KR100752840B1 - Q/D model design method for quanitzation parameter control of transcoder and encoding device of multimedia contents - Google Patents

Abstract

The method of designing a Q / D model for adjusting a quantization coefficient of a transcoder according to the present invention changes the quantization coefficient of a transcoder, calculates re-quantization bit reduction rates accordingly, and requantizes by a predetermined quantization coefficient without passing through the transcoder. Calculating a bit reduction rate; Calculating a MAE between requantization bit reduction rates according to the quantization coefficients of the transcoder and bit reduction rate requantized by the predetermined quantization coefficients; Analyzing a relationship with the MAE by calculating a ratio between the quantization coefficients of the encoder and the modified quantization coefficients of the transcoder; And modeling the relationship between the ratio between the coefficients and the distortion ratio as the relationship between the ratio between the coefficients and the MAE is analyzed.

According to the present invention, it is possible to design a Q / D model that conforms to each device and is signal dependent regardless of the type of conventional encoder and transcoder, so that the bit rate of the compressed signal can be dynamically controlled regardless of the image quality. It can be effective. In addition, the content provider does not need to produce a plurality of contents according to the type of bandwidth, can transmit a single multimedia content to various types of client devices, it is possible to significantly reduce the transmission load of the multimedia providing server. .

Description

Q / D model design method for quanitzation parameter control of transcoder and encoding device of multimedia contents}

1 is a block diagram illustrating a method of designing a Q / D model for adjusting a quantization coefficient of a transcoder according to an embodiment of the present invention.

2 is a graph showing the bit reduction rate and the direct requantization bit reduction value of the transcoder stage according to the quantization coefficient of the transcoder in designing a Q / D model according to the present invention.

3 is a graph showing the MAE between the bit reduction rate and the direct requantization bit reduction value of the transcoder stage according to the quantization coefficient of the transcoder in designing a Q / D model according to the present invention.

4 is a graph showing the correlation between the ratio of the quantization coefficient of the transcoder and the quantization coefficient of the encoder according to the MAE value in designing a Q / D model according to the present invention.

5 is a graph illustrating a Q / D model designed for adjusting the quantization coefficients of a transcoder according to the present invention.

6 is a block diagram schematically illustrating the components of a multimedia content encoding apparatus to which a Q / D model is applied according to an embodiment of the present invention.

7 is a graph simulating VBR transcoding results of a multimedia content device according to an embodiment of the present invention.

8 is a table illustrating CBR transcoding results of a multimedia content device according to an embodiment of the present invention.

<Explanation of symbols for main parts of drawing>

100: multimedia content encoding device

110: encoder 112: DCT section

114: first quantization unit 116: first variable length coding unit

120: transcoder 121: variable long coding unit

122: inverse quantization unit 123: transcoding unit

124: second quantization unit 124a: Q / D model analysis module

124b: coefficient adjusting module 124c: quantization module

125: second variable length coding unit 200: decoding unit

The present invention relates to a multimedia content encoding apparatus and an efficient transcoder stage design technique that provide an efficient transcoding function so as to remove distortion of an image while reducing a bit rate.

Recently, various digital products such as mobile phones, smart phones, digital TVs, and personal digital assistants (PDAs) are widely used. These digital products have communication functions, message services, video / audio playback and storage functions, search functions, personal Provide various services such as schedule management function.

In particular, as devices for playing multimedia content are diversified as described above, the content providing device has a difficulty in performing encoding differently in consideration of hardware resources of the respective playback devices.

One of the encoding techniques developed in response to this need is transcoding.

Transcoding refers to an encoding technique that dynamically responds to factors such as differences in network environments such as bandwidth and differences in the amount of data constituting content.

As such, the use of transcoding technology enables stable transmission of video streams, processing, and maintenance of image quality. Accordingly, research on bit rate control of transcoding has been actively conducted.

However, re-adjusting the bit rate of an already encoded bit stream at the transcoding stage is not only technically difficult, but also involves the risk of data loss or corruption. For example, adjusting the compression variables of a bit stream without accurate analysis on the compression domain (eg, DCT domain) often results in distortion and decoding of the streamed data at the final output.

Accordingly, there is a need for an intelligent quantization processing technique on a transcoder capable of controlling bit rate more efficiently while minimizing image distortion.

In theory, the image quality of the bit rate and the image form a trade-off relationship. As the bit rate increases (when the compression rate drops), the image quality improves, and when the bit rate decreases (when the compression rate goes up), the image quality is distorted.

Therefore, how to adjust the bit rate in the transcoder stage is recognized as the biggest problem, and the technique of processing the requantization dynamically (depending on the signal) according to the type of signal suggests a solution to the above problem. I think it will.

Accordingly, the present invention uses a Q / D model designed to provide a selection criterion for quantization coefficients based on a conventional transcoder, and is encoded with factors such as the type of multimedia playback device, network environment, data size of multimedia content, and the like. It is an object of the present invention to provide a multimedia content encoding apparatus that processes requantization of a transcoder stage without data loss / corruption by dynamically adjusting a quantization coefficient in response to a signal.

In addition, the present invention analyzes the quantization coefficients used in the conventional encoder, the quantization coefficients used in the transcoder, the coefficients used for direct requantization without going through the transcoder, and the bit rate according to them, thereby minimizing image distortion while providing the minimum. It is another object of the present invention to provide a method for designing a Q / D model that provides a criterion for adjusting a transcoder quantization coefficient to achieve a compression ratio of.

In order to achieve the above object, the multimedia content encoding apparatus according to the present invention includes a DCT unit for DCT operation of multimedia content, a first quantizer for quantizing the DCT calculated content with a first quantization coefficient, and the quantized content. An encoder having a variable length coding unit for coding a variable length code; And a variable length coding unit for inversely encoding the encoded content with a variable length code, an inverse quantizer for inversely quantizing the variable length coded content with the first quantization coefficient, and extracting a bit stream of the content converted into a DCT region. If the ratio of the second quantization coefficient to the first quantization coefficient is smaller than the first reference value of the distortion rate variation interval, the second quantization coefficient is increased on the transcoding unit and the Q (Quantization parameter) / D (Distortion) model. The second quantization coefficient is reduced when the distortion is greater than the second reference value of the variation range, and if the distance between the first reference value and the second reference value is decreased, the second quantization coefficient is sequentially reduced until the deviation is made. A transformer provided with a second quantization unit for quantizing the transcoded content using the adjusted second quantization coefficient. It characterized in that it comprises a further.

In addition, the second quantization unit of the multimedia content encoding apparatus according to the present invention,

if (Q1 / Q2> 0.6)

if (Q1 / Q2> 0.6 && Q1 / Q2 <0.7)

Q2 = Q2-1

if (Q1 / Q2> 0.6 && Q1 / Q2 <0.7)

Q2 = Q2-1

else if (Q1 / Q2 ≤ 0.6)

Q2 = Q2 + 1

(Q1: first quantization coefficient, Q2: second quantization coefficient)

The second quantization coefficient is adjusted by an algorithm expressed as described above.

In order to achieve the above object, the Q / D model design method for quantization coefficient adjustment of a transcoder according to the present invention includes an encoder for performing DCT operation, quantization processing, and variable length coding, and variable length inverse coding and inverse quantization. A multimedia content encoding apparatus comprising a transcoder that performs processing, transcoding, requantization processing, and variable length coding, wherein the requantization bit reduction rates are calculated while changing the quantization coefficient of the transcoder, and the transcoder Calculating a bit quantization rate requantized by a predetermined quantization coefficient without going through; Calculating a mean absolute error (MAE) between requantization bit reduction rates according to the quantization coefficients of the transcoder and bit reduction rate requantized by the predetermined quantization coefficient; Analyzing a relationship with the MAE by calculating a ratio between the quantization coefficients of the encoder and the modified quantization coefficients of the transcoder; And modeling the relationship between the ratio between the coefficients and the distortion ratio as the relationship between the ratio between the coefficients and the MAE is analyzed.

In addition, the Q / D model design method for adjusting the quantization coefficient of the transcoder according to the present invention designates a section having a large change amount of distortion on the modeled Q / D model, and sets the first reference value and the second reference value of the designated section. It characterized in that it further comprises the step.

In addition, the first reference value and the second reference value used in the Q / D model design method for adjusting the quantization coefficient of the transcoder according to the present invention are reference values of a section in which the distortion rate is sharply lowered, and the quantization coefficient of the transcoder versus the encoder If the ratio between the quantization coefficients is less than the first reference value, the quantization coefficient of the transcoder is increased by a predetermined value. If the ratio is greater than the second reference value, the quantization coefficient of the transcoder is decreased by a predetermined value. If it is between the first reference value and the second reference value, the quantization coefficient of the transcoder is sequentially reduced by a predetermined value until it exits this section.

Hereinafter, a method for designing a Q / D model for adjusting a quantization coefficient of a multimedia content encoding apparatus and a transcoder according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

The technical idea of the present invention is to perform video by adaptively performing quantization (such as content size or bandwidth of a network environment) in a transcoder stage of a multimedia content encoding apparatus based on a Q (Quantization parameter) / D (Distortion) model. Minimize the distortion of the decoding.

Therefore, after the Q / D model for adjusting the quantization coefficient of the transcoder is designed, the quantization process must be adaptively performed at the transcoder stage of the multimedia content encoding apparatus based on the Q / D model for the quantization coefficient adjustment of the transcoder. This is explained first.

1 is a block diagram illustrating a method of designing a Q / D model for adjusting a quantization coefficient of a transcoder according to an embodiment of the present invention.

First, a general multimedia content encoding apparatus (not shown) includes an encoder and a transcoder, which performs a DCT operation, a quantization process, and variable length coding (VLC), The transcoder performs variable length inverse coding, inverse quantization processing, transcoding, requantization processing, and variable length coding.

First, the requantization bit reduction rates are calculated according to the change of the quantization coefficient of the transcoder (S100), and the output signal of the encoder that does not go through the transcoder is requantized by a predetermined quantization coefficient (hereinafter, “directly” direct) is calculated (S200).

In this case, the transcoder has a quantization coefficient before the present invention is applied, and means a conventional transcoder as a design target.

2 is a graph showing the bit reduction rate and the direct requantization bit reduction rate of the transcoder stage according to the quantization coefficient of the transcoder in designing the Q / D model according to the present invention.

(A), (b) and (c) of FIG. 2 respectively show quantization coefficients (hereinafter referred to as "Q1") of the encoder when "4", "8" and "12" are set. The graphs are graphs, and the x-axis of each graph is a quantization coefficient of the transcoder (hereinafter referred to as "Q2") as a variable, and the y-axis represents a corresponding (encoding) bit reduction rate.

In the graphs, the numerical value a1 indicated by the dark line represents the bit reduction rate according to the quantization coefficient of the conventional transcoder, and the numerical value a2 indicated by the dim line represents the bit reduction rate according to the direct requantization.

According to Fig. 2, even if the quantization coefficient of the conventional transcoder is reduced to a small size, it can be seen that the bit rate a1 is greatly reduced on each graph (A1, A2, A3 in each graph). This is because the bit stream has been performed to be converted to the DCT region and has a Laplacian Distribution characteristic.

In designing the Q / D model according to the present invention, the analysis procedure is performed in the DCT region, and in the related art, the bit rate of the bit stream is greatly reduced, which means that the distortion is likely to occur in the image.

On the other hand, since the bit rate a2 according to direct requantization is reduced while showing a linear characteristic on each graph, it can be seen that almost no image distortion occurs.

That is, the design of the Q / D model according to the present invention aims at the bit rate a2 characteristic according to direct requantization.

Subsequently, a mean absolute error (MAE) is calculated between the directly requantized bit reduction rate and the requantized bit reduction rate according to Q2 without passing through the transcoder stage (S300).

3 is a graph showing the MAE between the bit reduction rate and the direct requantization bit reduction rate of the transcoder stage according to the quantization coefficient of the transcoder in designing the Q / D model according to the present invention.

(A), (b) and (c) of FIG. 3 are graphs when the Q1 "is set to" 4 "," 8 "and" 12 ", respectively, and the x-axis of each graph is the" Q2 ". "As a variable, and the y-axis shows the calculated MAE value.

According to FIG. 3, as expected in the measurement result of FIG. 2, the portions B1, B2, and B3 of which bit rate is most rapidly reduced on each graph can be identified. Distortion is severely generated.

This suggests what values Q2 of the transcoder according to the present invention to be designed will be set fluidly.

Next, in order to design a specific Q2D Q / D model, a ratio between the quantization coefficient Q1 and the Q2 of the encoder is calculated, and the relationship with the calculated MAE is analyzed (S400).

4 is a graph showing the correlation between the ratio of the quantization coefficient of the transcoder and the quantization coefficient of the encoder according to the MAE value in designing the Q / D model according to the present invention.

(A), (b) and (c) of FIG. 4 are graphs when the Q1 "is set to" 4 "," 8 ", and" 12 ", respectively, and the x-axis of each graph is the" Q2 ". "Is a variable, and the y-axis has shown the" Q1 / Q2 "numerical value.

According to FIG. 4, it can be seen what value "Q1 / Q2" has when the MAE has the largest value or the lowest value.

In each graph of FIG. 4, the Q2 and Q1 / Q2 values are displayed when the MAE has the largest value (C1, C2, C3).

In this way, when the relationship between the ratios Q1 / Q2 and the MAE between the coefficients is analyzed based on the numerical value applied to the conventional transcoder, the relationship between the ratio between the coefficients and the distortion ratio is modeled (S500).

5 is a graph illustrating a Q / D model designed for adjusting a quantization coefficient of a transcoder according to the present invention.

Referring to the Q / D model shown in FIG. 5, the distortion ratio can be largely divided into three regions according to the ratio of Q1 / Q2, and the second region in which the distortion ratio drops to a predetermined value rapidly drops in the transcoder stage according to the present invention. A reference is provided to the quantization coefficient setting (S600).

Therefore, the quantization coefficients used in the transcoder stage of the multimedia content encoding apparatus according to the present invention may be determined according to the ratio of Q1 / Q2 corresponding to the second region.

For example, when the quantization coefficient Q1 of the encoder stage is set to a predetermined value (by network environment, etc.), in which area on the Q / D model the ratio with the quantization coefficient Q2 of the conventional transcoder stage is located. By judging, the numerical value of Q2 can be flexibly adjusted.

The flow of adjusting the Q2 value according to the embodiment of the present invention is as follows.

5, when Q1 / Q2 is smaller than "0.6", the value of Q2 is increased by a predetermined unit, and when Q1 / Q2 is larger than "0.7", the value of Q2 is decreased by a predetermined unit.

In addition, when Q1 / Q2 is between "0.6" and "0.7", the Q2 is sequentially decreased by a predetermined unit until it exits this section.

Therefore, according to the present invention, the encoding bit rate can be controlled while preventing distortion of an image by adaptively setting the quantization coefficient of the transcoder stage according to the output signal characteristics of the encoder stage.

Next, a multimedia content encoding apparatus whose encoding bit rate is controlled by the Q / D model will be described.

6 is a block diagram schematically illustrating components of the multimedia content encoding apparatus 100 to which the Q / D model is applied according to an embodiment of the present invention.

Referring to FIG. 6, the multimedia content encoding apparatus 100 according to the embodiment of the present invention includes an encoder 110 and a transcoder 120. The encoder 110 includes a DCT unit 112 and a first quantization unit. 114 and a first variable length coding unit 116, wherein the transcoder 120 includes a variable length coding unit 121, an inverse quantization unit 122, a transcoding unit 123, and a second quantization unit. 124, a second variable length coding unit 125.

First, the DCT unit 112 of the encoder 110 converts the distribution of pixels constituting the multimedia content into a frequency domain to separate important and non-important information, and loses non-critical information by lossy compression. Perform compression primarily.

In addition, the first quantization unit 114 aligns irregularly and densely distributed pixel values through a predetermined quantization coefficient, which is based on neural visual redundancy, meaning that certain data is lost. At this time, the larger the quantization coefficient, the smaller the loss of data, but the compression rate is reduced.

The first variable length coding unit 116 codes the quantized content with a variable length code, which is a lossless compression method in which the shortest code word is assigned to the most appearing intensity, and the importance of data is increased. Depending on the variable length code is used.

In the variable length coding, for example, a Huffman scheme or an arithmetic coding scheme may be used.

In this way, when the multimedia content is primarily compressed in the encoder 110, transcoding proceeds according to the network environment (bandwidth) and the amount of data.

The variable length coding unit 121 of the transcoder 120 inversely encodes the encoded content into a variable length code, which is a component corresponding to the first variable length coding unit 116.

Subsequently, the inverse quantization unit 122 inverse quantizes the variable inverse coded content with the first quantization coefficient, which is a component corresponding to the first quantization unit 114.

The multimedia content encoded by the variable length coding unit 121 and the transcoding unit 123 is represented on the DCT region, and the transcoding unit 123 extracts a bit stream of the content expressed on the DCT, and then the data stream. It handles transcoding, such as reducing the bit rate or lowering (or increasing) the resolution.

Since the present invention has a characteristic element in the function of the second quantization unit 124, a detailed description of the function of the transcoding unit 123 will be omitted.

Referring to FIG. 6, the second quantization unit 124 includes a Q / D model analysis module 124a, a coefficient adjustment module 124b, and a quantization module 124c. The Q / D model analysis module 124a is provided. 1 to 5 have Q / D model data designed through the conventional transcoder 120, and analyze the Q / D model as the quantized transcoded content data. .

In analyzing the Q / D model, the size of the encoded data, the network environment, the quantization coefficient of the encoder 110, the quantization coefficient of the current second quantization unit 124 which is generally set, and the like are used.

First, the Q / D model analysis module 124a determines that the ratio of the second quantization coefficient to the first quantization coefficient on the Q / D model is smaller than the first reference value (“0.6” in FIG. 5) of the distortion rate variation interval. Then, the coefficient adjusting module 124b increases the value of Q2 by "1".

Second, the Q / D model analysis module 124a determines that the ratio of the second quantization coefficient to the first quantization coefficient on the Q / D model is greater than the second reference value (“0.7” in FIG. 5) of the distortion rate variation interval. Then, the coefficient adjusting module 124b decreases the value of Q2 by "1".

Third, the ratio of the second quantization coefficient to the first quantization coefficient in the Q / D model analysis module 124a is the first reference value (“0.6” in FIG. 5) and the second in the distortion rate variation interval. If it is between the reference value ("0.7" in Fig. 5), the coefficient adjusting module 124b sequentially decreases the value of Q2 by "1" until it leaves this section.

In this way, as the coefficient adjusting module 124b flexibly adjusts the value of Q2, the quantization module 124c processes the requantization of the transcoder 120 stage using the adjusted Q2.

The process of analyzing the Q / D model by the Q / D model analysis module 124a may be programmed as follows.

(One embodiment of the Q / D model analysis program according to an embodiment of the present invention)

if (Q1 / Q2> 0.6)

if (Q1 / Q2> 0.6 && Q1 / Q2 <0.7)

Q2 = Q2-1

if (Q1 / Q2> 0.6 && Q1 / Q2 <0.7)

Q2 = Q2-1

else if (Q1 / Q2 ≤ 0.6)

Q2 = Q2 + 1

Here, Q1 means "first quantization coefficient" and Q2 means "second quantization coefficient".

FIG. 7 is a graph simulating a variable bit rate (VBR) transcoding result of a multimedia content device according to an embodiment of the present invention, and FIG. 8 is a constant bit rate (CBR) transformer of a multimedia content device according to an embodiment of the present invention. A table illustrating the coding results.

In the simulation graph of FIG. 7, the x-axis represents the quantization coefficient of the transcoder 120, and the y-axis represents the Peak Signal to Noise Ratio (PSNR). The transformer according to the present invention is compared to the PSNR value E1 according to direct requantization. While the PSNR value E2 according to the requantization of the coder 120 is improved, the distortion of the image does not occur (there is no sharply changing section) compared to the conventional art.

In addition, according to Figure 8, the portion indicated by the thick square line on the table corresponds to the experimental results according to the quantization of the transcoder 120 according to the present invention, the remaining part represents the experimental results by the conventional transcoding, ( a) Figure CBR transcoding result table of the garden image, (b) Figure CBR transcoding result table of the human image.

And, (c) the figure is a table of CBR transcoding results of the soccer game image, (d) the figure is a table of CBR transcoding results of the date screen image.

As a result of performing the CBR transcoding according to the present invention in the state where the data amount of the original image is different, the compression rate of the encoder 110 is different, and the bit rate is also different, all the numerical values ("Δ (dB ) ").

The bit stream re-quantized by the second quantizer 124 is lossless compressed by the second variable coding unit 125 and transmitted to the decoder 200 through a network. The decoder 200 processes inverse variable coding, inverse quantization processing, and inverse DCT transformation to reproduce original multimedia contents.

The present invention has been described above with reference to preferred embodiments thereof, which are merely examples and are not intended to limit the present invention, and those skilled in the art do not depart from the essential characteristics of the present invention. It will be appreciated that various modifications and applications which are not illustrated above in the scope are possible. For example, each component specifically shown in the embodiment of the present invention can be modified. And differences relating to such modifications and applications will have to be construed as being included in the scope of the invention defined in the appended claims.

The present invention has the following effects.

First, according to the present invention, it is possible to design a Q / D model that conforms to each device and is dependent on a signal regardless of the type of the conventional encoder and transcoder 120, and thus the bit rate of the compressed signal regardless of the image quality of the image. This has the effect of dynamically controlling.

Second, the content provider does not need to produce a plurality of contents according to the type of bandwidth, and can transmit one multimedia content to various types of client devices, and greatly reduce the transmission load of the multimedia providing server.

Claims (5)

A transcoding unit transcoding and encoding the multimedia content; And, And a transform unit configured to perform requantization on the trans-coded multimedia content, and determine requantization coefficients by comparing the quantization coefficients used in the transcoding unit. The multimedia content encoding apparatus of claim 1, wherein the requantization coefficients are sequentially reduced or increased by a predetermined unit based on the quantization coefficients of the transcoded multimedia content. The method of claim 1, wherein the transform unit includes a decoding function used by the transcoding unit for each function, and after combining and decoding the decoding function according to an encoding method of the transcoding unit, the transform unit uses the requantization coefficients determined by the transform unit. And requantize the multimedia content. delete delete

Images