JP2002218464A - Method, device and program for performing transcoding of moving image coded data and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a transcoding device having little possibility that image quality deteriorates. SOLUTION: An inverse quantizing part 3 returns a bit stream generated from moving image data by encoding processing including a quantization process to unquantized data. A quantizing part 4 requantizes output data of the inverse quantizing part 3 with a quantization scale corresponding to a quantization scale code indicated by a CPU 7. The CPU 7 designates a threshold as a quantization scale code for requantization with respect to data subjected to quantization processing corresponding to a quantization scale code being smaller than the threshold among pieces of data included in the bit stream. The CPU 7 also changes thresholds in response to the temporal change of encoding difficulty of an object to be processed.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method, an apparatus, a program, and a recording medium for recording transcoding of encoded video data.

[0002]

2. Description of the Related Art As means for converting the bit rate of encoded video data, the encoded video data is decoded to generate video data before encoding, and this video data is converted to the original video data. There is a transcoding device that re-encodes data having a smaller bit amount than encoded data to generate encoded video data having a different bit rate.

However, in such a transcoding device, an inverse DCT (Discrete Cosine Tr
moving image data is decoded through an irreversible process such as an ansform (discrete cosine transform) and inverse quantization, and this moving image data is further converted into coded video data through an irreversible process such as DCT and quantization. There is a problem of deterioration.

As means for solving the above problems, various transcoding devices have been provided. For example,
MPEG (Moving Pictu), a well-known encoding standard
re Experts Group), the quantized data of the DCT coefficient included in the encoded video data is inversely quantized and converted back to the DCT coefficient.
In some cases, CT coefficients are requantized on a coarser quantization scale than at the time of encoding to obtain encoded video data having a smaller bit amount than the original encoded video data.

[0005]

By the way, at the time of encoding a moving image, a DC obtained from each picture constituting the moving image is used.
The T coefficient is not always quantized on the same quantization scale. Some DCT coefficients are quantized on a large quantization scale, while others are quantized on a small quantization scale.
There are also coefficients. However, in the above-described conventional transcoding device, the DCT coefficient is re-quantized by uniformly increasing the quantization scale. Because of this,
DCT coefficients that have been quantized on a large quantization scale are re-quantized on a coarser quantization scale, and there is a problem in that image quality deteriorates accordingly.

[0006] The present invention has been made in view of the above circumstances, and it is possible to perform transcoding with a low possibility of image quality degradation when performing re-encoding by converting a quantization scale. It is an object to provide a method, an apparatus, a program, and a recording medium on which the program is recorded.

[0007]

According to the present invention, there is provided an inverse quantization step of returning first encoded data generated from moving image data to data before quantization by an encoding process including a quantization step; A re-encoding step of including a re-quantization step of quantizing data before quantization, and generating second encoded data corresponding to the moving image data using data obtained by the re-quantization step; , Among the first encoded data,
For those having undergone a quantization process corresponding to a quantization scale code smaller than a threshold value, a quantization scale for changing the quantization scale code of the quantization scale in the requantization process applied to the encoded data to the threshold value Transcoding, comprising: a changing step; and a threshold changing step of changing the threshold according to a temporal change in the degree of difficulty in encoding an image corresponding to data to be processed in the re-encoding step. A method, a program for causing a computer to execute each procedure constituting the method, and a recording medium on which the program is recorded are provided.

Further, the present invention provides an inverse quantization means for returning the first encoded data generated from the moving picture data by the encoding process including the quantization process to the data before the quantization, A re-encoding unit for re-quantizing the data, the re-encoding unit using the data obtained by the re-quantization unit to generate second encoded data corresponding to the moving image data; Of the coded data of which has undergone the quantization process corresponding to the quantized scale code smaller than the threshold, the quantized scale code of the quantized scale in the re-quantizing means applied to the coded data is Quantization scale changing means for changing to a threshold,
A transcoding device comprising: a threshold changing unit that changes the threshold according to a temporal change in the degree of difficulty in encoding an image corresponding to data to be processed by the re-encoding unit. .

[0009]

Embodiments of the present invention will be described below with reference to the drawings. A: Configuration of Embodiment FIG. 1 is a block diagram illustrating a configuration of a transcoding device 100 according to an embodiment of the present invention. This transcoding device 100
This is a device that converts the bit rate of a bit stream (hereinafter, simply referred to as a bit stream) obtained by a moving image compression encoding method based on PEG1. As shown in FIG. 1, the transcoding device 100 includes an input buffer memory 1, a variable length decoding unit 2, and an inverse quantization unit 3.
, A quantization unit 4, a variable length coding unit 5, an output buffer memory 6, a CPU 7, a memory 8, a CPU bus 9
And a control panel 10.

Here, the input buffer memory 1 is a memory for storing a bit stream to be processed by the transcoding device 100. FIG. 2 schematically shows the structure of the bit stream input to the input buffer memory 1. As shown in FIG. 2, the bit stream has a six-layer structure including a GOP (Group Of Picture; picture group), a picture, a slice, a macroblock, and a block. Hereinafter, data corresponding to each of these layers will be described.

A sequence is a set of time-series pictures constituting a moving image. The bit stream includes sequence layer data corresponding to this sequence as the highest layer data. The data of the sequence layer includes data specifying the size (horizontal direction, vertical direction), pixel aspect ratio, picture rate (time resolution), etc. of each picture belonging to the sequence.

The sequence layer data includes GOP layer data corresponding to a plurality of GOPs constituting this sequence. One GOP is composed of a series of pictures as illustrated in FIG.

There are three types of pictures constituting the GOP: an intra picture (I picture), a forward prediction picture (P picture), and a bidirectional prediction picture (B picture). Among these, the I picture is a picture on which intra-picture encoding is performed without referring to other pictures.
The P picture and the B picture are pictures to be subjected to inter-picture predictive encoding with reference to another picture. G
In the data of the OP layer, the data of the picture layer corresponding to each picture constituting the GOP as described above is inserted.

One picture is composed of a plurality of slices as shown in FIG. Correspondingly, slice layer data corresponding to each slice constituting the picture is inserted in the picture layer data. The data of the slice layer corresponding to each slice includes data specifying a quantization scale to be applied to the slice.

One slice is composed of a plurality of macroblocks each consisting of 16.times.16 pixels. In the slice layer data, macro block layer data corresponding to each macro block constituting the slice is inserted.

Here, data of a macroblock layer corresponding to a P picture or a B picture includes, for each macroblock, data of a reference picture in another picture referred to for inter-picture predictive coding of the macroblock. It may include motion information specifying the position.

In the data of the macro block layer corresponding to each macro block, the data of the block layer corresponding to each block constituting the macro block is inserted. Here, the data of the block layer corresponding to the I picture is obtained by quantizing a plurality of DCT coefficients obtained from each pixel value in the block and performing variable length coding for each block. It contains variable length encoded data. Further, the data of the block layer corresponding to the P picture or the B picture is obtained by quantizing DCT coefficients corresponding to difference information between the block and the reference image for each block and performing variable length coding. It contains the obtained variable-length coded data.

The above is the details of the configuration of the bit stream to be processed by the transcoding device 100 according to the present embodiment. This bit stream is read from the input buffer memory 1 under the control of the CPU 7,
It is supplied to the variable length decoding unit 2.

The variable length decoding unit 2 is a device for converting variable length coded data in the bit stream supplied in this way to fixed length data. The bit stream that has been processed by the variable-length decoding unit 2 includes quantized data obtained by quantizing DCT coefficients corresponding to each block constituting a picture.

The inverse quantization unit 3 inversely quantizes these quantized data in the bit stream, and calculates each D before the quantization.
This is a device for returning to the CT coefficient.

The quantization unit 4 is a device that re-quantizes the DCT coefficients thus obtained in accordance with a command from the CPU 7. The command from the CPU 7 to the quantization unit 4 includes a quantization scale code for specifying the quantization scale. These quantization scale codes are “1” to “3”.
1 ”. The quantization unit 4 requantizes the DCT coefficients using a quantization scale corresponding to the quantization scale code specified by the CPU 7. Here, the larger the quantization scale code, the coarser the quantization scale, and the smaller the code amount of the quantized data obtained when requantization is performed using the quantization scale.

The variable length coding unit 5 is a device for performing variable length coding on the quantized data obtained by requantization.

The output buffer memory 6 is a memory for storing the bit stream obtained by the above processing.
The bit stream stored in the output buffer memory 6 is transmitted to another device as output stream data.

The control panel 10 is a device for receiving a command input for designating the operation of the transcoding device 100.

The CPU 7 is a control center for controlling each part of the transcoding device 100. The main control functions of the CPU 7 are as follows.

A. Data distribution control between the elements The CPU 7 sequentially takes out the bit stream to be processed from the input buffer memory 1 and delivers it to the variable length decoding unit 2, and delivers the output data of the variable length decoding unit 2 to the inverse quantization unit 3. The output data of the inverse quantization unit 3 is transferred to the quantization unit 4, the output data of the quantization unit 4 is transferred to the variable length coding unit 5, and the output data of the variable length coding unit 5 is written to the output buffer memory 6.

B. Processing for Specifying the Quantization Scale Used for Requantization by the Quantization Unit 4 The bit stream to be processed is a quantum that specifies the quantization scale used for quantization of the DCT coefficient corresponding to each part of the video sequence. Includes a scale code. The CPU 7 requantizes the quantized scale code of the original quantized scale code included in the bit stream, which is equal to or larger than a threshold value (lower limit) calculated in advance, as a quantized scale code for requantization. Part 4
Specify for. For the quantized scale code smaller than the above threshold value among the original quantized scale codes, the threshold value is designated as the quantized scale code for requantization instead of the original quantized scale code.

C. Threshold control of quantization scale code In the process of the data distribution control, the CPU 7 obtains a parameter necessary for determining the threshold,
The threshold of the quantization scale code is calculated from these parameters.

B: Operation of the Embodiment FIGS. 5 to 12 are flowcharts showing the contents of processing executed by the CPU 7 in this embodiment. Hereinafter, the operation of the present embodiment will be described with reference to these flowcharts.

Before starting the processing of the bit stream in the input buffer memory 1, the CPU 7 obtains the file size (total bit amount) of the bit stream (step S1). Next, the CPU 7 initializes each variable required for transcoding control (step S).
2). When the above processing is completed, the CPU 7 starts processing the bit stream in the input buffer memory 1 (step S3).

In this step S3, the processing of the sequence layer whose flow is shown in FIG. 6 is executed. First, the CPU 7
For example, the target conversion rate tr_r is obtained by a method such as accepting an input from a user through the control panel 10.
ate is obtained (step S101). This target conversion rate t
r_rate is a ratio of the bit amount of the bit stream to be obtained by re-encoding to the bit amount before re-encoding.

Next, the CPU 7 reads the bit stream in the input buffer memory sequentially from the head. And
If you find the sequence header in the bitstream,
Processing of the sequence header such as writing of the sequence header into the output buffer memory 6 is performed (step S
102).

Next, the CPU 7 reads out data bit_rate_value specifying the bit rate of the bit stream from the input buffer memory 1 (step S103), multiplies the data by "400", and sets the bit rate of the bit stream to be processed. Find bit_rate (Step S
104).

Next, the CPU 7 compares the bit rate bit_rate obtained in step S103 with step S103.
Multiply by the target conversion rate tr_rate obtained in 101,
A target conversion bit rate tr_bit_rate which is a bit rate of the bit stream after re-encoding is obtained (step S).
105).

Thereafter, the CPU 7 continues reading the bit stream from the input buffer memory 1 until a GOP start code indicating the start of data in the GOP layer is found (step S106). Then, the CPU 7
When the OP start code is read, the processing of the GOP layer shown in FIG. 7 is executed (step S107).

In the processing of the GOP layer, the CPU 7
First, the GOP header corresponding to the first GOP is read from the input buffer memory 1, and the GOP header is processed such as writing the GOP header to the output buffer memory (step S201). Next, the CPU 7 obtains a bit allocation amount r_add per GOP by an operation shown in the following equation. r_add = tr_bit_rate × N / picture_rate (1) In the above equation, tr_bit_rate is the target conversion bit rate obtained in step S105, N is the number of pictures per GOP, and picture_rate is the number of pictures per second.

Next, the CPU 7 obtains the permissible amount capa by the calculation shown in the following equation (step S203). capa = fsize / 10000 (2) where fsize is the file size of the input bit stream.

The permissible amount capa indicates that the bit amount of the bit stream after re-encoding is equal to the bit allocation amount r_GOP per GOP.
This is the permissible amount when the bit amount obtained from add exceeds the bit amount. More specifically, in the present embodiment, basically, the bit amount of the bit stream after re-encoding is the bit amount obtained from the bit allocation amount r_add per GOP (for example, when the k-th GOP from the top is processed, (k × r_add), control of the quantization scale at the time of requantization is advanced. However, if the coarse quantization scale causes deterioration of image quality, the quantization scale is not coarsened and the bit amount of the re-encoded bit stream is 1 G within a certain allowable range.
It is allowed to exceed the bit amount obtained from the bit allocation amount per OP r_add. This is the allowance capa.

Next, the CPU 7 obtains a margin R (step S204). This calculation is performed as follows. That is, assuming that data to be processed corresponds to, for example, the k-th GOP, the target bit amount of the re-encoded bit stream at the time when the re-encoding corresponding to the (k-1) -th GOP ends. From the target conversion bit rate tr_bit_rate. Next, before G
The margin R is obtained by subtracting the bit amount of the re-encoded bit stream at the time when the re-encoding corresponding to the OP, that is, the (k-1) -th GOP is completed, from the target bit amount. Next, the CPU 7 adds the bit allocation amount r_add per GOP to the margin R at the end of the re-encoding corresponding to the (k-1) -th GOP, and outputs the result as the margin R corresponding to the k-th GOP. _Let it be _k (step S204). The calculation of the margin R can be represented by an equation. R _k = R _k−1 −B _k−1 + r_add (3) In the above equation (3), R _k is a margin for the k-th GOP, and B _k is a code amount of the k-th GOP. .

Next, the CPU 7 calculates the threshold value qth of the quantization scale code applied to the data corresponding to the k-th GOP to be processed (step S205). The details of this calculation will be described later.

Thereafter, the CPU 7 continues until a picture start code indicating the start of the picture layer data is found.
The reading of the bit stream from the input buffer memory 1 is continued (step S206). And CPU7
Reads the picture start code and executes the processing of the picture layer shown in the flow of FIG. 8 (step S20)
7). When the processing corresponding to the first GOP is completed in this way, the processing of steps S201 to S207 described above is repeated for the second and subsequent GOPs.

Next, the processing of the picture layer will be described with reference to FIG. In the processing of this picture layer, the CP
U7 first reads a picture header corresponding to the first picture from the input buffer memory 1, and performs a picture header process such as writing the picture header to the output buffer memory (step S301).

Thereafter, the CPU 7 continues until the slice start code indicating the start of the slice layer data is found.
The reading of the bit stream from the input buffer memory 1 is continued (step S302). And CPU7
Reads out the slice start code, and executes the slice layer processing shown in the flow of FIG. 10 (step S3).
03). When the processing corresponding to the first picture is completed in this way, the processing of steps S301 to S303 described above is repeated for each of the second and subsequent pictures.

Next, the processing of the slice layer will be described with reference to FIG. In processing this slice layer, the CP
U7 first reads a slice header corresponding to the first slice from the input buffer memory 1, and performs slice header processing such as writing the slice header to the output buffer memory (step S401).

Next, the CPU 7 extracts a quantized scale code q_scale_code from the slice header (step S402). The quantization scale code extracted from the slice header specifies the quantization scale used to obtain data corresponding to each slice following the slice header.

Next, the CPU 7 determines whether or not the quantized scale code q_scale_code extracted from the slice header in step S402 is smaller than the threshold value qth obtained in the processing of the GOP layer (FIG. 7) (step S403). . If the result of this determination is “NO”, the CP
U7 is a quantization scale code aq_scale_c for requantization.
Let ode be the quantization scale code q_scale_code extracted from the slice header (step S404). On the other hand, if the result of this determination is “YES”, the CPU 7
Is the quantization scale code for requantization aq_scale_code
Is set to the threshold value qth (step S405).

As described above, of the quantized scale code q_scale_code extracted from the slice header, the threshold value qth
The above (that corresponds to a coarse quantization scale exceeding a certain limit) is used as it is as a quantization scale code aq_scale_code for requantization. On the other hand, those smaller than the threshold qth (that is, those corresponding to fine quantization scales beyond a certain limit)
Threshold instead of its quantized scale code q_scale_code
qth is the quantization scale code for requantization aq_scale_code
Used as Therefore, if the threshold value qth is increased, the bit amount of the re-encoded bit stream is reduced accordingly.

As described above, the quantization scale code aq_scale_code for requantization to be applied to the first slice
Is determined, the CPU 7 executes the processing of the macro block layer whose flow is shown in FIG. 10 (step S406). When the processing corresponding to the first slice is completed as described above, the processing of steps S401 to S406 described above is repeated for the second and subsequent slices.

Next, the processing of the macroblock layer will be described with reference to FIG. The data of the macroblock layer may include a quantized scale code. If such a quantized scale code is included in the data of the macroblock layer corresponding to a certain macroblock, each macroblock after that macroblock until a new quantized scale code is defined The corresponding encoded moving image data is quantized by the quantization scale corresponding to the quantization scale code.

Therefore, in the processing of the macro block layer, C
PU7 performs such a quantization scale code q_scale_co
It is determined whether or not de is included in the data of the macroblock layer (step S501).

If the result of this determination is "NO",
The block layer processing is executed using the latest quantized scale code q_scale_code obtained at that time (step S505).

On the other hand, if the decision result in the step S501 is "YES", the quantization scale code q_scale_code in the data of the macroblock layer is smaller than the threshold value qth obtained in the processing of the GOP layer (FIG. 7). It is determined whether it is smaller (step S502).

If the result of this determination is “NO”, the CP
U7 sets the quantized scale code q_scale_code in the data of the macroblock layer as the quantized scale code aq_scale_code for requantization (step S503). on the other hand,
If the result of this determination is “YES”, the CPU 7 sets the threshold value qth to a quantization scale code aq_scale_co
It is set to de (step S504).

Then, step S503 or S504
Quantization scale code for requantization aq_scale_cod
When e is determined, the CPU 7 executes the processing of the block layer (step S505).

In the processing of the block layer, the data of the block layer is returned to the data before the variable-length encoding by the variable-length decoding unit 2, and the data before the quantization is returned by the inverse quantization unit 3. Then, the quantizing unit 4 applies a quantization scale code aq_scale for requantization to the data before quantization.
Performs quantization using the quantization scale corresponding to _code,
The quantized data obtained as a result is subjected to variable-length encoding by the variable-length encoding unit 5 and written to the output buffer memory 6.

As described above, when the processing corresponding to the first macroblock is completed, the remaining macroblocks belonging to the same slice are also subjected to step S described above.
The processing of 501 to S505 is repeated.

Next, referring to FIG. 11 and FIG.
The calculation processing of the threshold value qth of the quantized scale code executed in step S205 of the processing of the P layer (FIG. 7) will be described. In the following, for convenience of explanation, the CPU 7
Performs a process corresponding to the k-th GOP from now on.

First, the CPU 7 calculates the average bit rate of all the re-encoded bit streams at the time when the re-encoding corresponding to the (k-1) -th GOP is completed. Then, it is determined whether or not this average bit rate is higher than the target conversion bit rate tr_bit_rate obtained in step S105 of the processing of the sequence layer (FIG. 6) (step S601).

If the result of this determination is "YES", that is, if the bit amount of the bitstream after re-encoding has been larger than expected so far,
The CPU 7 sets the bit rate control parameter qth_r to “1” (step S602). On the other hand, step S
If the determination result of step 601 is "NO", the bit rate control parameter qth_r is set to "0" (step S6).
03). This bit rate control parameter qth_r is used to calculate the average bit rate after re-encoding to the target conversion bit rate tr.
This is a parameter for performing control to approach _bit_rate, and is used to increase or decrease the quantization scale code used for requantization.

Next, the CPU 7 calculates the encoding difficulty (complexity) of the bit stream before re-encoding corresponding to the (k−1) -th GOP and the k−1 GOPs from the first to the (k−1) -th GOP. Then, the average encoding difficulty of the bit stream before re-encoding corresponding to is calculated (step S604). Here, the former encoding difficulty is k-
The average quantization scale code applied to the quantization of the data corresponding to the first GOP, and the (k-1) th GOP
Is multiplied by the bit amount of the bit stream corresponding to. The latter encoding difficulty is k
An average quantization scale code applied to quantization of data corresponding to (k−1) GOPs up to the first,
-1 is obtained by multiplying by the bit amount of the bit stream corresponding to one GOP.

Next, the CPU 7 determines whether or not the average encoding difficulty for all the GOPs so far is greater than the encoding difficulty corresponding to the immediately preceding (k-1) -th GOP (see FIG. 4). Step S605). The result of this determination is “YE
S ”, the CPU 7 determines the image quality adjustment parameter qth_
Let c be "2". Here, the image quality adjustment parameter qth_c
Is the quantization scale code threshold qt according to the change in image quality.
This parameter specifies the range of increase or decrease when h is increased or decreased.

If the decision result in the step S605 is "NO", the CPU 1 determines that the margin R is larger than "0",
Further, it is determined whether or not the image quality adjustment parameter qth_c applied to the previous GOP (k-1st GOP) is “2” (step S607). If the result of this determination is "YES"
If, the CPU 7 sets the image quality adjustment parameter qth_c to “−2” (step S608).

If the decision result in the step S607 is "NO", the CPU 7 decides whether or not the margin R is larger than the bit allocation amount r_add per GOP (step S609). If the determination result is “YES”, the CPU 7 sets the image quality adjustment parameter qth_c to “−1”.
(Step S610).

If the decision result in the step S609 is "NO", the CPU 7 judges whether or not the margin R is smaller than -capa, that is, the bit amount exceeds the target bit amount, and the excess is not permitted. It is determined whether or not the capacity is exceeded (step S611). When the determination result is “YES”, the CPU 7 determines the image quality adjustment parameter qth_c
Is set to “2” (step S612), and if “NO”, the image quality adjustment parameter qth_c is set to “0” (step S612).

As described above, the image quality adjustment parameter qth_
When c is obtained, the CPU 7 sets the image quality adjustment parameter q
Using th_c, the bit rate control parameter qth_r obtained in steps S601 to S603, and the threshold value qth of the current quantization scale code, the following expression is calculated to obtain the threshold value qth to be applied to the k-th GOP. (Step S614). qth = qth + qth_r + qth_c (4)

Next, when the threshold value qth obtained as described above is out of the range of quantization scale codes "1" to "31" usable in MPEG1 and MPEG2, it falls within the range. Is performed. That is, when the threshold value qth is greater than “31”, the CPU 7 sets qth = “31” (step S615,
S616) If the threshold value qth is smaller than “1”, qth =
"1" (steps S615, S617, S61)
8) In other cases, the threshold value qth obtained in step S614 is not changed (steps S615 and S615).
617, S619). The above is the processing of the GOP layer (FIG. 8)
15 is a detail of the calculation process of the threshold value qth of the quantized scale code executed in step S206 of FIG.

The above calculation processing is summarized as follows. (1) The threshold value qth of the quantized scale code is the average encoding difficulty level of each GOP up to the encoding difficulty level corresponding to the GOP immediately before the GOP currently being requantized. Is smaller than, the value is increased by “2” (step S606). Therefore, according to the present embodiment, the degree of difficulty in encoding an image can be quickly multiplied by the decrease, the quantization scale for requantization can be coarse, and the bit amount can be reduced.

(2) Although the degree of difficulty in encoding an image did not decrease in the previous GOP, the bit amount of the bit stream after re-encoding is smaller than the target bit amount, and the bit amount in the processing of the immediately preceding GOP is small. Scale code threshold
When qth is increased by “2”, the threshold value qth of the quantized scale code in the processing of the current GOP is decreased by “2” (step S608).

(3) The degree of difficulty in encoding has not decreased, and the shortage of the bit amount of the re-encoded bit stream from the target bit amount is larger than r_add, and quantization is performed in the immediately preceding GOP processing. Threshold qth of scale code is "2"
Otherwise, the threshold value qth of the quantized scale code in the processing of the current GOP decreases by “1” (step S610).

(4) If the degree of difficulty in encoding has not decreased and the excess of the bit amount of the re-encoded bit stream from the target bit amount exceeds the allowable amount capa, the current GOP Threshold qt of the quantized scale code in the processing of
h increases by "2" (step S612).

(5) If the average bit rate exceeds the target conversion bit rate, the threshold value qth of the quantized scale code in the processing of the current GOP is increased by "1" (step S602).

C: Effects of the Embodiment According to the embodiment described above, in the requantization processing of each data belonging to each GOP, the data is quantized by the quantization scale corresponding to the quantization scale code equal to or larger than the threshold value qth. The existing data is re-quantized by the original quantization scale. Therefore, it is possible to prevent a problem that data that has been originally quantized by the coarse quantization scale is quantized by the coarser quantization scale, thereby deteriorating the image quality. On the other hand, data that has been quantized by the quantization scale corresponding to the quantization scale code smaller than the threshold qth is requantized by the quantization scale corresponding to the threshold qth. Then, the threshold value qth of the quantized scale code is such that the coding difficulty corresponding to the GOP immediately before the GOP to be requantized is the average coding difficulty of each GOP up to that point. If it is smaller than the value, it is increased by “2”. Therefore, according to the present embodiment, the degree of difficulty in encoding an image can be quickly multiplied by the decrease, the quantization scale for requantization can be coarse, and the bit amount can be reduced.

D: Modifications While one embodiment of the present invention has been described above, the above-described embodiment is merely an example, and modifications can be made within the scope of the present invention. For example, the following modifications are conceivable.

(Modification 1) The latter half (FIG. 12) of the process of calculating the threshold value of the quantized scale code may be replaced with that shown in FIG. The processing illustrated in FIG. 13 does not include the processing of steps S609 to S613 in FIG.
The contents of the processing shown in FIG. 13 are summarized as follows.

(1) If the encoding difficulty corresponding to the GOP immediately before the GOP currently being requantized is smaller than the average encoding difficulty for each GOP up to then, Threshold qth of quantization scale code is "2"
(Steps S605 and S606).

(2) Although the degree of difficulty in encoding an image did not decrease in the previous GOP, the bit amount of the bit stream after re-encoding is smaller than the target bit amount, and the quantum amount in the processing of the immediately preceding GOP is small. Scale code threshold
When qth is increased by “2”, the threshold value qth of the quantized scale code in the processing of the current GOP is decreased by “2” (steps S605, S607, and S6).
08).

(3) Although the degree of difficulty in image coding did not decrease in the previous GOP, the bit amount of the bit stream after re-encoding is equal to or larger than the target bit amount, If the threshold value qth of the quantized scale code is not increased by “2”, the threshold value qth of the quantized scale code in the processing of this GOP is
Is increased by “2” (steps S605 and S607)
And S606). Also in the present modification, the threshold value of the quantization scale code can be increased by multiplying the decrease in the degree of difficulty in encoding, and the amount of bits after re-encoding can be reduced.

(Modification 2) In the above embodiment, the threshold value qth of the quantized scale code is controlled in accordance with the change of the code of the margin R and the degree of difficulty of encoding. May be divided into ranks, and the range of increase or decrease of the threshold may be changed depending on which rank the change amount of the margin R or the degree of difficulty of encoding belongs to.

(Modification 3) The method of obtaining the average encoding difficulty can be modified. For example, each past GO
The encoding difficulty corresponding to P may be averaged by multiplying by a weight coefficient. In this case, each weighting factor is
The weighting factor for multiplying the encoding difficulty corresponding to the OP is maximized, and the weighting factor for multiplying the encoding difficulty is reduced as going back in the past. The purpose is to consider that the relationship with the current GOP is weaker in the past GOP.

(Modification 4) In the above embodiment, GO
The threshold qth of the quantized scale code was controlled by calculating the degree of coding difficulty and the margin R for each P.
The threshold qth may be controlled for other layers according to G1 and MPEG2, for example, for each slice.

(Modification 5) The above embodiment may be applied to the re-encoding of a bit stream obtained by an encoding method based on MPEG2.

(Modification 6) The present invention is implemented in such a manner that the transcoding device described above is produced and sold, and a program for executing the transcoding method according to the present invention by a computer is read by a computer. The present invention can also be implemented by recording on a possible recording medium and distributing it to general users, or providing it to general users via an electric communication line.

[0083]

As described above, according to the present invention,
When re-encoding is performed to reduce the bit amount, there is an effect that the possibility that image quality deteriorates can be reduced.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a transcoding device according to an embodiment of the present invention.

FIG. 2 is a diagram showing a data structure of a bit stream to be processed in the embodiment.

FIG. 3 shows a GOP represented by the same bit stream.
FIG.

FIG. 4 is a diagram illustrating pictures, slices, macroblocks, and blocks represented by the same bit stream.

FIG. 5 is a flowchart showing the operation of the embodiment.

FIG. 6 is a flowchart showing the operation of the embodiment.

FIG. 7 is a flowchart showing the operation of the embodiment.

FIG. 8 is a flowchart showing the operation of the embodiment.

FIG. 9 is a flowchart showing the operation of the embodiment.

FIG. 10 is a flowchart showing the operation of the embodiment.

FIG. 11 is a flowchart showing the operation of the embodiment.

FIG. 12 is a flowchart showing the operation of the embodiment.

FIG. 13 is a flowchart showing an operation of a modification of the embodiment.

[Explanation of symbols]

1 ... input buffer memory, 2 ... variable length decoding unit, 3
... Inverse quantization section 4... Quantization section 5... Variable length coding section 6... Output buffer memory 7.
……Control panel.

Claims (4)

[Claims] An inverse quantization step of returning first encoded data generated from moving image data to data before quantization by an encoding process including a quantization step; and quantizing the data before quantization. A re-encoding step of generating second encoded data corresponding to the moving image data using the data obtained by the re-quantization step; and the first encoded data Of those, for those having undergone the quantization process corresponding to the quantization scale code smaller than the threshold value, the quantization scale code of the quantization scale in the requantization process applied to the encoded data is changed to the threshold value. A quantizing scale changing step, and a threshold changing step of changing the threshold value according to a temporal change in the degree of difficulty in encoding an image corresponding to data to be processed in the re-encoding step. A transcoding method characterized by the following. 2. An inverse quantization step of returning first encoded data generated from moving image data to data before quantization by an encoding process including a quantization step, and quantizing the data before quantization. A re-encoding step of generating second encoded data corresponding to the moving image data using the data obtained by the re-quantization step; and the first encoded data Of those, for those having undergone the quantization process corresponding to the quantization scale code smaller than the threshold value, the quantization scale code of the quantization scale in the requantization process applied to the encoded data is changed to the threshold value. A computer comprising: a quantization scale changing step; and a threshold changing step of changing the threshold value according to a temporal change in the degree of difficulty in encoding an image corresponding to data to be processed in the re-encoding step. A computer-readable recording medium on which a program to be executed by a computer is recorded. 3. An inverse quantization step of returning the first encoded data generated from the moving image data to the data before quantization by an encoding process including a quantization step, and quantizing the data before quantization. A re-encoding step of generating second encoded data corresponding to the moving image data by using data obtained by the re-quantization step; and the first encoded data Of those, for those having undergone the quantization process corresponding to the quantization scale code smaller than the threshold value, the quantization scale code of the quantization scale in the requantization process applied to the encoded data is changed to the threshold value. A computer comprising: a quantization scale changing step; and a threshold changing step of changing the threshold value according to a temporal change in the degree of difficulty in encoding an image corresponding to data to be processed in the re-encoding step. Program to be executed. 4. An inverse quantization means for returning first encoded data generated from moving image data to data before quantization by an encoding process including a quantization process, and quantizing the data before quantization. Re-encoding means for generating second encoded data corresponding to the moving image data using data obtained by the re-quantization means; and the first encoded data. Among them, for those having undergone the quantization processing corresponding to the quantization scale code smaller than the threshold, the quantization scale code of the quantization scale in the requantization means applied to the encoded data is changed to the threshold. Quantization scale changing means, and threshold changing means for changing the threshold in accordance with a temporal change in the degree of difficulty of encoding an image corresponding to data to be processed by the re-encoding means. A transcoding device characterized by the following.