monitoring technology based only on the received image after compression coding, since the image analysis is performed on the received image, there is a problem that a scene or a portion including a pattern such as deterioration in the original image is detected as deterioration. It was. For this reason, there has been a need for a technique for quantifying deterioration in image quality due to compression coding at the receiving end of a video having no original image regardless of the pattern of the original image.

Therefore, the present applicant has invented a coding error estimation method and a coding error estimation device that can easily and accurately estimate deterioration in image quality due to compression coding without using an original image (Patent Documents 3 and 4). reference). The encoding error estimation method and the encoding error estimation apparatus disclosed in Patent Documents 3 and 4 estimate the encoding error by estimating the orthogonal transform coefficient of the original image from the orthogonal transform coefficient after encoding.
JP 2000-102041 A Special table 2003-501850 gazette JP 2004-80741 A JP 2005-184260 A

  Patent Document 4 described above is an improved method of Patent Document 3, and is intended to improve the measurement accuracy of coding errors. However, the compression coding technique that uses motion compensated prediction has a mode in which pixels are copied from the motion reference frame without coding when the compression rate is extremely increased, and information on orthogonal transform coefficients is not transmitted. Many such non-orthogonal transform coded blocks are produced.

  As a result, in the compression coding technique using motion compensation prediction, the information amount of the orthogonal transform coefficient required for measuring the coding error is reduced, and an error may occur in the measurement (prediction) of the error amount. It was. This phenomenon frequently occurs in a frame that is not used as a motion reference frame, and there is a problem that a measurement error increases.

  The present invention has been made in view of the above points, and even in compression encoding combined with motion compensation prediction, it is possible to easily and accurately measure image quality degradation due to compression encoding without using an original image. It is an object of the present invention to provide a coding error measurement device and a coding error measurement program that are possible.

  In order to solve the above-mentioned problems, the present invention is an encoding error measuring apparatus for measuring an encoding error of a compressed encoded signal obtained by compressing and encoding a video signal, and is based on the compressed encoded signal used in combination with motion compensation prediction. The decoded motion reference image is orthogonally transformed to create its frequency distribution, while the image to be measured that refers to the motion reference image decoded from the compression-encoded signal used in combination with motion compensated prediction is orthogonally transformed. The difference between the means for creating the frequency distribution, the average value of the variance of the frequency distribution of the image that is the measurement target, and the average value of the variance of the frequency distribution of the motion reference image is the measurement target. And means for measuring as a lost coding error that is not measured in the image.

  The present invention also provides a coding error measurement program for measuring a coding error of a compression coded signal obtained by compression coding a video signal, which is executed by a computer having at least a storage device and an arithmetic processing unit, The processing apparatus orthogonally transforms the motion reference image decoded from the compression-encoded signal used in combination with motion-compensated prediction and creates a frequency distribution thereof, and is decoded from the compressed encoded signal used in combination with motion-compensated prediction. Orthogonally transforming the image that is the measurement target referring to the motion reference image, creating a frequency distribution thereof, an average value of the variance of the frequency distribution of the image that is the measurement target, and the motion reference image A step of measuring a difference from a mean value of a variance of a frequency distribution as a lost coding error that is not measured in the measurement target image. Characterized in that it is a difference measurement program.

  According to the present invention, an encoding error measuring apparatus capable of easily and accurately measuring degradation in image quality due to compression encoding without using an original image, even in compression encoding combined with motion compensation prediction. An encoding error measurement program can be provided.

  Next, the best mode for carrying out the present invention will be described based on the following embodiments with reference to the drawings.

  First, in order to facilitate understanding of the present invention, the principle of the present invention will be described. The present invention relates to a coding error measuring apparatus and a coding error measuring program for measuring a coding error of a compression coded signal (hereinafter referred to as a bit stream) obtained by compression coding a video signal using a compression coding technique.

  The present invention uses a bitstream that is compression-encoded by a compression-encoding algorithm that uses motion-compensated prediction, orthogonal transform, and quantization. In the present invention, the coefficient frequency distribution of orthogonal transform coefficients obtained from the bitstream is measured, and the coefficient frequency distribution before compression coding is estimated from the statistics of the coefficient frequency distribution. In the present invention, the coding error is measured from the statistics of the frequency distribution of the orthogonal transform coefficients of the motion reference image used for motion compensation prediction and the statistics of the frequency distribution of the orthogonal transform coefficients of the image to be measured.

  That is, the present invention utilizes the property that the variance of the orthogonal transform coefficient distribution of the frame that refers to the motion reference frame and the variance of the orthogonal transform coefficient distribution of the motion reference frame substantially coincide. In the present invention, using the above-described properties, the encoding error energy of a signal that cannot be measured by non-encoding is estimated from the motion reference frame, and in addition to the encoding error energy measured by the conventional method, The encoding error energy of the frame to be measured is used.

  As described above, according to the present invention, even when compression coding is used in combination with motion compensated prediction, coding error energy of a signal that cannot be measured by non-coding is estimated by using a motion reference frame. Error measurement error can be improved.

  FIG. 1 is a configuration diagram of an embodiment of an image transmission system 1 including a coding error measuring apparatus according to the present invention. In the image transmission system 1 of FIG. 1, the transmission side encoding unit 10 transmits a compressed encoded signal (bit stream) obtained by compressing and encoding a transmission image to the reception side via the transmission network 60.

  The decoding unit 20 and the coding error measurement device 40 on the receiving side receive the bit stream from the transmission network 60. The decoding unit 20 supplies the received image obtained by decoding the bit stream to the monitor 30. The monitor 30 displays the received image decoded by the decoding unit 20. In addition, the encoding error measurement device 40 quantifies the degree to which the image quality of the transmission image has deteriorated by encoding using a bit stream.

  Here, MPEG-2 will be described as an example of a decoding procedure of hybrid encoding using both motion compensation prediction and orthogonal transform. FIG. 2 is a configuration diagram of an embodiment of the decoding unit. The decoding unit 20 in FIG. 2 includes an orthogonal transform coefficient extraction unit 21 including a variable length decoding unit 22 and an inverse quantization unit 23, an inverse orthogonal transform unit 24, an adder 25, a video memory 26, and a motion compensation prediction unit. 27 and a format conversion unit 28.

  The variable length decoding unit 22 decodes the variable length code of the received bitstream into a fixed length code and supplies the decoded code to the inverse quantization unit 23. The variable length decoding unit 22 supplies a prediction mode and a prediction vector to a motion compensation prediction unit 27 described later. The inverse quantization unit 23 inversely quantizes the supplied fixed-length code into a DCT coefficient value (orthogonal transform coefficient value) and supplies it to the inverse orthogonal transform unit 24.

  The inverse orthogonal transform unit 24 inversely transforms the supplied DCT coefficient value into a decoded image and supplies the decoded image to the adder 25. The adder 25 adds the decoded image supplied from the inverse orthogonal transform unit 24 and the motion compensated prediction signal supplied from the motion compensation prediction unit 27 and supplies the result to the video memory 26 and the format conversion unit 28 as a decoded image.

  The video memory 26 supplies the supplied decoded image to the motion compensation prediction unit 27 in units of frames. The motion compensation prediction unit 27 performs motion compensation prediction based on the prediction mode and the prediction vector supplied from the variable length decoding unit 22 and supplies a motion compensation prediction signal to the adder 25.

  The format conversion unit 28 is supplied with the decoded image from the adder 25. In this embodiment, since motion compensation prediction is used, the order of frames is rearranged when encoding is performed. Therefore, the format conversion unit 28 rearranges the supplied decoded image frames in a temporally correct order and outputs them to the monitor 30.

  FIG. 3 is a block diagram of an embodiment of an encoding error measuring apparatus. 3 includes an orthogonal transform coefficient extraction unit 41, a moment calculation unit 42, a coding error calculation unit 43, an inverse orthogonal transform unit 44, an adder 45, a motion compensation prediction unit 46, a video memory 47, The configuration includes an orthogonal transform unit 48, a moment calculation unit 49, a moment predicted value calculation unit 50, a differentiator 51, and switches 52 and 53. Note that the coding error measuring device 40 in FIG. 3 has a configuration obtained by extending a conventional coding error measuring device.

  The orthogonal transform coefficient extraction unit 41 decodes the variable length code of the received bitstream into a fixed length code, decodes the prediction mode and the prediction vector, decodes the orthogonal transform (DCT) coefficient by inverse quantization, and performs inverse orthogonal transform. Supplied to the unit 44. The orthogonal transform coefficient extraction unit 41 supplies the quantized value (fixed length code) and the DCT coefficient value extracted from the received bitstream to the moment calculation unit 42.

  The moment calculation unit 42 divides the supplied DCT coefficient value into intra-coded block information and non-intra-coded block information. The moment calculation unit 42 classifies the DCT coefficient values divided into the intra-coded block information and the non-intra coded block information for each quantized value. In the case of a simple configuration, the moment calculator 42 calculates a second-order product factor (second moment) of the frequency distribution of DCT coefficient values classified for each quantized value. The second moment (hereinafter simply referred to as moment) is equal to the variance of DCT coefficient values. The moment calculator 42 supplies the calculated moment to the encoding error calculator 43.

  On the other hand, the inverse orthogonal transform unit 44 inversely transforms the supplied DCT coefficient value into a decoded image and supplies it to the adder 45. The adder 45 adds the decoded image supplied from the inverse orthogonal transform unit 44 and the motion compensated prediction signal supplied from the motion compensation prediction unit 46, and supplies the result to the video memory 47 and the orthogonal transform unit 48 as a decoded image.

  The video memory 47 supplies the supplied decoded image to the motion compensation prediction unit 46 in units of frames. Further, the motion compensation prediction unit 46 performs motion compensation prediction based on the prediction mode and the prediction vector supplied from the orthogonal transform coefficient extraction unit 41, and supplies a motion compensation prediction signal to the adder 45.

  The orthogonal transform unit 48 orthogonally transforms the supplied decoded image into a DCT coefficient value, and supplies the DCT coefficient value to the moment calculation unit 49. The moment calculator 49 calculates the moment of the frequency distribution of the supplied DCT coefficient value for each frequency.

  In the case of a motion reference frame, the coding error measurement device 40 performs control to close the switch 52 and open the switch 53. The moment calculation unit 49 records the calculated moment in the moment predicted value calculation unit 50 and supplies the moment to the differentiator 51. The moment predicted value calculation unit 50 outputs the recorded moment to the differentiator 51. However, since the switch 53 is open, the difference between the moment supplied from the moment calculator 49 and the moment supplied from the moment predicted value calculator 50 is not supplied to the encoding error calculator 43.

  On the other hand, in the case of a frame that refers to a motion reference frame, the encoding error measuring device 40 performs control to open the switch 52 and close the switch 53. The moment calculator 49 supplies the moment to the subtractor 51 without recording the moment in the moment predicted value calculator 50 because the switch 52 is open.

  The moment predicted value calculation unit 50 outputs the recorded moment to the differentiator 51. Since the switch 53 is closed, the difference between the moment supplied from the moment calculator 49 and the moment supplied from the moment predicted value calculator 50 is supplied from the subtractor 51 to the encoding error calculator 43.

  In the case of MPEG-2, the opening / closing of the switch 52 and the switch 53 is controlled as shown in Table 1 below according to the I picture, P picture, and B picture.

なお、スイッチ５２及びスイッチ５３の開閉は、Ｉピクチャ，Ｐピクチャ及びＢピクチャに応じて、以下の表２のように制御してもよい。

The opening and closing of the switch 52 and the switch 53 may be controlled as shown in Table 2 below according to the I picture, P picture, and B picture.

ここで、スイッチ５２，５３の動作について、順を追って説明する。まず、初期状態としてスイッチ５２は閉，スイッチ５３は開である。ＭＰＥＧ−２符号化ストリームの場合、予測フレームとして最大２フレームを用いるため、モーメント予測値演算部５０は２フレーム分のモーメント情報を記録，演算することができる。

Here, the operation of the switches 52 and 53 will be described in order. First, as an initial state, the switch 52 is closed and the switch 53 is open. In the case of an MPEG-2 encoded stream, a maximum of two frames are used as predicted frames, so that the moment predicted value calculation unit 50 can record and calculate moment information for two frames.

  First, the moment obtained by calculating the I picture information by the moment calculating unit 49 is recorded in the moment predicted value calculating unit 50. The output of the moment calculator 49 is input to the subtractor 51, but does not affect the encoding error calculator 43 because the switch 53 is open.

  The P picture information to be processed next is calculated by the moment calculator 49 and recorded in the moment predicted value calculator 50. The moment predicted value calculation unit 50 calculates the moment of the delayed I picture and outputs it to the differentiator 51. The subtractor 51 outputs the difference between the moment output from the moment calculator 49 and the moment of the I picture recorded in the moment predicted value calculator 50 as a moment predicted value. Since the switch 53 is closed, the moment predicted value is output to the encoding error calculation unit 43.

  Next, when processing the B picture information, the switch 52 is opened. The moment output by the moment calculator 49 is output to the differentiator 51. The moment predicted value calculation unit 50 calculates a moment predicted value from the recorded I and P picture moments according to the following equation (1), and outputs the calculated moment to the differentiator 51. Since the switch 53 is closed, the moment predicted value is output to the encoding error calculation unit 43.

  When the P picture is processed after the P picture, the switch 52 is closed. The moment output by the moment calculator 49 is recorded in the moment predicted value calculator 50. When the moment predicted value calculation unit 50 records a new moment, it discards the oldest moment (in this case, the moment of the first recorded I picture) and records the moment output by the moment calculation unit 49.

  Then, the encoding error calculation unit 43 uses the moment supplied from the moment calculation unit 42, and the difference between the moment supplied from the moment calculation unit 49 and the moment supplied from the moment predicted value calculation unit 50, An error amount of coding error is calculated.

  As motion compensation prediction, there is a method of performing motion compensation prediction from a plurality of frames. In that case, the encoding error measuring apparatus 40 has a plurality of moment predicted value calculation units 50, weights the recorded moments according to the temporal distance from the measurement target frame, and supplies the weights to the subtractor 51.

FIG. 4 is a diagram illustrating the relationship between motion compensation prediction and picture type. In FIG. 4, a frame of P picture is a frame P _k , and a frame of B picture is a frame B _k . For example, in the case of an MPEG-2 B picture (bidirectional frame), the moments σ _k and σ _{k + 1 of the} frames P _k and P _{k + 1 that} are reference frames are recorded in the moment predicted value calculation unit 50.

When the moments of the frames B _l and B _{l + 1} are σ _l and σ _{l + 1} , the output of the moment predicted value calculation unit 50 is expressed by the following equation (1).

σ _pred = ασ _k + βσ _{k + 1} (1)
In this case, α and β are weighting coefficients. For example, in the case of the frame _B1 , the weighting coefficient α = 2/3 and the weighting coefficient β = 1/3. In the case of frame B _{1 + 1} , the weighting coefficient α = 1/3 and the weighting coefficient β = 2/3. The method of determining α and β is a simple weighting method according to the time distance. In the case of a P picture, the moment of the reference frame becomes a predicted value as it is.

In addition, there are the following advanced methods for determining α and β. The moment prediction value calculation unit 50 sets α and β according to the prediction mode decoded by the orthogonal transform coefficient extraction unit 41 and the ratio of the frames referred to by the prediction vector. The moment predicted value calculation unit 50 sets the ratio of the number of vectors referring to the frame P _k to the total number of predicted vectors as α, and the ratio of the number of vectors referring to the frame P _{k + 1} to the total number of predicted vectors as β.

For the measurement of the frame B _l, the difference Δσ supplied to the encoding error calculation unit 43 to become the following equation (2).

Δσ = σ _pred −σ _l (2)
Note that the moment supplied from the moment calculator 49 is an average of 64 DCT coefficient frequency distribution moments in the case of 8 × 8 DCT. In order to further improve the measurement accuracy of the coding error, it is desirable that the moment supplied from the moment calculator 49 is the average of the 63 DCT coefficient frequency distribution moments of the AC component excluding the DC component.

  In the coding error measuring apparatus 40 of FIG. 3, the orthogonal transform coefficient extracting unit 41 classifies DCT coefficient values by quantized values, the moment calculating unit 42 calculates each moment, and supplies the calculated moment to the coding error calculating unit 43. ing.

The encoding error calculation unit 43 calculates the encoding error for each DCT coefficient value for each quantized value, adds up the error amount MSE _intra of the intra-coded block, and the error amount MSE _{non_intra} of the non-intra-coded block. To do. The ratio of the number of intra-coded blocks, the number of non-intra blocks, and the number of non-coded blocks is α, β, and γ, and the coding error amount (PSNR) of a frame that refers to a motion reference frame by the following equation (3) Calculate

PSNR = ₁₀ log ₁₀ ((Sp−p × Sp−p) / (αMSE _intra + βMES _{non_intra} + γΔσ)) (3)
In the above equation (3), in the case of an 8-bit image, α + β + γ = 1, 0 ≦ α ≦ 1, 0 ≦ β ≦ 1, 0 ≦ γ ≦ 1, Sp−p = 255 (= 2 ⁸ ). .

  The encoding error measuring device 40 can also be realized by a computer system as shown in FIG. FIG. 5 is a configuration diagram of an example of a computer system that implements an encoding error measurement apparatus.

  The computer system of FIG. 5 includes an input device 101, an output device 102, a drive device 103, an auxiliary storage device 104, a memory device 105, an arithmetic processing device 106, and an interface device 107, which are mutually connected by a bus B. .

  The input device 101 includes a keyboard and a mouse, and is used to input various signals. The output device 102 includes a display device and is used to display various windows and data. The interface device 107 includes a modem, a LAN card, and the like, and is used for connecting to a predetermined network.

  The encoding error measurement program of the present invention is provided by distributing the recording medium 108 or downloading it from a network. The recording medium 108 on which the encoding error measurement program is recorded is a recording medium such as a CD-ROM, a flexible disk, a magneto-optical disk, etc. for recording information optically, electrically or magnetically, a ROM, a flash memory, etc. Various types of recording media such as a semiconductor memory that electrically records information can be used.

  When the recording medium 108 on which the encoding error measurement program is recorded is set in the drive device 103, the encoding error measurement program is installed from the recording medium 108 to the auxiliary storage device 104 via the drive device 103. The encoding error measurement program downloaded from the network is installed in the auxiliary storage device 104 via the interface device 107.

  The encoding error measurement device stores the installed encoding error measurement program and stores necessary files, data, and the like. The memory device 105 reads and stores the coding error measurement program from the auxiliary storage device 104 when the computer is started. Then, the arithmetic processing unit 106 can realize the various processing blocks shown in FIG. 3 according to the coding error measurement program stored in the memory device 105.

  The present invention obtains an encoding parameter from a bit stream for the purpose of measuring and monitoring the quality of a compressed encoded signal (bit stream) obtained by compressing and encoding a video signal by compression encoding technology, and analyzing the encoded parameter Thus, how much the image quality is deteriorated by the compression encoding is quantified.

  For example, the present invention can be applied to a monitoring facility such as a broadcasting station or a network operator that automatically monitors how much the image quality of a transmission image is deteriorated by compression encoding using an automatic monitoring technique.

It is a block diagram of one Example of the image transmission system containing the encoding error measuring apparatus of this invention. It is a block diagram of one Example of a decoding part. It is a block diagram of one Example of an encoding error measuring apparatus. It is a figure which shows the relationship between motion compensation prediction and a picture type. It is a block diagram of an example of the computer system which implement | achieves an encoding error measuring apparatus.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Coding part 20 Decoding part 21,41 Orthogonal transformation coefficient extraction part 22 Variable length decoding part 23 Inverse quantization part 24,44 Inverse orthogonal transformation part 25,45 Adder 26,47 Video memory 27,46 Motion compensation prediction part 28 Format conversion unit 30 Monitor 40 Coding error measuring device 42, 49 Moment calculation unit 43 Coding error calculation unit 48 Orthogonal transformation unit 50 Moment prediction value calculation unit 51 Differentiator 52, 53 Switch 60 Transmission network 101 Input device 102 Output device 103 Drive device 104 Auxiliary storage device 105 Memory device 106 Arithmetic processing device 107 Interface device

Claims (3)

A coding error measuring device for measuring a coding error of a compression coded signal obtained by compression coding a video signal,
The motion reference image decoded from the compression-encoded signal that uses motion compensation prediction is orthogonally transformed to create a frequency distribution thereof, while the motion reference image decoded from the compression-encoded signal that also uses motion compensation prediction Means for orthogonally transforming the image to be measured with reference to create a frequency distribution thereof,
The difference between the average value of the variance of the frequency distribution of the image to be measured and the average value of the variance of the frequency distribution of the motion reference image is a lost coding error that is not measured in the image to be measured. And an encoding error measuring device.   Means for calculating a coding error of the measurement target image based on a lost coding error that is not measured in the measurement target image and a coding error measured from within the measurement target image; The encoding error measuring apparatus according to claim 1, comprising: An encoding error measurement program for measuring an encoding error of a compressed encoded signal obtained by compressing and encoding a video signal executed in a computer having at least a storage device and an arithmetic processing unit,
The arithmetic processing unit orthogonally transforms a motion reference image decoded from the compression-encoded signal used in combination with motion compensation prediction, and creates a frequency distribution thereof.
Orthogonally transforming an image to be measured referring to the motion reference image decoded from the compression-encoded signal used in combination with motion compensation prediction, and creating a frequency distribution thereof;
The difference between the average value of the variance of the frequency distribution of the image to be measured and the average value of the variance of the frequency distribution of the motion reference image is a lost coding error that is not measured in the image to be measured. A coding error measurement program that executes the step of measuring as

Images