JP2002247584A - Method and device for encoding image, program for image encoding processing, and recording medium for the program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image encoding technology capable of improving the efficiency of encoding by accurately estimating the number of generated codes. SOLUTION: Constitution for approximately estimating an I picture from the dispersion value of an input image in a frame to be encoded and approximately estimating a P picture and a B picture from the predictive error value of movement search executed between the input image of the frame to be encoded and the input image of a reference frame is used as a means for predicting the number of generated codes in each picture type. Since the number of codes of the frame to be encoded is predicted on the basis of not the information of past frames but the information of the frame to be encoded in accordance with the constitution, a problem that prediction is not hit due to a change in a scene or the like can be solved.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method in which a variance value of an input image and a result of motion search between input images are used in a subsequent encoding process.
An image encoding method and apparatus that determines the frame complexity index of each frame based on the preceding frame and controls the encoding based on the frame index, and an image encoding process used to realize the image encoding technique And a recording medium for an image encoding processing program that stores the program.

[0002]

2. Description of the Related Art In a conventional coding control method of image coding such as MPEG-2 Test Model 5, for each picture type such as I, P, and B, the generated code amount of the picture, the average quantization scale, and the like are calculated. Is calculated after the encoding of the picture is completed, and based on the value, the generated code amount of the picture to be encoded next is predicted and reflected in the encoding control.

In the invention described in Japanese Patent Application Laid-Open No. H11-298904, prior to encoding at a later stage, a code amount is estimated based on a statistic output from a reduced picture motion detecting unit, and the encoding is performed. By reflecting the results in the control, a real-time video encoding device having the same hardware scale as that of the one-pass video encoding device and having the same meaning as two-pass video is realized.

[0004]

However, in the former method of the prior art, since the amount of generated code is estimated based on the information of past pictures, the pattern of the previous picture greatly changes. In such a case, or when a scene is switched, the generated code amount cannot be estimated, so that the code amount distribution becomes inaccurate and the coding efficiency decreases.

In the latter prior art method, only the statistical information based on the reduced picture motion detection is directly used for code amount prediction.
Accuracy is not sufficient to accurately predict the code amount for each picture type such as a picture, and as a result, there is a problem that the code amount distribution is inaccurate and the coding efficiency may decrease.

The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a new image coding technique which realizes an improvement in coding efficiency by accurately estimating a generated code amount. And

[0007]

[Means for Solving the Problems] MPEG-2 Test Model
According to the method, a complexity index is obtained based on information of pictures of the same picture type in the past, and the amount of generated code is estimated. Therefore, the pattern of the previous picture of the same picture type greatly changes. However, when the scene is switched or when the scene is switched, the generated code amount cannot be estimated, so that the code amount distribution becomes inaccurate and the coding efficiency decreases.

Therefore, in the present invention, as means for predicting the generated code amount of each picture type, an I picture is approximately estimated from a variance of an input image of a frame to be coded, and a P picture and a B picture are Is used for approximation from the prediction error value of the motion search performed between the input image of the reference frame and the input image of the reference frame.

According to this configuration, since the code amount of the encoding target frame is predicted based on information about the encoding target frame instead of the past frame, the problem that the prediction cannot be achieved due to a scene change or the like is solved. You.

In the invention described in Japanese Patent Application Laid-Open No. H11-298904, a statistic obtained from a reduced motion search on a frame to be coded is performed by performing a reduced picture motion search prior to a subsequent encoding process. Predicts the code amount from
This is reflected in the code amount control.
When predicting the code amount of each picture type such as a picture, a P picture, a B picture, etc., accuracy was sometimes insufficient.

Therefore, in the present invention, the I picture uses the variance value of the input image of the encoding target frame, and the P picture and the B picture are used between the input image of the encoding target frame and the input image of the reference frame. A configuration is used in which the generated code amount is approximately estimated according to the estimation formula for each picture type using the prediction error value of the performed motion search.

According to this configuration, the code amount can be accurately predicted by a method suitable for the picture type of the picture.

Further, according to the present invention, a complexity index is obtained from a product of a frame generation code amount obtained from a result of actual encoding processing and a frame average quantization scale, and the information, such a variance value and such a prediction value are obtained. An error value is accumulated for several frames for each picture type, and the prediction amount used for predicting the generated code amount is dynamically updated, so that the code amount can be accurately estimated.

Further, in the present invention, the code amount is accurately estimated by using a configuration in which the coefficient update of the prediction formula is controlled or the prediction formula used for prediction is changed using the scene change detection information. I can do it.

Further, in order to obtain information used in these methods, there is an advantage that, as described below, when hardware implementation is considered, there is little additional hardware that must be newly added.

In general, in MPEG coding, motion compensation is performed between an input image of a frame to be coded and a locally decoded image of a reference frame. However, when performing a hierarchical motion search, In a primary search with a coarser accuracy than a secondary search with a final accuracy, a motion search is often performed between an input image of an encoding target frame and an input image of a reference frame thereof.

Therefore, unlike the case of using the locally decoded image, the motion search between the input images can be performed in advance of the encoding independently. In consideration of hardware implementation, a primary search of coarse accuracy between input images and a final accuracy of 2 between the input image and the locally decoded image are performed.
Providing a frame delay between the next search is relatively easy to implement because it is not necessary to store the prediction error image and only the motion vector needs to be stored.

In MPEG, the quantization scale of each MB (macroblock) is often increased or decreased based on an activity value which is input image information of the MB. For example, in MPEG-2 Test Mode 1, 1
Although the minimum value of the variance of the four sub-blocks in the macroblock of 6 pixels × 16 pixels is used, in an encoding apparatus that calculates this value at the preceding stage of the encoding, the sum of this information is calculated. The variance value of the input image used for the prediction of the code amount of the I picture can be easily obtained, and can be used as it is for the prediction value of the intra-frame code amount.

When the difference between adjacent pixels of the input image is used as the variance of the input image, the calculation function can be implemented at relatively low cost by means of adding a line buffer in terms of hardware. .

For this reason, according to the present invention, it is possible to easily predict the amount of generated code of each picture type of I picture, P picture, and B picture ahead of the coding by several frames in implementation. It can contribute to the improvement of efficiency.

[0021]

DESCRIPTION OF THE PREFERRED EMBODIMENTS (1) First Embodiment FIG. 1 shows MPEG-2 according to a first embodiment of the present invention.
1 illustrates a device configuration of an image encoding device that performs encoding.

As shown in the figure, in the image coding apparatus according to the first embodiment, a preprocessing unit 10/1 primary search unit 11/2 secondary search prepared for performing MPEG-2 coding. In addition to the unit 12 / DCT unit 13 / quantization unit 14 / variable length encoding unit 15 / inverse quantization unit 16 / inverse DCT unit 17 / quantization parameter control unit 18, in order to realize the present invention,
An I-picture complexity index calculator 20 and a P / B picture complexity index calculator 21 are provided.

The I-picture complexity index calculating section 20 calculates
Using the variance of the input image input to the preprocessing unit 10,
A process of estimating the complexity index of the I picture (the product of the generated code amount of the picture and the average quantization scale) is performed.

The P / B picture complexity index calculator 21 calculates
A process of estimating the complexity index of the P picture and the complexity index of the B picture is performed using the prediction error value obtained from the motion vector search result of the primary search unit 11.

In the present invention, the quantization parameter control unit 18
Performs processing for controlling the quantization unit 14 using the complexity index estimated in this way.

Here, each processing means provided for realizing the image encoding apparatus of the present invention shown in FIG. 1 can be realized by a computer program, and the computer program is a recording medium such as a semiconductor memory. Can be recorded and provided.

In the first embodiment, as shown in FIG. 2, at the time of encoding immediately after a scene change, the I-picture complexity index calculating unit 20 uses the variance value x of the input image and calculates the frame complexity index = a _I × x ² + b _I × x + c _I ... where a _I, b _{I, and} c _I are fixed values that are set in advance.
A process for estimating the complexity index of the I picture is performed.

Further, a P / B picture complexity index calculating section 2
1 uses the prediction error value x of the P picture obtained from the motion vector search result at the time of encoding immediately after the scene change, and calculates the frame complexity index = a _P × x ² + b _P × x + c _P. , A _P, b _{P, and} c _P are fixed values that are set in advance.
A process for estimating the complexity index of the P picture is performed.

However, at the time of encoding for the first time after the scene change, processing for estimating the complexity index of the P picture is performed in accordance with the above-described equation for estimating the complexity index of the I picture.

The P / B picture complexity index calculation unit 2
1 uses the prediction error value x of the B picture obtained from the motion vector search result at the time of encoding immediately after the scene change, and calculates the frame complexity index = a _B × x ² + b _B × x + c _B. , A _B, b _{B, and} c _B are fixed values that are set in advance.
A process for estimating the complexity index of the B picture is performed.

On the other hand, according to the first embodiment,
At the time of encoding other than immediately after a scene change, it is considered that estimating the complexity index using a quadratic equation with a fixed coefficient value may not provide sufficient accuracy. The configuration employs a complexity index calculated by the product of the quantization amount and the average quantization parameter.

That is, the I picture complexity index calculation unit 2
0 obtains a complexity index calculated by multiplying the generated code amount of the immediately preceding I picture by the average quantization parameter during encoding other than immediately after the scene change, and passes it to the quantization parameter control unit 18. Is processed as follows. Also, P
The B picture complexity index calculation unit 21 acquires a complexity index calculated by a product of the code amount generated of the immediately preceding P picture and the average quantization parameter during encoding other than immediately after a scene change, and In addition to passing to the quantization parameter control unit 18, a complexity index calculated by the product of the generated code amount of the immediately preceding B picture and the average quantization parameter is obtained, and is passed to the quantization parameter control unit 18. Processing.

Therefore, in the first embodiment, FIG.
As shown in the processing flow, when the encoding process is started, it is determined whether or not a scene change has occurred, and if a scene change has occurred, the complexity of the I picture and the P picture is calculated using the above equation. Then, a complexity index of the B picture is calculated using the above-described formula.

Then, until the prescribed number of frames after the scene change occurs (for example, until a P picture and a B picture appear twice each),
Calculate the complexity index of the I picture using the above equation,
Calculate the complexity index of the P picture using the above equation,
The complexity index of the B picture is calculated using the above equation.

Here, with respect to the P picture, since the first P picture after the scene change is coded by an intra macroblock, the complexity index is calculated in the same manner as the I picture using the above equation. Calculate,
For the second and subsequent P-pictures after the scene change,
The complexity index is calculated using the above equation.

When the specified number of frames has elapsed, a measured value of the complexity index calculated by multiplying the generated code amount by the average quantization parameter is read, and the measured value is used as the calculated complexity index. You do it.

(A) Complexity Index Calculation Processing by I-picture Complexity Index Calculation Unit 20 FIG. 4 shows an embodiment of a processing flow executed by the I-picture complexity index calculation unit 20.

As shown in this processing flow, the I-picture complexity index calculating section 20 processes each frame of the input image input to the pre-processing section 10 with respect to each pixel, and obtains the luminance of each pixel with the pixel on the right. Calculate the sum of the absolute value of the difference between the values and the absolute value of the difference between the luminance values of the pixels immediately below and the sum of the entire screen, and calculate the sum of the absolute values of the differences between adjacent pixels and input it. A process of estimating the complexity index of the I picture is performed by using the variance value of the image and substituting it into the above equation.

(B) P · B picture complexity index calculation unit 2
1 for calculating a complexity index among frames of an input image sequence encoded as a P-picture or a B-picture.
Then, a motion search with 2-pixel accuracy is performed using the reduced image of the input image of the encoding target frame and the reduced image of the input image of the reference frame.

Here, the reduced image is an image obtained by reducing the input image by half in the horizontal and vertical directions. Also, both reduced images of the encoding target frame and the reference frame are created for each field from the first field and the second field of the input image, and the motion search is performed on a field basis.

In the search for two-pixel accuracy, the first and second fields of the encoding target frame are used as starting points, and the P picture is a total of four using the first and second fields of one reference frame as ending points. Searches are performed for two types of motion vectors, and for B pictures, a search is performed for a total of eight types of motion vectors ending at the first and second fields of two reference frames.

FIG. 5 shows a motion vector searched for a B picture. For P-pictures,
Only the reference frame A shown in the figure and v1, v2, v3, and v4 exist.

The P / B picture complexity index calculator 21 calculates
From the prediction errors corresponding to these motion vectors, a final prediction error is calculated, and it is substituted into the above-mentioned expression or expression, thereby performing a process of estimating the complexity index of the P picture or the B picture.

FIG. 6 shows an embodiment of the processing flow of the estimation process of the complexity index of the P picture executed by the P / B picture complexity index calculation unit 21, and FIG. 7 and FIG.
FIG. 6 illustrates an embodiment of a processing flow of a B picture complexity index estimation process executed by the B picture complexity index calculation unit 21. FIG.

The P / B picture complexity index calculator 21 calculates
When estimating the complexity index of the P picture, as shown in the processing flow of FIG. 6, the smaller one of the prediction errors of v1 and v2 shown in FIG. 5 and the prediction error of v3 and v4 shown in FIG. Is calculated as the prediction error for the macroblock, the sum within the frame is calculated, and the sum is used as the prediction error value of the frame to estimate the complexity index of the P picture.

On the other hand, the P / B picture complexity index calculating section 2
When estimating the complexity index of a B-picture, first, as shown in the processing flows of FIGS. 7 and 8, first, the smaller one of the prediction errors of v1 and v2 shown in FIG. 5, the sum of the smaller values of the prediction errors v3 and v4 is calculated, and then the smaller values of the prediction errors v5 and v6 shown in FIG. 5 and v7 and v8 shown in FIG. And the sum of the two smaller sums is calculated as the prediction error for the macroblock, the sum within the frame is calculated, and the sum is used as the prediction error value of the frame. Estimate the complexity index of the B picture.

As described above, the I-picture complexity index calculating unit 20 calculates the sum of absolute differences between adjacent pixels of the input image input to the pre-processing unit 10 at the time of encoding immediately after a scene change, and calculates the sum. By substituting into the quadratic expression, a process of estimating the complexity index of the I picture is performed.

Then, at the time of encoding immediately after the scene change, the P / B picture complexity index calculating section 21 obtains a prediction error value obtained from a motion vector search result of the primary search section 11, and calculates it by a quadratic equation. , A process of estimating the complexity index of the P picture and the complexity index of the B picture is performed.

As described above, in the first embodiment,
When a scene change is detected by any means, the variance value of the input image for an I picture, or the prediction error value of the motion search result between the input images for a P or B picture for a specified number of frames after the scene change. Is applied to a quadratic expression that is an estimation expression of a frame complexity index of each picture type, and a complexity index of each picture is estimated.

Here, the coefficient of the quadratic expression is a predetermined fixed initial setting value. However, the P picture that is coded for the first time after a scene change is coded as an intra macroblock in a large proportion.
The complexity index is estimated by an equation for estimating the complexity index of the I picture.

In the first embodiment, as for the picture other than immediately after the scene change, the estimation is not sufficiently performed by the estimation formula using the coefficient value as a fixed value. Using the product of the generated code amount of the picture immediately before the picture type and the average quantization parameter as a complexity index, inter-frame code amount distribution is performed to determine a picture target code amount and the like.

Here, the complexity index is the product of the amount of generated code of the picture and the average quantization parameter. However, the present invention is not limited to this. The value obtained by normalizing the generated code amount by the value of the quantization parameter by some method is used. Should be fine.

Further, the reference value of the quantization step of each macroblock in the frame or the initial value of the quantization step at the start of intra-frame coding is based on the value obtained by dividing the above-mentioned complexity index by the picture target code amount. To decide. As a result, it is possible to prevent the quantization step in the frame from greatly changing due to the code amount feedback control in the frame.

Although not described above, a frame delay of one or more frames is provided between the primary search and the secondary search, and after the frame prediction error value of the primary search is determined, the target code amount of the next picture is obtained. Needs to be estimated. At this time, if a frame delay is provided so that the prediction error value of the P picture two frames ahead is obtained in advance, even if a scene change occurs, the generated code amounts of all picture types after the scene change can be predicted in advance. Therefore, the coding efficiency is improved.

(2) Second Embodiment FIG. 9 shows MPEG-2 according to a second embodiment of the present invention.
1 illustrates a device configuration of an image encoding device that performs encoding.

As shown in this figure, in the image coding apparatus according to the second embodiment, the image encoding device according to the first embodiment is provided.
An I-picture complexity index calculating unit 30 is provided in place of the picture complexity index calculating unit 20, and P · B is used instead of the P / B picture complexity index calculating unit 21 provided in the first embodiment.
The configuration including the picture complexity index calculation unit 31 is adopted.

Here, each processing means provided for realizing the image encoding apparatus of the present invention shown in FIG. 2 can be realized by a computer program, and the computer program is a recording medium such as a semiconductor memory. Can be recorded and provided.

In the second embodiment, as shown in FIG. 10, the I-picture complexity index calculating section 30 uses the variance value x of the input image at the time of encoding other than scene change, and calculates the frame complexity index = a _I × x ² + b _I × x + c _I ... where a _I, b _{I, and} c _I are dynamic setting values, and the I picture is encoded according to a quadratic expression (coefficients change dynamically). At the end, a process of estimating the complexity index of the I picture is performed.

The P / B picture complexity index calculating section 3
1 uses a prediction error value x of a P picture obtained from a motion vector search result at the time of encoding other than a scene change, and calculates a frame complexity index = a _P × x ² + b _P × x + c _P. , A _P, b _{P, and} c _P are subjected to a process of estimating the complexity index of the P-picture at the end of encoding of the P-picture according to a quadratic expression (dynamic coefficient dynamically changes) of a dynamic setting value.

The P / B picture complexity index calculating section 3
1 uses a prediction error value x of a B picture obtained from a motion vector search result at the time of encoding other than a scene change, and calculates a frame complexity index = a _B × x ² + b _B × x + c _B. , A _B, b _{B, and} c _B perform a process of estimating the complexity index of the B-picture at the end of encoding of the B-picture according to a quadratic expression (coefficient dynamically changes) of a dynamic setting value.

Then, the I picture complexity index calculating section 30
Uses the variance value x of the input image at the time of encoding at the time of a scene change, and uses the frame complexity index = a _I0 × x ² + b _I0 × x + c _I0 ... Where a _I0, b _I0, c _I0 Performs a process of estimating the complexity index of the I-picture at the end of encoding of the I-picture in accordance with a quadratic expression (the coefficient is fixed to the initial value).

Further, the P / B picture complexity index calculating section 3
1 uses a prediction error value x of a P picture obtained from a motion vector search result at the time of encoding at a scene change, and a frame complexity index = a _P0 × x ² + b _P0 × x + c _P0. , A _P0, b _{P0, and} c _P0 are processes for estimating the complexity index of the P-picture at the end of the encoding of the P-picture according to a quadratic expression (initial value is fixed). Do.

However, at the time of encoding for the first time after the scene change, processing for estimating the complexity index of the P picture is performed in accordance with the above-described equation for estimating the complexity index of the I picture.

The P / B picture complexity index calculating section 3
1 uses a prediction error value x of a B picture obtained from a motion vector search result during encoding at the time of a scene change, and a frame complexity index = a _B0 × x ² + b _B0 × x + c _B0. , A _B0, b _{B0, and} c _B0 are processes for estimating the complexity index of the B picture at the end of encoding of the B picture in accordance with a quadratic expression (initial value is fixed). Do.

Therefore, in the second embodiment, FIG.
As shown in the processing flow of FIG. 1 and FIG. 12, when the encoding process starts, it is determined whether or not a scene change has occurred. If a scene change has occurred, the accumulated variance value / prediction error value / After clearing the complexity metric measurements,
The complexity index of the I picture and the P picture is calculated using the above equation, and the complexity index of the B picture is calculated using the above equation.

Then, until the prescribed number of frames after the scene change occurs (for example, until a P picture and a B picture appear twice, respectively), the complexity of the I picture is calculated using the above equation. Then, a complexity index of the P picture is calculated using the above equation, and a complexity index of the B picture is calculated using the above equation. At this time, the measured value of the complexity index is read and accumulated, and the variance value and the prediction error value used for estimating the complexity index are accumulated.

When the specified number of frames elapses, it is necessary to dynamically set the calculated variance / prediction error / complexity index measurement value using a calculation method such as linear regression. wherein the coefficient of above "a _I, b _I, c _I"
And the coefficients “a _P, b _P, c _P ” of the above equation and “a _B, b _B, c _B ” of the above equation are updated, and the updated equation is used to update the above equation. Is used to calculate the complexity index of the I picture, the above equation is used to calculate the complexity index of the P picture, and the above equation is used to calculate the complexity index of the B picture. At this time, a new complexity index measurement value / variance value /
When the prediction error value is obtained, the accumulated variance / prediction error value / complexity index measurement value is updated accordingly.

(A) Complexity Index Calculation Processing by I-picture Complexity Index Calculation Unit 30 FIG. 13 shows an embodiment of a processing flow executed by the I-picture complexity index calculation unit 30.

As shown in this processing flow, the I-picture complexity index calculation unit 30 uses each frame of the input image input to the pre-processing unit 10 as a processing target to determine the quantization scale in macroblock units. By calculating the activity value of the macroblock, and using the sum within the frame as the variance of the input image and substituting it into the above equation,
A process for estimating the complexity index of the I picture is performed.

Here, the macroblock activity is defined as the minimum value of the variance of four 8 × 8 pixel blocks in a 16 × 16 macroblock. The variance value of the 8 pixel × 8 pixel block used at this time is the sum of the absolute value of the difference between the average value of the luminance value in the 8 pixel × 8 pixel block and the luminance value of each pixel in the block. It is.

The quantization scale of a macroblock has been raised or lowered based on the activity value of the macroblock. Therefore, a mechanism for obtaining the activity value of the macroblock is separately provided. May be prepared, in such a case,
The activity value of the macroblock obtained by the mechanism is received, and processing is performed to obtain the sum within the frame.

(B) P / B picture complexity index calculating section 3
1 for calculating a complexity index among frames of an input image sequence encoded as a P-picture or a B-picture.
Then, a motion search with one-pixel accuracy is performed using the input image of the encoding target frame and the input image of the reference frame. Here, a motion search on a field basis and a motion search on a frame basis are performed between the encoding target frame and the input image of the reference frame.

In this one-pixel accuracy search, a total of four P-pictures, starting from the first and second fields of the encoding target frame and ending with the first and second fields of one reference frame, are obtained. Search for two types of field motion vectors and starting from the frame to be encoded,
A search for one frame motion vector ending at one reference frame is performed.

For the B picture, a total of eight types of field motion vectors starting from the first and second fields of the encoding target frame and ending at the first and second fields of two reference frames. And a search for two frame motion vectors starting from the encoding target frame and ending at two reference frames.

FIG. 14A shows a field motion vector searched for a B picture, and FIG. 14B shows a frame motion vector searched for a B picture. It is. In the case of a P picture, a search for a field motion vector is performed as shown in FIG.
Reference frame A shown in (a) and v1, v2, v3, v
In the search for the frame motion vector, only the reference frames A and v9 shown in FIG. 14B exist.

The P / B picture complexity index calculator 31 calculates
A final prediction error is calculated from the prediction errors corresponding to these motion vectors, and the calculated prediction error is substituted into the above-described expression or expression (or expression or expression), thereby estimating the complexity index of the P picture or the B picture. Perform the following processing.

FIGS. 15 and 16 show an embodiment of the processing flow of the P picture complexity index estimation processing executed by the P / B picture complexity index calculating section 31, and FIG.
FIG. 18 shows an example of the processing flow of the estimation process of the complexity index of the B picture executed by the P / B picture complexity index calculation unit 31.

The P / B picture complexity index calculator 31 calculates
When estimating the complexity index of the P picture, as shown in the processing flows of FIGS. 15 and 16, the smaller one of the prediction errors of v1 and v2 shown in FIG.
The sum of v3 and the smaller value of the prediction error of v4 shown in (a) is calculated, and the smaller value of the prediction error value of v9 shown in FIG. The sum within the frame is taken as the error, and the complexity index of the P picture is estimated as the prediction error value of the frame.

On the other hand, the P / B picture complexity index calculating section 3
When estimating the complexity index of the B picture, first, as shown in the processing flow of FIGS. 17 and 18, first, the smaller value of the prediction error between v1 and v2 shown in FIG. And the sum of the smaller values of v3 and v4 of the prediction error shown in FIG. 14 (a) is calculated. Then, the smaller value of the prediction errors of v5 and v6 shown in FIG. , FIG.
(A) finds the sum of v7 and v8, which is the smaller value of the prediction error, and then calculates the smaller value of the two sums, which is used as the field prediction error value for the macroblock. .

Further, v9 and v1 shown in FIG.
0 is determined as the frame prediction error value for the macroblock, and the smaller of the field prediction error value and the frame prediction error value is determined as the prediction error for the macroblock. Then, the complexity index of the B picture is estimated using the result as a prediction error value of the frame.

In this way, the I-picture complexity index calculating section 30 obtains the intra-frame sum of the activity values of the macroblocks of the input image input to the preprocessing section 10, and substitutes the sum into a quadratic equation. By doing so, a process of estimating the complexity index of the I picture is performed.

Then, the P / B picture complexity index calculating section 31 obtains a prediction error value obtained from the motion vector search result of the primary search section 11, and substitutes it into a quadratic expression, thereby obtaining a P picture And the complexity index of the B picture are estimated.

Here, regarding the coefficients of the quadratic expression used for estimating the complexity index of each picture, the variance value and the prediction error value of the past several frames and the frame generation after actually encoding are calculated for each picture type. From the data of the value of the complexity index obtained by the product of the code amount and the average quantization parameter,
Use the coefficients obtained by a method such as linear regression. However,
The method of calculating the coefficient from the data is not limited to this.

When a scene change is detected, the accumulated past frame data is cleared.
The n-th order coefficient of each picture type is also set to a predetermined initial value.

However, the first P-picture after the scene change is coded as an intra-macroblock at a high rate. The complexity index is estimated in the same way as the estimation method of.

Although the present invention has been described with reference to the illustrated embodiment, the present invention is not limited to this. For example,
In the embodiment, the configuration using the quadratic expression for estimating the complexity index is employed, but the present invention is not limited to the quadratic expression.
It may be a linear expression, or may be an n-th order expression of a third order or higher.

[0087]

As described above, according to the present invention,
Since the complexity index of the frame to be encoded can be estimated in advance, the inter-frame code amount distribution and the quantization control within the frame can be performed more efficiently. It is possible to improve the efficiency of gasification.

[Brief description of the drawings]

FIG. 1 shows a first embodiment of the present invention.

FIG. 2 is an explanatory diagram of a complexity index estimation process in the first embodiment.

FIG. 3 is an overall processing flow of a complexity index estimating process in the first embodiment.

FIG. 4 is a processing flow of a complexity index estimation process in the first embodiment.

FIG. 5 is an explanatory diagram of a motion vector searched in a primary search.

FIG. 6 is a processing flow of a complexity index estimation process according to the first embodiment.

FIG. 7 is a processing flow of a complexity index estimation process in the first embodiment.

FIG. 8 is a processing flow of a complexity index estimation process in the first embodiment.

FIG. 9 shows a second embodiment of the present invention.

FIG. 10 is an explanatory diagram of a process of estimating a complexity index in the second embodiment.

FIG. 11 is an overall processing flow of processing for estimating a complexity index in the second embodiment.

FIG. 12 is an overall processing flow of estimation processing of a complexity index in the second embodiment.

FIG. 13 is a processing flow of a complexity index estimation process in the second embodiment.

FIG. 14 is an explanatory diagram of a motion vector searched in the primary search.

FIG. 15 is a processing flow of a complexity index estimation process in the second embodiment.

FIG. 16 is a processing flow of a complexity index estimation process in the second embodiment.

FIG. 17 is a processing flow of a complexity index estimation process in the second embodiment.

FIG. 18 is a processing flow of a complexity index estimation process in the second embodiment.

[Explanation of symbols]

 Reference Signs List 10 preprocessing unit 11 primary search unit 12 secondary search unit 13 DCT unit 14 quantization unit 15 variable length coding unit 16 inverse quantization unit 17 inverse DCT unit 18 quantization parameter control unit 20 I picture complexity index calculation unit 21 P · B picture complexity index calculation unit 30 I picture complexity index calculation unit 31 P · B picture complexity index calculation unit

 ──────────────────────────────────────────────────続 き Continued on the front page F-term (reference)

Claims (12)

[Claims] 1. A variance value of an encoding target frame of an input image and a prediction error value of a motion vector search result between the encoding target frame of the input image and a reference frame are converted to a subsequent encoding process. The frame complexity index when performing intra-frame encoding is estimated from such a variance, and the frame complexity index when performing inter-frame encoding is estimated from such a prediction error value. An image coding method characterized in that coding control is performed using the estimated value of the complexity index. 2. A variance value of an encoding target frame of an input image and a prediction error value of a motion vector search result between the encoding target frame of the input image and a reference frame are converted to a subsequent encoding process. The frame complexity index in the case of performing the intra-frame encoding, which is obtained at least one frame earlier, is approximately estimated by the n-th order expression of the variance (where n is an integer of 1 or more), and the inter-frame encoding is performed. An image encoding method characterized in that the frame complexity index in the case is approximately estimated by the n-th order expression of the prediction error value, and encoding control is performed using the estimated value of the complexity index. 3. The image encoding method according to claim 2, wherein a fixed value set in advance is used as the coefficient of the n-th order expression. 4. The image encoding method according to claim 2, wherein a relationship between a variance value and a prediction error value of the input image and a value of a frame complexity index obtained from an encoding result is determined for each picture type. An image encoding method characterized by storing a plurality of frames in the past, and calculating a coefficient of the above-mentioned n-order expression for each picture type from the stored data. 5. The image encoding method according to claim 4, wherein, when a scene change is detected, the stored data is cleared, and the coefficient of the n-th order expression is returned to an initial value. Image encoding method. 6. The image encoding method according to claim 1, wherein, when a scene change is detected, an I picture to be encoded for the first time after the scene change is distributed. For a P picture to estimate a frame complexity index from a value and to encode for the first time after a scene change, a frame complexity index is estimated from the variance value of an input image according to the same procedure as for an I picture, and a B to be encoded for the first time after a scene change.
For a picture, an image coding method characterized by estimating a frame complexity index from a prediction error value of a motion vector search result. 7. The image coding method according to claim 1, wherein a sum of absolute values of adjacent pixel differences in a frame is used as a variance of the input image. Image coding method. 8. The image encoding method according to claim 1, wherein a sum of activity values of each macroblock in a frame is used as a variance value of an input image. Image coding method. 9. A variance value of an encoding target frame of an input image, and a prediction error value of a motion vector search result between the encoding target frame of the input image and a reference frame, for a subsequent encoding process. Means for obtaining one frame or more in advance; and estimating a frame complexity index for performing intra-frame encoding from the variance value, and estimating a frame complexity index for performing inter-frame encoding from the prediction error value. And a means for performing coding control using the estimated value of the complexity index. 10. A variance value of an encoding target frame of an input image and a prediction error value of a motion vector search result between the encoding target frame of the input image and a reference frame are calculated for a subsequent encoding process. Means for obtaining one or more frames ahead of time; approximation of a frame complexity index for performing intra-frame encoding by an n-th order expression of the variance (where n is an integer of 1 or more); Means for approximately estimating the frame complexity index by performing the n-th order expression of the prediction error value when performing, and means for performing coding control using the estimated value of the complexity index. Image coding device. 11. An image encoding processing program for causing a computer to execute processing used to implement the image encoding method according to claim 1. Description: 12. A recording medium for an image encoding processing program which stores a program for causing a computer to execute processing used for implementing the image encoding method according to claim 1. Description: