JP4676513B2 - Encoded picture type determination method, apparatus, program thereof, and recording medium thereof - Google Patents

Description

  The present invention relates to a method for determining an encoded picture type for video encoding.

  MPEG-2 and H.264 In a moving image encoding method such as H.264, encoding can be performed by switching picture types due to differences in encoding tools. I picture that is closed and encoded in the picture without motion compensation prediction, P picture that performs inter-picture prediction by referring to the previous encoded picture in time, and past or future encoded pictures can be referenced There is a B picture. These multiple picture types are combined to form a GOP (Group of Pictures) and used as a unit for random access and bit rate control.

  There are N and M to indicate the GOP structure. In general, the encoding interval for I pictures is N, and the encoding interval for P pictures is M. FIG. 8 shows an example of the reference relationship of each picture type. FIGS. 8A to 8D are frequently used in the conventional moving picture coding system. FIG. An example using a B picture (hereinafter referred to as Bs) that can be referred to from other pictures adopted in H.264 is shown.

  In this GOP structure, the optimum reference relationship differs depending on the video to be encoded. For example, in a fast-moving video, since the correlation between pictures is small, a reference relationship with a short inter-picture distance is advantageous (for example, FIGS. 8A and 8B). Therefore, a method for determining a more efficient GOP structure has been proposed.

  For example, in the scene adaptive moving image encoding device described in Patent Document 1, the picture type is determined by examining the efficiency of motion compensation prediction in advance for a picture to be encoded. In this apparatus, the motion compensated prediction characteristics are detected before the encoding process, and the picture type of the picture to be encoded is determined.

  FIG. 9 is a flowchart of the prior art, and FIG. 10 is a block diagram of the prior art.

  The pre-analysis unit 200 in FIG. 10 detects the motion compensation prediction characteristics of the current picture from the input image signal before the encoding process (step S100). The picture type determination unit 201 determines the picture type of the encoding target picture based on the motion compensation prediction characteristics detected by the pre-analysis unit 200 (step S101). The picture rearrangement unit 202 rearranges the pictures according to the determined picture type, and changes the encoding order (step S102). The encoding processing unit 203 performs predictive encoding on the image signals of the pictures rearranged by the picture rearrangement unit 202 (step S103).

  The subtractor 204, orthogonal transform unit 205, quantization unit 206, information source coding unit 207, inverse quantization unit 208, inverse orthogonal transform unit 209, adder 210, frame memory 211, motion in the encoding processing unit 203 The predictive encoding process performed by the search unit 212 and the motion compensation unit 213 is a well-known process and will not be described in detail.

According to such a method, it is possible to select an optimal picture type for the input video, and the coding efficiency is improved as compared with a method in which the GOP structure is fixed.
JP 2002-77924 A

  In the technique of Patent Document 1 described above, the motion compensation prediction efficiency is estimated and the picture type is selected by performing motion detection and motion compensation prediction on the encoding target picture before the encoding process. Here, the motion compensation prediction characteristic is obtained from the result in the coding loop.

  In this method, when the picture type is changed, the encoding order needs to be changed, and the processing becomes complicated. For example, when a B picture is selected from the motion compensated prediction characteristics, since the picture type to be encoded next is a P picture, information in the encoding loop must be deleted. As described above, the conventional methods have problems such as an increase in calculation cost and a complicated processing structure.

  The present invention solves the above-mentioned problem, and determines the optimal picture type (GOP structure) for a picture to be coded without checking the motion compensation efficiency before coding, and thereby before coding. It is an object of the present invention to eliminate the need for a prior motion vector search for a picture to be encoded, and to reduce the calculation cost and simplify the processing structure.

  In a video encoding system having a plurality of inter-picture prediction modes, encoding results such as motion vector and prediction mode selection vary depending on the encoding efficiency of the picture type and the video characteristics of the input video. Therefore, the video property of the encoding target picture is estimated using the encoding result of the encoded picture. As the encoding result, information obtained on the encoder side, such as the selection ratio of the prediction mode and the norm of the motion vector, is used.

That is, in order to solve the above-described problem, the present invention pays attention to the prediction mode and motion vector of the inter-picture prediction picture encoded in the past when switching the picture type of inter-picture prediction, and encodes the encoding target picture. Before calculating the statistic only of the encoding result of the inter-picture prediction picture that has been encoded before the current encoding target picture in the same encoding target picture , a value obtained from the statistical amount and a predetermined threshold value And the picture type of the encoding target picture is determined from the comparison result. The value obtained from the statistic may be the statistic itself or a value such as a cost value obtained from the statistic.

  In the prior art, since the efficiency of motion compensated prediction of the encoding target picture is calculated and the picture type is determined before the encoding process, the signal of the encoding target picture is used to determine the encoding target picture. Is required. On the other hand, in the present invention, since the picture type of the encoding target picture is determined using the encoding result of the encoded picture, there is no need to input the encoding target picture itself, which is the conventional technique. And very different. This eliminates the need for a prior motion vector search for the encoding target picture before encoding, thereby reducing the calculation cost and the circuit scale of the preliminary analysis unit.

Et al is, it is also effective to select the optimum picture type determines picture types by using only the results of the bidirectional predictive pictures.

  In the above-described invention, a method of determining a picture type using one or more values of a motion vector, a prediction mode, a prediction block size, and a prediction error power for an encoded inter-picture prediction picture is also preferable.

  In the above invention, when determining a picture type using a plurality of encoding results (motion vector, prediction mode, prediction block size, prediction error power, etc.) of an encoded picture, the type of encoding result to be used It is also possible to use a method in which a predetermined threshold value is set for each time, the encoding result is compared with the threshold value, and the picture type of the encoding target picture is determined from the comparison result.

  In the above invention, when a picture type is determined using a plurality of encoding results (motion vector, prediction mode, prediction block size, prediction error power, etc.) of an encoded picture, the encoding result to be used is set as a parameter. It is also preferable to determine the picture type of the picture to be encoded from the comparison result using a means for calculating the cost by substituting it into a cost function and a means for comparing the calculated cost with a predetermined threshold. is there.

  In the above invention, when using a motion vector of an encoded inter-picture prediction picture, means for calculating a statistic such as an average value, a maximum value, and a variance for each component or norm of the motion vector is used as a statistic. It is also preferable to determine the picture type using the calculated statistic.

  In the above invention, when using the prediction mode and the prediction block size of the encoded inter-picture prediction picture, a means for calculating a selection ratio of each prediction mode and prediction block size is used as a statistic, and the calculated selection ratio is calculated. It is also preferable to determine the picture type using.

  In the above invention, the means for calculating the selection ratio between the bidirectional prediction mode and the reduced overhead prediction mode from the encoded bidirectional prediction picture, the means for comparing the calculated selection ratio with a predetermined threshold, and the comparison result in advance If it falls within the defined range, use a means for calculating the selection ratio of prediction modes other than forward prediction and intra prediction, and a means for comparing the selection ratio with a predetermined threshold, and using the statistics of the prediction mode. , Determining the picture type step by step also gives good results.

  In the above invention, it is also preferable to determine the picture type using the encoding results of both the encoded one-way prediction picture and bidirectional prediction picture. Furthermore, the picture type is determined from the calculated statistic by using the means for measuring the statistic of the motion vector of the encoded unidirectional prediction picture and the means for measuring the statistic of the prediction mode of the bidirectional prediction picture. Also good results can be obtained.

  In the above invention, when setting the threshold value, a means for calculating the threshold value from the picture size and frame rate of the input video and the encoding bit rate or the allocated code amount of each picture is used to match the encoding condition and the input video. The threshold value is changed to determine the picture type. By automatically switching the threshold values in this way, an appropriate threshold value can be used according to the situation, and more efficient picture type selection is possible.

  According to such a method, an efficient picture type can be selected without performing motion compensation prediction or the like of the encoding target picture before encoding. Since it is not necessary to check the coding efficiency of the picture to be coded in advance, it can be expected to reduce the calculation cost and the circuit scale.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. A basic flowchart of the embodiment of the present invention is shown in FIG.

  First, statistics such as a motion vector and a prediction mode are calculated for a picture that performs inter-picture prediction, such as a P picture or a B picture (step S1).

The following are examples of statistics.
(1) Average, maximum, and variance of horizontal and vertical components of motion vector (2) Average, maximum, and variance of motion vector norm (3) Selection ratio of each prediction mode and prediction block size (4 ) Prediction error power In the above statistics, the average value, maximum value, and norm of the motion vector indicate the speed of motion of the video, and the variance of the motion vector indicates variation in motion. The characteristics of the video can be known from the selection ratio of the prediction mode and the prediction block size. For example, in the B picture, the similarity with the preceding and following reference pictures can be seen from the prediction direction. The size of the moving area can be estimated from the predicted block size. When many small prediction block sizes of 4 × 4 size are selected, there is a possibility that the motion in the screen is not uniform. The prediction error power can be used to know the prediction efficiency.

  Further, when calculating the statistic, the determination ratio may be calculated using a threshold value. For example, in the case of motion vectors, a threshold value for determining the speed of motion can be provided, and the ratio of motion vectors determined to be fast can be used.

  Next, the picture type of the picture to be coded is determined based on the statistic obtained from the coding result of the coded picture (step S2). The picture type is determined by using a statistic of a specific encoded picture type or a plurality of picture types.

(1) Use a statistic of a specific picture type Using a statistic of one type of picture type such as a B picture, the suitability of the encoding target picture is determined, and the picture type of the encoding target picture is determined.

(2) Use statistics of a plurality of picture types Compare statistics of each picture type, select a picture type that seems to have high encoding efficiency, and determine a picture type of a picture to be encoded.

The statistics used to determine the picture type are a combination of one or more statistics. When combining multiple statistics,
(I) calculating the cost using a cost function with each statistic as a parameter;
(Ii) Judgment by threshold for each statistic.
There is a method.

  When using a cost function, a cost value is calculated from a plurality of statistics. For example, the cost Cost is obtained using the prediction error power D and the generated code amount R, and the picture type is determined from this cost. λ is Lagrange's undetermined multiplier.

Cost = D + λ · R (Formula 1)
In the case of threshold determination, a threshold is determined for each statistic, and the picture type is determined based on the magnitude relationship. It is also possible to prioritize the statistics to be compared and change the statistics threshold in the latter stage according to the conditions. For example, when using statistic A and statistic B, if statistic A is greater than or equal to threshold TH _A , the threshold of statistic B is TH _B1 , and if statistic A is less than threshold TH _A , the threshold of statistic B is Can be switched to TH _B2 .

  In this way, by determining the picture type using the encoding result of the encoded picture, the picture type of the encoding target picture can be determined without analyzing the input video before encoding.

  FIG. 2 is a block diagram showing a configuration example of an encoding apparatus using the present invention. In FIG. 1, a statistic calculation unit 10 is means for calculating a statistic of the encoding result of an encoded picture. The picture type determination unit 11 compares the value obtained from the statistic with a predetermined threshold using the statistic calculated by the statistic calculation unit 10, and determines the picture type of the picture to be encoded from the comparison result. . The picture rearrangement unit 12 switches the picture type for inter-picture prediction based on the determined picture type, and changes the coding order according to the reference relationship.

  The encoding processing unit 100 is the same as the configuration of the conventional encoding processing unit using motion compensation prediction, except that the encoding result of the encoded picture is output to the statistic calculation unit 10. The encoding processing unit 100 inputs image signals of encoding target pictures rearranged by the picture rearranging unit 12.

  In the encoding processing unit 100, the subtractor 101 calculates a prediction error between the input image signal and the predicted image signal. The orthogonal transform unit 102 performs orthogonal transform such as DCT on the prediction error signal, and the quantization unit 103 quantizes the output of the orthogonal transform unit 102. The information source encoding unit 104 performs variable length encoding on the quantized signal and outputs it as encoded data.

  On the other hand, the output of the quantization unit 103 is inversely quantized by the inverse quantization unit 105, and the inverse orthogonal transform unit 106 performs inverse orthogonal transform such as IDCT on the inverse quantized value. The adder 107 adds the predicted image signal to the prediction error obtained by the inverse orthogonal transform, obtains a decoded image signal for reference, and stores it in the frame memory 108.

  The motion search unit 109 performs motion estimation from the input image signal and the decoded image signal stored in the frame memory 108, obtains a motion vector, and outputs the motion vector to the motion compensation unit 110. The motion compensation unit 110 generates a predicted image signal from the motion vector and outputs it to the subtracter 101. The predicted image signal is also output to the adder 107 for local decoding.

  Information on motion vectors and prediction modes obtained by the motion search unit 109 is encoded by the information source encoding unit 104 and output as encoded data. Also, information on these encoding results is also output to the statistic calculation unit 10.

[Example 1]
Specific examples of the present invention are shown below. In example 1, H. Assuming that the picture type is switched in H.264, M = 2 in FIG. 8B and M = 4 in FIG. 8E are switched. Only the statistic of the encoded B picture is used for switching, and the following two types of statistic are used.
(1) Selection ratio BMC _{bi of} bidirectional prediction + overhead saving prediction mode
(2) Selection ratio BMC _all of prediction modes other than forward prediction and intra prediction
The reduced overhead prediction mode is H.264. This is a skip mode or direct mode used in H.264. The selection ratio BMC _all other than the forward prediction and the intra prediction is obtained by adding the backward prediction mode to the selection ratio BMC _bi of the bidirectional prediction + saving overhead prediction mode.

  The picture type switching is performed by threshold determination, and M = 2 and M = 4 are switched in units of 4 pictures. FIG. 3 shows an example of switching the M value in units of 4 pictures.

  In this embodiment, since the M value is switched in units of 4 pictures, the M value switching process executes the following process after the last B picture process in units of 4 pictures.

FIG. 4 shows a flowchart of this embodiment.
[Step S10] Measure the B picture prediction mode statistic:
The statistic of the encoded B picture 20 shown in FIG. 3 is measured.
[Step S11] Determination of M Value:
The M value is determined from the measured statistics. Details of this processing will be described later with reference to FIG.
[Step S12] Determination of picture type:
Here, picture types up to four frames ahead are determined based on the M value determined in step S11.
[Step S13] Rearrange pictures:
The coding order is changed according to the reference relationship.
[Step S14] Encoding process:
The encoding process is performed in order according to the picture type determined in step S12.

Here, the method for determining the M value in step S11 will be described. First, the selection ratio BMC _bi in the bidirectional prediction + saving overhead prediction mode is compared with a threshold value. When the selection ratio BMC _bi is between two thresholds TH _H and TH _L (where TH _H > TH _L ), the threshold TH for the selection ratio BMC _all of prediction modes other than forward prediction and intra prediction is used. Compare with _H. FIG. 5 shows a flowchart of processing for determining the M value.
[Step S20] Calculate BMC _bi :
The selection ratio BMC _bi of the bidirectional prediction + saving overhead prediction mode is calculated.
[Step S21] BMC _bi > TH _H :
Comparing the selection ratio BMC _bi and the threshold TH _H, when the selection ratio BMC _bi exceeds the threshold TH _H, the step S25 advances to (M = 4), if more than the threshold TH _H, the next step S22 Proceed to
[Step S22] BMC _bi > TH _L :
Comparing the selection ratio BMC _bi and the threshold TH _L, if the selected percentage BMC _bi is equal to or less than the threshold value TH _L, the step S26 advances to (M = 2), if it exceeds the threshold value TH _L, the next step S23 Proceed to
[Step S23] Calculate BMC _all :
A selection ratio BMC _all of prediction modes other than forward prediction and intra prediction is calculated.
[Step S24] BMC _all > threshold TH _H :
Comparing the selection ratio BMC _all the threshold TH _H, when the selection ratio BMC _all exceeds a threshold value TH _H, the step S25 advances to (M = 4), if more than the threshold TH _H, the step S26 (M = Go to 2).
[Step S25] M = 4:
The encoding interval M of the P picture is determined to be 4.
[Step S26] M = 2:
The encoding interval M of the P picture is determined as 2.

  FIG. 6 shows the relationship between the statistics and the GOP structure, and particularly shows the relationship between each selection ratio and the M value. For example, as shown in FIG. 6, the M value is determined by the positional relationship between the two types of selection ratios.

[Example 2]
A second example 2 is shown. In the second embodiment, as in the first embodiment, M = 2 in FIG. 8B and M = 4 in FIG. 8E are switched. In addition to the previous B picture encoding result, the norm of the motion vector is used as the statistic of the P picture. A threshold TH _MVnorm is set for the norm of the motion vector, and a motion vector ratio MV _f exceeding this threshold is calculated. Using the ratio MV _f of the motion vector, it switches the picture type.

FIG. 7 shows a flowchart of the second embodiment.
[Step S30] Measure the motion vector ratio of the P picture:
The norm of the motion vector of the encoded P picture is calculated, and it is determined whether the calculated norm exceeds the threshold value TH _MVnorm . A motion vector ratio MV _f exceeding the threshold TH _NVnorm is calculated.
[Step S31] Compare the motion vector ratio MV _f with the threshold TH _r :
If the motion vector ratio MV _f is greater than the threshold value TH _r , the process proceeds to step S32; otherwise, the process proceeds to step S33.
[Step S32] M = 2:
If the ratio MV _f of the motion vector is larger than the threshold TH _r, as a fast motion determines the M = 2, the process proceeds to step S35.
[Step S33] Measure the B picture prediction mode statistic:
The statistic of the encoded B picture 20 shown in FIG. 3 is measured.
[Step S34] Determination of M Value:
The M value is determined from the measured statistics. Here, for example, the M value is determined by the same algorithm as the determination of the M value of the first embodiment described with reference to FIG.
[Step S35] Determination of picture type:
Here, picture types up to four frames ahead are determined based on the M value determined in steps S32 and S34.
[Step S36] Rearranging pictures:
The coding order is changed according to the reference relationship.
[Step S37] Encoding Processing Encoding processing is performed in order according to the picture type determined in step S36.

  In this embodiment, the speed of motion is confirmed using a motion vector of a P picture. For images with fast motion, M = 2 is determined by the same method as in the first embodiment except that M = 2 is fast and efficient.

  In this embodiment, the norm of the motion vector of the P picture is used, but the value changes according to the distance between the reference pictures. Therefore, when using motion vectors, it is better to normalize with the distance between reference pictures. By normalizing, even if the distance between reference pictures changes, it can be handled in the same way.

  In the determination of the encoded picture type in the above embodiment, a method of determining the picture type by changing the threshold according to the encoding condition and the input video is also suitable. For this reason, when setting the threshold value, means for calculating the threshold value from the picture size and frame rate of the input video and the encoding bit rate or the allocated code amount of each picture is provided. Thereby, for example, the threshold value is changed as follows.

(1) Picture size Since the block size changes when the picture size changes, the statistic changes. When the picture size is large, the threshold value is lowered, and when the picture size is small, the threshold value is raised.

(2) Frame rate When the frame rate changes, the distance between frames changes and the statistics change. When the frame rate is high, the inter-frame correlation is high, so the threshold value is lowered. When the frame rate is low, the inter-frame correlation is low, and the threshold value is raised.

(3) Encoding bit rate The overhead cost ratio varies according to the bit rate, the prediction mode selection ratio changes, and the statistics change. When the bit rate is high, the threshold value is lowered, and when the bit rate is low, the threshold value is raised.

(4) Allocated code amount of each picture Similar to the encoding bit rate, the selection ratio of the prediction mode varies depending on the allocated code amount. For example, when the allocation ratio to the B picture is large, the threshold is decreased, and when the allocation ratio is small, the threshold is increased.

  When the inter-frame correlation is small and the B code allocation code amount is small, the direct mode and skip frequently occur because control for reducing the overhead cost works. For this reason, even if many of these prediction modes are selected, the optimal picture type is not necessarily selected.

  Parameters such as picture size, frame rate, coding bit rate, or assigned code amount of each picture can be compared with the assigned code amount per pixel. The number of pixels per unit time is “picture size × frame rate”. When the bit rate is divided by the number of pixels, bit / pel is obtained. A threshold value can be determined by creating a table of the assigned code amount per pixel and the threshold value in advance and referring to this table with the code amount per pixel.

  The coded picture type determination process described above can be realized by a computer and a software program. The program can be provided by being recorded on a computer-readable recording medium or via a network. is there.

It is a flowchart of an embodiment of the invention. It is a block diagram which shows the structural example of the encoding apparatus using this invention. It is a figure which shows the example which switches M value per 4 pictures. 3 is a flowchart of the first embodiment. It is a flowchart of the process which determines M value. It is a figure which shows the relationship between a statistic and GOP structure. 10 is a flowchart of Example 2. It is a figure which shows the example of the reference relationship (GOP structure) of each picture type. It is a flowchart of a prior art. It is a block diagram of a prior art.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Statistic calculation part 11 Picture type determination part 12 Picture rearrangement part 100 Encoding process part 101 Subtractor 102 Orthogonal transformation part 103 Quantization part 104 Information source encoding part 105 Inverse quantization part 106 Inverse orthogonal transformation part 107 Adder 108 Frame memory 109 Motion search unit 110 Motion compensation unit

Claims (24)

A coding picture type determination method in a video coding system having a plurality of inter-picture prediction modes,
Calculating a statistic of only the encoding result of the inter-picture prediction picture that has been encoded before the encoding target picture in the same encoding target video before encoding the encoding target picture ;
Using the statistic, comparing a value obtained from the statistic with a predetermined threshold, and determining a picture type of the encoding target picture from a comparison result;
A method for determining an encoded picture type, comprising: switching a picture type for inter-picture prediction based on the determined picture type. The encoded picture type determination method according to claim 1 ,
The code haze when using only coding result of viewing Picture predictive picture coded picture type determination method characterized by utilizing only the coding result of the bidirectional predictive pictures. The encoded picture type determination method according to claim 1 ,
As the coding result of the code haze seen Picture predictive picture, the motion vector, prediction mode, prediction block size of the prediction error power, the encoding picture type determination, which comprises using one or more values Method. The encoded picture type determination method according to claim 3 ,
When a picture type is determined using a plurality of encoding results of the encoded inter - picture prediction picture, a threshold value set in advance for each type of encoding result to be used is used. A coding picture type determination method, comprising: comparing a threshold value corresponding to a type of the coding result and determining a picture type of the coding target picture from the comparison result. The encoded picture type determination method according to claim 3 ,
A process of calculating a cost value by substituting a statistic of the encoding result to be used as a parameter when determining a picture type using a plurality of encoding results of the encoded inter - picture prediction picture Have
An encoded picture type determination method, wherein the cost value is compared with the threshold value, and a picture type of the encoding target picture is determined from a comparison result. The encoded picture type determination method according to claim 1 ,
When using a motion vector as the encoding result of the code haze seen Picture predictive picture, as statistics, to have a process of calculating an average value for each component or norm of the motion vector, the maximum value or dispersion Characterized coded picture type determination method. The encoded picture type determination method according to claim 1 ,
When using the prediction mode or prediction block size as a coding result of the code haze seen Picture predictive picture, and characterized in that a statistic comprises the step of calculating the respective prediction mode or selecting the ratio of the prediction block size A method for determining an encoded picture type. The encoded picture type determination method according to claim 7 ,
From the bidirectional predictive pictures of said code haze seen Picture predictive picture, comprising the steps of: calculating a first selection ratio of the bidirectional predictive mode and saving overhead prediction mode,
Comparing the first selection ratio with a predetermined threshold;
Calculating a second selection ratio obtained by adding a selection ratio of the backward prediction mode to the first selection ratio when the comparison result falls within a predetermined range;
Comparing the second selection ratio with a predetermined threshold;
An encoded picture type determination method characterized in that the picture type is determined step by step using prediction mode statistics. The encoded picture type determination method according to claim 1 ,
As the encoding result of the prediction picture between the coded Kasumi see Picture, coding picture type determination method which comprises using the encoding results of both the picture type of coded unidirectional prediction picture and bidirectionally predictive picture . The coded picture type determination method according to claim 9 ,
Measuring a motion vector statistic of the encoded unidirectional prediction picture;
Measuring a prediction mode statistic of the bi-predictive picture,
An encoded picture type determination method, characterized by determining a picture type from a calculated statistic. The encoded picture type determination method according to claim 1 ,
Calculating the threshold from the picture size of the input video, the frame rate, the encoding bit rate or the assigned code amount of each picture, or the assigned code amount per pixel,
An encoded picture type determination method, wherein the calculated threshold value is used for comparison with a value obtained from the statistics, and the picture type is determined by changing the threshold value according to an encoding condition or an input video. An encoded picture type determination apparatus in a video encoding system having a plurality of inter-picture prediction modes,
Means for calculating a statistic of only the encoding result of the encoded inter-picture prediction picture encoded before the encoding target picture in the same encoding target video before the encoding target picture encoding process;
Means for comparing a value obtained from the statistic with a predetermined threshold using the statistic, and determining a picture type of the picture to be encoded from a comparison result;
An encoded picture type determining apparatus comprising: means for switching a picture type for inter-picture prediction based on the determined picture type. The encoded picture type determination device according to claim 12 ,
The code haze when using only coding result of viewing Picture predictive picture coded picture type determination apparatus characterized by utilizing only the coding result of the bidirectional predictive pictures. The encoded picture type determination device according to claim 12 ,
As the coding result of the code haze seen Picture predictive picture, the motion vector, prediction mode, the predicted block size, among the prediction error power, the encoding picture type determination, which comprises using one or more values apparatus. The encoded picture type determining apparatus according to claim 14 ,
When a picture type is determined using a plurality of encoding results of the encoded inter - picture prediction picture, a threshold value set in advance for each type of encoding result to be used is used. An encoded picture type determining apparatus, which compares a threshold corresponding to a type of the encoded result and determines a picture type of the encoding target picture from the comparison result. The encoded picture type determining apparatus according to claim 14 ,
Means for calculating a cost value by substituting a statistic of the encoding result to be used as a parameter when determining a picture type using a plurality of encoding results of the encoded inter - picture prediction picture Prepared,
An encoded picture type determining apparatus, which compares the cost value with the threshold and determines a picture type of the encoding target picture from a comparison result. The encoded picture type determination device according to claim 12 ,
When using a motion vector as the encoding result of the prediction picture between the coded Kasumi see Picture, as statistics, further comprising means for calculating the average value for each component or norm of the motion vector, the maximum value or dispersion An encoded picture type determination device as a feature. The encoded picture type determination device according to claim 12 ,
When using the prediction mode or prediction block size as a coding result of the code haze seen Picture predictive picture, and characterized in that a statistic comprises means for calculating each prediction mode or selecting the ratio of the prediction block size An encoded picture type determination device. The encoded picture type determination device according to claim 18 ,
From the bidirectional predictive pictures of said code haze seen Picture predictive picture, it means for calculating a first selection ratio of the bidirectional predictive mode and saving overhead prediction mode,
Means for comparing the first selection ratio with a predetermined threshold;
Means for calculating a second selection ratio obtained by adding a selection ratio of the backward prediction mode to the first selection ratio when the comparison result falls within a predetermined range;
Means for comparing the second selection ratio with a predetermined threshold;
An encoded picture type determination device characterized in that a picture type is determined step by step using a prediction mode statistic. The encoded picture type determination device according to claim 12 ,
As the encoding result of the prediction picture between the coded Kasumi see Picture, coding picture type determination apparatus characterized by using the coding results of both the picture type of coded unidirectional prediction picture and bidirectionally predictive picture . The encoded picture type determination device according to claim 20 ,
Means for measuring a statistic of a motion vector of the encoded unidirectional prediction picture;
Means for measuring a statistics of a prediction mode of the bidirectional prediction picture,
An encoded picture type determining apparatus, wherein a picture type is determined from a calculated statistic. The encoded picture type determination device according to claim 12 ,
Means for calculating the threshold from the picture size of the input video, the frame rate, the encoding bit rate or the assigned code amount of each picture, or the assigned code amount per pixel;
An encoded picture type determining apparatus, wherein the calculated threshold value is used for comparison with a value obtained from the statistics, and the picture type is determined by changing the threshold value according to an encoding condition or an input video. An encoded picture type determination program for causing a computer to execute the encoded picture type determination method according to any one of claims 1 to 11 . A computer-readable recording medium on which the encoded picture type determination program according to claim 23 is recorded.

Images