KR20030083109A - apparatus for transcoding bit rate - Google Patents

Abstract

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a bit rate conversion encoding apparatus, comprising: a bit rate conversion encoding apparatus having a variable length decoding unit, a first addition unit, and a quantization unit, the target bit amount which can be allocated to a current frame output from the variable length decoding unit Estimating a bit amount and a distortion value generated according to a quantization parameter applied in units of macroblocks from a target bit amount determiner, a target bit amount determined by the target bit amount determiner, and a difference signal output from the first adder. A target distortion determiner for determining a target distortion of the entire n-th frame currently transform-coded from the signal output from the target bit rate distortion determiner and the signal output from the target bit rate distortion determiner; The bit rate-distortion curve for the k-th macroblock according to the bit rate and distortion value output from the bit rate-distortion estimator And a quantization parameter determiner for determining a quantization parameter for generating a distortion close to the target distortion output from the target distortion determiner and outputting the quantization parameter to the quantization unit. There is an effect that can be greatly reduced.

Description

Apparatus for transcoding bit rate

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a bit rate conversion encoding apparatus, and more particularly, to a bit rate conversion encoding apparatus capable of greatly reducing memory and computation amount while performing the same operation as the conventional bit rate conversion encoding apparatus.

It is well known that digital video has an excessive amount of data compared to the capacity of a storage medium or a transmission medium. Therefore, a digital source signal is compressed using a technique such as MPEG, which is a typical compression method.

The broadcast studio keeps the original signal for later use. However, if you store the bit stream instead of the original signal, you can save tens of times. Recently, there is a lot of research to use this bit string as it is without using the original signal. Transcoding of an MPEG bit string having a specific bit rate to another bit rate is called transcoding. .

In the case of digital broadcasting, when a broadcaster encodes a program in MPEG-2, this bit string can be transmitted through various transmission media. The digital bit string only needs to satisfy a bit rate that a medium can transmit. For example, MPEG-2 bit strings for satellite broadcasting can be applied everywhere, such as terrestrial broadcasting, VOD service, and program supply to the headend of each CATV through satellite. In this case, a bit rate conversion encoder for changing the bit rate according to the transmission capability of each medium is required.

Broadcasting contents stored in storage media after being broadcasted once will also be frequently re-broadcasted through other transmission media. Even in this case, the transmission bit rate is likely to be different. For example, a video signal encoded and stored at 6 Mbps is recorded on a 3 Mbps digital video disc (DVD), applied to a 3 Mbps VOD service, or Internet broadcasting between tens of Kbps and 1 Mbps. As such, since the transmission bit rate varies depending on the application target, a bit rate conversion encoding apparatus is required.

Hereinafter, a bit rate conversion encoding apparatus according to the prior art will be described.

1 is a view showing an example of a bit rate conversion encoding apparatus according to the prior art. As shown, a bit rate conversion encoding apparatus 2 is present between an MPEG decoder 1 and an MPEG encoder 3, The basic form of the bit rate conversion encoding apparatus 2 is simply a cascade of the decoding unit 2a and the encoding unit 2b. The decoding unit 2a encodes the bit rate R1 in the MPEG encoding unit 1. The encoded video is reproduced as a video corresponding to the bit rate R1. The coder 2b encodes the video reproduced at the bit rate R1 in the decoder 2a at the bit rate R2. The MPEG decoder 3 encodes the code. The video encoded by the bit rate R2 in the section 2b is reproduced as a video corresponding to the bit rate R2.

At this time, the coder 2b controls the quantization parameter so that the generated bit rate becomes R2. Where R1> R2.

FIG. 2 is a diagram showing a detailed configuration of the bit rate conversion encoding apparatus of FIG. 1, wherein the decoding unit 2a of the bit rate conversion encoding apparatus 2 has the same structure as that of the MPEG decoding unit 3. As shown in FIG. The input bit string encoded at the bit rate R1 in the MPEG coder 1 is decoded to a variable length in a variable length decode unit VLD 2a-1 and then decoded from the first dequantization unit DQ1. In -2), inverse quantization is performed according to the quantization parameter QP1. In this case, the quantum and the parameter QP1 used in the dequantization process are included in the input bit string R1 and transmitted. This value may be changed in units of macroblocks (MBs).

Discrete Cosine Transform (DCT) coefficients obtained by the inverse quantization are transformed into a pixel region by a first Inverse Discrete Cosine Transform (IDCT, 2a-3). Pixel value ( ) Corresponds to the difference between the pixel of the current input frame and the pixel of the previous frame whose motion compensation is performed. This difference ( ) Is a previous frame image (stored in the first image memory units MEM1 and 2a-4). ) Is a value of a pixel whose motion compensation is performed by the first moving compensator (MC1, 2a-6) ) And the first adder 2a-4 to add an image (for the current frame) ) Will be played. At this time, the motion vector used for the motion compensation ( ) Is transmitted through the bit string R1. Here, the motion compensation is performed in units of macroblocks (MBs), and this process is performed by using a motion vector (MacroBlock: MB) at the position of a macroblock (MB) currently decoded. This is a process of extracting a macroblock (MB) of a previous frame shifted by).

Therefore, the image reproduced by the decoder 2a may be expressed as in Equation 1 below.

Where operator Is the motion vector at the position of the macroblock (MacroBlock: MB) that is currently decoded for the previous frame. Performs the operation of extracting the macroblock (MacroBlock: MB) at the position shifted by).

In addition, the code unit 2b of the bit string encoding apparatus 2 has the same structure as that of the MPEG code unit 1, and as shown, the output of the second adder 2b-1. Is represented by Equation 2 below, the image reproduced by the decoder (2a) ( ) From the previous frame image stored in the second image memory unit (MEM2, 2b-9) ) Is a pixel value of motion compensation (Moving Compensation 2: MC2, 2b-10). ) By dividing

The difference signal thus obtained The discrete cosine transform (DCT, 2b-2) is discrete cosine processed to be quantized according to the quantization parameter (QP2) in the quantization unit (Quantization: Q, 2b-3) and then the variable length coding unit (Variable Length) Coding: converted into a bit string by VLC (2b-4), and then stored in the buffer (Buf, 2b-5) and transmitted at the transmission bit rate R2.

On the other hand, the DCT coefficients quantized by the quantization unit (Quantization: Q, 2b-3) are dequantized by the second dequantization unit (DeQuantization2: DQ2, 2b-6) for motion compensation processing, and then the second inverse discrete cosine transform unit. Pixel value compensated for by motion in the third adder 2b-8 after inverse discrete cosine in (Inverse Discrete Cosine Transform2: IDCT2, 2b-7) ) Is stored in the second image memory units MEM2 and 2b-9.

Since the loss of information during the discrete cosine transform and quantization process occurs only in the quantization process, the output of the second inverse discrete cosine transform unit (IDCT2, 2b-7) is the discrete cosine transform unit (DCT). , 2b-2) Quantization noise () Equal to the sum of). Finally, the image (stored in the second image memory units MEM2 and 2b-9) ) Is the same image as the output image of the MPEG decoder 3, and this value is represented by the following equation (3), and the output of the decoder 2a ( Quantization noise () Equal to the sum of).

Operator meaning motion compensation in the above Is a linear operator, so substituting Equation 3 into Equation 2 results in a difference signal. Can be expressed as Equation 4 below.

In addition, substituting Equation 1 into Equation 4 indicates a difference signal. Can be expressed as Equation 5 below.

Accordingly, the bit rate conversion encoding apparatus as shown in FIG. 2 can be simplified to the bit rate conversion encoding apparatus as shown in FIG.

In addition, the discrete cosine transform unit (DCT, 2b-2) is a linear transformation, and the result of the discrete cosine transform unit (DCT, 2b-2) is converted into a second inverse discrete cosine transform unit (DCT). Inverse Discrete Cosine Transform2: IDCT2, 2b-7) results in the same result as the original input. In other words, there is no loss of information in the Discrete Cosine Transform (DCT) 2b-2 and the Inverse Discrete Cosine Transform (IDCT) 2b-7.

Therefore, between the Discrete Cosine Transform (DCT, 2b-2) and the Inverse Discrete Cosine Transform (IDCT, 2b-7), the following equations (6a, 6b, 6c) Relationships are established.

Using FIG. 3 as above, the arrangement can be simplified as shown in FIG.

Bit rate control applied to the bit rate conversion encoder as shown in FIG. 4 is a process of obtaining a quantization parameter QP2 for a coding unit of an image. Here are some things to consider:

1) It must satisfy the given target bit rate. That is, if the bit rate of the input video is R1 and the output bit rate is R2, the buffer 2'-12 of the bit rate conversion encoder and the input buffer (not shown) of the decoder (not shown) on the receiving side overflow. The quantization parameter QP2 should be assigned so that overflow does not occur. However, in the rate conversion encoding process, since R1> R2, there is a relationship of QP1 ≦ QP2 between the quantization parameters for the input bit stream and the output bit stream.

2) Maintain good image quality if possible under the given bit rate. That is, by allocating quantization parameters that can be allocated in units of macroblocks, it is necessary to maintain consistent image quality and to generate less encoding distortion.

However, due to the rate-distortion characteristic as shown in FIG. 5, a complex image has a larger distortion reduction effect due to an increase in bit amount than a simple image.

Therefore, in order to equalize the distortion caused by the increase in the bit amount as compared with the distortion among the images constituting one screen, the bit amount allocated to the simple image is If the amount of bits is allocated to a complex image, the increase in distortion in a simple image And the reduction of distortion in complex images to be.

However, since a complex image has a larger distortion reduction effect due to an increase in bit amount than a simple image, Becomes

Therefore, the average distortion is reduced by allocating a portion of the bit amount allocated to a simple image to maintain a constant image quality.

However, the bit rate conversion encoder according to the related art analyzes the received bit string to obtain a graph of bit usage per macroblock in one screen, and transforms the graph according to the ratio of the input bit rate R1 and the output bit rate R2. After that, the value of the quantization parameter is adjusted while considering the difference between the target bit amount and the actual bit amount. This method does not consider the distortion characteristics, but only the bit generation amount, so that the converted coded image does not maintain good image quality. There is a problem.

Accordingly, an object of the present invention is to provide a bit rate conversion encoding apparatus capable of greatly reducing memory and computation amount while performing the same operation as the conventional bit rate conversion encoding apparatus.

1 is a view showing an example of a bit rate conversion encoding apparatus according to the prior art;

FIG. 2 is a diagram showing the detailed configuration of the bit rate conversion encoding apparatus of FIG.

3 is a diagram showing another detailed configuration of the bit rate conversion encoding apparatus of FIG.

FIG. 4 is a diagram showing another detailed configuration of the bit rate conversion encoding apparatus of FIG. 1; FIG.

5 is a diagram illustrating bit rate-distortion characteristics according to complexity of an image;

6 is a diagram showing an embodiment of a bit rate conversion encoding apparatus according to the present invention;

FIG. 7 illustrates an embodiment of a target distortion determination method of the target distortion determiner of FIG. 6.

Explanation of symbols for main parts of the drawings

110: target bit amount determination unit 120: bit rate-distortion estimation unit

130: target distortion determiner 140: quantization parameter determiner

A feature of a bit rate conversion encoding apparatus according to the present invention for achieving the above object is a bit rate conversion encoding apparatus including a variable length decoding unit, a first addition unit and a quantization unit, the current being output from the variable length decoding unit. A target bit amount determiner for determining a target bit amount assignable to a frame, and a quantization parameter applied in macroblock units from a target bit amount determined by the target bit amount determiner and a difference signal output from the first adder. A target distortion of the entire n-th frame currently transform-coded from the bit rate-distortion estimator for estimating the generated bit amount and the distortion value, and the signal output from the target bit amount-distortion estimator; A target distortion determiner for determining and a k-th macro according to the bit amount and distortion value outputted from the bit rate-distortion estimator Is determined by a quantization parameter to generate a distortion close to the target distortion output from the target distortion determining unit is composed by comprising a determined quantization parameter to output to the quantizing distortion in the curve-bit-rate for the block.

The target bit amount determining unit may determine a target bit amount assignable to the current frame output from the variable length decoding unit according to the following equation.

here, Is the total amount of bits generated from the n th video frame that has been transform coded, Means the total amount of bits occurring in the nth frame of the input bit stream.

For preventing overflow and depletion of the buffer of the bit rate converter and the input buffer of the decoder of the receiver The allowable range of is another in that it is characterized by the following equation.

here, Is the average amount of bits per frame (i.e. )ego, Is the state of the buffer Buf of the bit rate conversion encoder at the instant when the (n-1) th frame is transform-coded, Is the size of the buffer Buf of the bit rate conversion encoder, which is determined by the product of the transmission bit rate and the buffer delay.

The target bit rate determiner determines the total target bit rate for the nth frame when the error of the bit rate control is taken into account. It is another gong to decide the following equation.

here, Denotes the difference between the target bit amount and the actual bit amount for the i-th frame.

The target bit amount determination unit determines the total target bit amount for the nth frame. It is another feature to determine the following equation.

here, The amount of spare bits considering the bit rate control error is determined as approximately 10% of the buffer size.

The target distortion determiner may determine the target distortion of the entire nth frame that is currently transform-coded from the signal output from the target bit amount determiner and the signal output from the bit rate-distortion estimator. It is to feature.

here, Is the total bitrate-distortion curve, and the target bitrate Denotes a target bit amount.

The quantization parameter determiner uses a bit rate-distortion curve for the k-th macroblock according to the bit amount and the distortion value output from the bit rate-distortion estimator. Target distortion output from the target distortion determiner And a quantization parameter (QP) for generating distortion close to the following equation.

here, Is an estimate of the distortion for the kth macroblock (MacroBlock: MB), Silver inverse quantizer Means the quantization parameter used in.

Other objects, features and advantages of the present invention will become apparent from the following detailed description of embodiments taken in conjunction with the accompanying drawings.

Hereinafter, a preferred embodiment of a bit rate conversion encoding apparatus according to the present invention will be described.

FIG. 6 is a diagram showing an embodiment of a bit rate conversion encoding apparatus according to the present invention. As shown in FIG. 4, a target bit amount assignable to a current frame output from the variable length decode unit 2'-1 is shown. Quantization applied in units of macroblocks from the target bit amount determination unit 110 to determine the target bit amount determined by the target bit amount determination unit 110 and the difference signal output from the first adder 2'-3. A bit rate-distortion estimator 120 for estimating a bit amount and a distortion value generated according to a parameter, a signal output from the target bit amount determiner 110 and a signal output from the bit rate-distortion estimator 120 A target distortion determiner 130 for determining a target distortion of the entire n-th frame to be currently transcoded, and a bit rate for the k-th macroblock according to the bit rate and the distortion value output from the bit rate-distortion estimator 120. Recall from the distortion curve Determining a quantization parameter to generate a distortion close to the desired output from the distortion table distortion determining unit 130 is configured by including a quantization parameter determination unit 140 for output to the quantization unit (2'-4).

FIG. 7 is a diagram illustrating an example of a target distortion determination method of the target distortion determiner of FIG. 6.

The operation of the bit rate conversion encoding apparatus according to the present invention configured as described above will be described in detail with reference to FIGS. 6 and 7.

First, the target bit amount determining unit 110 determines and outputs a target bit amount assignable to the current frame output from the variable length decoding unit 2'-1, as shown in FIG. As shown in the figure, the bit rate conversion encoder performs re-quantization by adjusting the quantization parameter QP2 applied to the quantization unit 2'-4, thereby converting the bit string having the bit rate R1 to the bit rate having the bit rate R2. Will be converted to heat. Finally, when converting the bit rate from R1 to R2 (where R1> R2), the total amount of bits in the nth frame of the input bit stream Amount of bits generated from the n th video frame The relationship of must satisfy Equation 7 below.

Therefore, the target bit amount determination unit 110, when considering the error of the bit rate control, the total target bit amount for the n-th frame It is determined as Equation 8 below.

here, Denotes the difference between the target bit amount and the actual bit amount for the i-th frame.

On the other hand, as shown in Fig. 4, the overall target for the n-th frame for preventing overflow and depletion of the buffers Buf (2'-12) of the bit rate conversion encoding apparatus and the input buffer (not shown) of the decoder at the receiving end. Bit quantity Is as shown in Equation 9 below.

here, Is the average amount of bits per frame (i.e. )ego, Is the state of the buffers Buf 2'-12 of the bit rate conversion encoder as shown in FIG. 4 at the time when the (n-1) th frame is transcoded. As shown in Fig. 4, the size of the buffer Buf 2'-12 of the bit rate conversion encoder is input by the amount of transmission bits and the buffer delay (buffer Buf 2'-12 of the bit rate conversion encoder and the receiving end. Multiplication by a buffer (not shown). At this time, the size of the input buffer of the receiver is also Decide on As a result, the target bit amount determination unit 110 determines the total target bit amount for the nth frame. From Equation 8 and Equation 9 is determined as shown in Equation 10 below.

here, The amount of spare bits considering the bit rate control error is determined as approximately 10% of the buffer size.

On the other hand, the signal component affected by quantization in MPEG video compression is a DCT coefficient excluding intra DC. Therefore, in the target bit amount determination unit 110, the total bit amount generated from the n th video frame that is transcoded. As shown in Equation 11 below, the amount of bits generated from the DCT coefficient excluding intra DC From information other than and Can be divided into

here, Denotes the amount of bits generated from information not affected by quantization, such as header information and motion vectors.

Therefore, in the target bit amount determining unit 110, the bit amount allocated to the DCT coefficient directly affected by the quantization from the above equations (10) and (11). It is determined as shown in Equation 12 below.

here, Is not affected by quantization, so Has almost the same bit rate as Can get the exact value from the input bit stream, Since the exact value can be obtained from Equation 10, the bit amount to be allocated to the DCT coefficients affected by the quantization unit (Quantization Q) 2'-4 as shown in FIG.

Then, the bit rate-distortion estimator 120 applies a quantization parameter applied in macroblock units from the target bit amount determined by the target bit amount determiner 110 and the difference signal output from the first adder 2'-3. The bit amount and distortion value generated according to (QP2) are estimated and output.

Here, the quantization parameter QP2 may have a value between 1 and 31. The larger the value of the quantization parameter QP2, the smaller the amount of bits generated and the greater the distortion. Since the input bit rate R1 is larger than the output bit rate R2, the input bit rate R1 is always larger than the quantization parameter QP1 of the quantization parameter QP2.

Accordingly, the bit rate-distortion estimation unit 120 according to the quantization parameter Q applied in units of macroblocks (MBs) according to the quantization parameter Q The amount of bits generated by encoding And distortion Will be output. Here, the unit of the quantization parameter Q is to be.

Subsequently, the target distortion determiner 130 determines the target distortion of the entire n-th frame currently transform-coded from the signal output from the target bit amount determiner 110 and the signal output from the bit rate-distortion estimator 120. Will print.

That is, the target distortion determiner 130 outputs the output of the target bit amount determiner 110. ) And the output of the bit rate-distortion estimator 120 ( And distortion ), The target distortion of the entire nth frame to be currently transform-coded Will print

This process will be described with reference to an example of determination of target distortion for three macroblocks (MacroBlock: MB) as shown in FIG.

First, a bit rate-distortion curve (R-D Curve) for each macro block (MacroBlock: MB) is obtained from the bit rate-distortion estimation result for each macro block (MacroBlock: MB).

The overall bitrate-distortion curve is obtained by adding the obtained bitrate-distortion curves to the bitrate axis. Obtain Here, the bit rate-distortion curve for each macroblock (MacroBlock: MB) has a step shape. If there are 31 types of quantization parameters applicable to the macroblock (MacroBlock: MB), the number of macroblocks (MacroBlock: MB) is N. In this case, the overall bit rate-distortion curve may have a distortion value of up to N * 31 steps.

In the target distortion determination process, the target bit amount obtained from the above-described Equation 12 for the macroblocks (MacroBlock: MB) belonging to the nth frame to be currently transcoded. In the encoding process, the distortion value for minimizing the distortion difference between the bit rate and distortion curves is obtained. That is, as shown in Fig. 7, the overall bit rate-distortion curve And target bit rate Is the distortion value corresponding to the point where This is obtained by the following equation (13).

Accordingly, the quantization parameter determiner 140 determines the target distortion determiner 130 in the bit rate-distortion curve for the k-th macroblock according to the bit amount and the distortion value output from the bit rate-distortion estimator 120. The quantization parameter for generating distortion close to the output target distortion is determined and output to the quantization unit 2'-4.

That is, the quantization parameter determiner 140 determines the bit rate-distortion curve for the k-th macroblock according to the bit amount and the distortion value output from the bit rate-distortion estimator 120. Target distortion output from the target distortion determiner 130 at The quantization parameter QP for generating distortion close to is obtained from Equation 14 below.

here, Is an estimate of the distortion for the kth macroblock (MacroBlock: MB), Silver inverse quantizer Quantization parameter used in.

Therefore, if each macroblock (MacroBlock: MB) is transform-coded by applying the quantization parameter (QP) value selected by Equation 14, the distortion of all macroblocks (MacroBlock: MB) as shown in FIG. The amount of bits allocated to each macroblock (MacroBlock: MB) is If you add up the bits, the target bit amount is Does not match However, the amount of free bits to prevent the buffer fullness from being too high or too low during the overall target bit rate determination process. You can use to avoid depletion and overflow of the buffer.

As described above, the bit rate conversion encoder according to the present invention performs the same operation as that of the conventional bit rate conversion encoder, but has an effect of greatly reducing the amount of memory and computation.

Those skilled in the art will appreciate that various changes and modifications can be made without departing from the spirit of the present invention.

Therefore, the technical scope of the present invention should not be limited to the contents described in the embodiments, but should be defined by the claims and their equivalents.

Claims (7)

A bit rate conversion encoding apparatus including a variable length decode unit, a first adder, and a quantizer, A target bit amount determiner for determining a target bit amount assignable to the current frame output from the variable length decode unit; A bit rate-distortion estimator for estimating a bit amount and a distortion value generated according to a quantization parameter applied in macroblock units from the target bit amount determined by the target bit amount determiner and the difference signal output from the first adder; A target distortion determiner for determining a target distortion of the entire n-th frame currently transform-coded from the signal output from the target bit amount determiner and the signal output from the bit rate-distortion estimator; Determining a quantization parameter for generating a distortion close to a target distortion output from the target distortion determiner in the bitrate-distortion curve for the k-th macroblock according to the bit rate and the distortion value output from the bitrate-distortion estimator; And a quantization parameter determiner for outputting to the quantizer. The method of claim 1, And the target bit rate determining unit determines a target bit amount assignable to the current frame output from the variable length decode unit according to the following equation. here, Is the total amount of bits generated from the n th video frame that has been transform coded, Means the total amount of bits occurring in the nth frame of the input bit stream. The method of claim 2, For preventing overflow and depletion of the buffer of the bit rate converter and the input buffer of the decoder of the receiver The allowable range of the bit rate conversion encoding apparatus characterized by the following equation. here, Is the average amount of bits per frame (i.e. )ego, Is the state of the buffer Buf of the bit rate conversion encoder at the instant when the (n-1) th frame is transform-coded, Is the size of the buffer Buf of the bit rate conversion encoder, which is determined by the product of the transmission bit rate and the buffer delay. The method of claim 2, The target bit rate determiner determines the total target bit rate for the nth frame when the error of the bit rate control is taken into account. Bit rate conversion encoding apparatus characterized in that determined by the following equation. here, Denotes the difference between the target bit amount and the actual bit amount for the i-th frame. The method of claim 2, The target bit amount determination unit determines the total target bit amount for the nth frame. Bit rate conversion encoding apparatus characterized in that determined by the following equation. here, The amount of spare bits considering the bit rate control error is determined as approximately 10% of the buffer size. The method of claim 1, The target distortion determiner determines the target distortion of the entire n-th frame currently transform-coded from the signal output from the target bit amount determiner and the signal output from the bit rate-distortion estimator by the following equation. Bit rate conversion encoding apparatus. here, Is the total bitrate-distortion curve, and the target bitrate Denotes a target bit amount. The method of claim 1, In the quantization parameter determiner Bit rate-distortion curve for k-th macroblock according to bit amount and distortion value output from the bit rate-distortion estimator Target distortion output from the target distortion determiner 130 at And a quantization parameter (QP) for generating distortion close to the following equation. here, Is an estimate of the distortion for the kth macroblock (MacroBlock: MB), Silver inverse quantizer Means the quantization parameter used in.