KR960015393B1 - Quantization level controller for digital video encoding apparatus - Google Patents

Abstract

a movement compensation area number counting part(40) which judges the macro block which is movement-compensated and the macro block which is not movement-compensated on the basis of the average absolute error rate between the present source signal and the recovered signal stored in a frame memory(118), and counts the number of the movement compensation macro blocks and the number of the non-movement compensation macro blocks; and a quantization level determining part(50) which provides a second quantization level as a signal controlling the quantization level of a quantizer(104) by outputting the second quantization level to the quantizer(104).

Description

Quantization Level Adjuster of Digital Image Coding Device

1 is a block diagram of a conventional digital video encoding apparatus.

2 is a block diagram of a digital video encoding apparatus having a quantization level adjuster according to the present invention.

3 is a MC / non-MC determination curve.

* Explanation of symbols for main parts of the drawings

10: frame buffer 20: signal source encoder

30: buffer 40: motion compensation area number counting unit

50: quantization level determiner 60: quantization level regulator

100: subtractor 102: DCT unit

104: quantizer 106: VLC unit

110: reverse quantizer 112: IDCT unit

114: adder 116: motion estimation and compensation unit

118: frame memory

The present invention relates to a digital image encoding apparatus, and more particularly, to a quantization level adjuster of a digital image encoding apparatus for differently adjusting a quantization level of a quantizer according to the number of motion compensation regions when encoding a digital image signal.

In general, a video encoding apparatus is a signal processing technology that is the basis of a video service such as a video telephone, video conferencing, and next-generation digital high-definition television (HDTV), in order to transmit the video information to a channel having a limited transmission band It is a device for encoding information.

1 shows a conventional video encoding apparatus for compressing and encoding video information. A typical video encoding apparatus includes a frame buffer 10 for dividing an input digital video signal into predetermined image processing units, and a signal source for compressing and encoding the amount of data transmitted by removing redundancy present in the divided data word. The source coder 20 so that the amount of data generated by the signal source encoder can be controlled in accordance with the amount of data while temporarily storing the data generated by the source coder 20 and the signal source encoder 20. A buffer 30 for generating a control variable and transmitting it to the signal source encoder 20.

The signal source encoder 20 aims to reduce the amount of data by removing redundancy present in the signal source. Signal source encoding methods currently used include prediction encoding, transform encoding, and band division encoding.

The buffer 30 generates a control signal that can control the amount of data generated by the signal source encoder 20 while temporarily accommodating the data so that the data generated from the signal source encoder 20 can be transmitted to a transmission channel having a constant bandwidth. do. The control of the amount of data input to the buffer is such that data is not depleted or overflowed in the buffer, and mainly controls the quantization step of a quantizer (not shown) which is a component of the signal source encoder 20.

In order to adjust the quantization step size of the quantizer, the buffer 30 feeds a quantization rate control variable (hereinafter referred to as QP) to increase or decrease the quantization step according to the current buffer state. That is, after coding one group of block (GOB consists of 33 macroblocks (MB)), the buffer status is checked, the QP value is changed according to the status, and the changed QP is converted into a quantizer. Feedback. Accordingly, the quantizer performs quantization with the size of the quantization step corresponding to the converted QP.

Since the QP that adjusts the quantization level of the conventional quantizer is simply applied at the same quantization level to all the macroblocks in one GOB according to the amount of data generated by the signal source encoder 20, the signal characteristics for each macroblock in the GOB are ignored. .

The present invention has been made to solve the above-mentioned problems, and each macroblock in one GOB is divided into a motion compensation area and a non-motion compensation area, and then the motion compensation area having important information is finely quantized for the current image. The present invention provides a quantization level adjuster of a digital image encoding apparatus that adjusts a quantization level so that quantization can be performed according to an image quality.

In order to achieve the above object, the present invention provides a motion for estimating and compensating a motion amount between a current frame video signal and a previous frame video signal in a video signal stream including a GOB (Group Of Block) having a plurality of macro blocks. An estimation and compensation unit; A subtractor for detecting a difference signal between the current frame signal and a motion compensated signal provided to the motion estimation and compensation unit; A DCT unit for converting the difference signal provided from the subtractor into transform coefficients of a DCT transform region: a quantizer for quantizing a transform coefficient provided from the DCT unit to a predetermined quantization level for quantizing: a code included in the quantized transform coefficients A VLC unit having a high frequency as a short length code and a low frequency as a long length code according to a frequency of occurrence: storing and outputting code data provided from the VLC unit and outputting the quantization level of the quantizer A buffer for outputting: an inverse quantizer for inversely quantizing the quantized transform coefficients; An IDCT unit for performing an Invert Discrete Cosine Tranform (IDCT) on the inverse quantized transform coefficients to restore a differential signal before the DCT encoding; An adder configured to add a difference signal provided from the IDCT unit and a motion compensated signal provided from the motion estimation and compensation unit to generate a decoded signal: storing the decoded signal provided from the adder and providing the decoded signal to the motion estimation and compensation unit A digital video encoding device comprising a frame memory; Based on the average absolute error between the current original signal and the previous reconstructed signal stored in the frame memory, the macrocompartment and the motion compensation performed by the motion estimation and compensation unit among the plurality of megacompartments in the GOB are performed. A motion compensation area number counting unit for judging uncommitted giant blocks and counting these motion compensation giant blocks and the number of empty motion compensation giant blocks: a previous first quantization level provided by the buffer and the number of motion compensation area numbers Based on the number of motion compensation macrocompartments within one GOB provided by a division, the second motion is generated by generating a second quantization level in a macrocompartment unit by dividing the motion compensation macrocompartment into the motion compensation macrocompartment, and outputting the result to the quantizer. And providing the second quantization level as a signal for adjusting the quantization level of the quantizer. Level determination is characterized in that the configuration comprising a.

Hereinafter, with reference to the accompanying drawings the present invention will be described in detail.

2 is a block diagram of an adaptive video encoding apparatus according to an embodiment of the present invention, which includes a signal source encoder 20, a buffer 30, and a quantization level controller 60. Here, the signal source encoder 20 is a subtractor 100, a discrete cosine transform (hereinafter referred to as DCT) unit 102, a quantizer 104, a variable length coder (hereinafter referred to as VLC). 106), an inverse quantizer 110, an inverse DCT transform (hereinafter referred to as IDCT) unit 112, an adder 114, a frame memory 118, a motion estimation and compensation unit 116.

The subtractor 100 removes temporal redundancy between frames by subtracting the input data into a signal input through the motion estimation and compensator 116 before performing transform encoding. The DCT unit 102 converts a difference signal between two frames input from the subtractor 100 into a conversion coefficient of a conversion region. The quantizer 104 quantizes the transform coefficients provided from the DCT unit 102 and supplies them to the VLC unit 106. The VLC unit 106 reduces the overall code length by displaying a short frequency code and a short frequency code as a long code according to the frequency of occurrence of each code included in the quantized transform coefficient. Meanwhile, the inverse quantizer 110 and the IDCT unit 112 restore the signal output from the quantizer 104 as a reverse process of quantization and DCT to a difference signal before transform encoding. The adder 114 and the frame memory 118 add the IDCT signal and the signal through the motion estimation and compensator 116 to store the signal at the time of being input to the signal source encoder. The motion estimation in the motion estimation and compensator 116 compares a current video signal with a previous video signal stored in the frame memory 118 to estimate how much the current video signal has moved in which direction compared to the previous video signal. Motion Compensation (hereinafter, referred to as MC) is to move a signal of a frame stored in the frame memory 118 by the estimated motion.

In this case, the estimated motion may be represented by two-dimensional motion vectors representing the displacement of the pixels between the previous frame and the current frame. One of the motion estimation techniques is, for example, a block matching algorithm, in which the current frame is divided into a plurality of search blocks having the same size. Typically, the size of the search block is 16 * 16 pixels in size. In order to determine a motion vector for the search block in the current frame, a similarity calculation is performed with each of the search block of the current frame and each of a plurality of candidate blocks having the same size included in a generally large search area in the previous frame. When performing similarity calculations, various error functions such as mean squared error or mean absolute error can be used. And, by definition, the motion vector represents the displacement between the search block and the candidate block causing the minimum error function. The motion vector is used to reconstruct the image block by block from the previous frame in the receiver.

In addition, motion compensation moves candidate blocks by a displacement vector and a motion vector between candidate blocks having a minimum error function with respect to the search block detected by the similarity calculation.

The buffer 30 is connected to the VLC 106, which is the last block of the signal source encoder 20, and generates a QP in accordance with the output data amount of the VLC.

The QP generated in the buffer 30 is composed of 32 grades and is determined by Equation 1 for each GOB (Group of BIock).

[Equation 1]

BS (Buffer Size) = Capacity of the set buffer

BC (Bit Count) = value obtained by adding the bits generated by the source coder 20 and subtracting the average number of target bits for the transmission rate (the initial value of BC is 1/4 of the buffer capacity).

As can be seen from Equation 1, the value of QP increases when the amount of data generated by the encoding of the signal source encoder 20 exceeds the capacity of the set buffer.

The quantization level controller 60 uses an MC / non-MC determination method to determine a large segment MC in the GOB and counts the number of regions. The quantization level determination unit 50 is configured to determine the quantization level so that finer quantization can be performed on the region determined by the motion compensation using QP.

The MC / non-MC determination, i.e., whether to compensate for motion or not, is determined by the average absolute error between the current original signal and the previous restored signal stored in the frame memory 118.

An MC / non-MC decision curve for this is illustrated in FIG. 3, and the description of the MC / non-MC decision will be described using the block matching algorithm described above as an example.

According to the block matching algorithm, the motion vector represents the displacement between the search block and the candidate block resulting in a minimum error function, which is used by the receiver to reconstruct the image block by block from the previous frame. Therefore, a motion vector must be transmitted to the receiver, which is one of the factors that increases the amount of data to be transmitted.

Therefore, if possible, if the motion vector is zero, the amount of data transmitted can be reduced.

Therefore, the block matching algorithm detects an average absolute error between the search block and the candidate block causing the minimum error function, while detecting a displacement vector between the search block and the candidate block having the minimum error function. Also, a mean absolute error between the search block and the candidate block at the same position as the search block, that is, the displacement vector is zero, among the plurality of candidate blocks in the search area is detected.

Then, the average absolute error value between the candidate block having the zero displacement vector and the search block and the mean absolute error between the search block and the candidate block causing the minimum error function are compared. As a result of the comparison, if the two average absolute error values are not different from each other, the displacement vector between the search block and the candidate block causing the minimum error function is determined as the motion vector, and motion compensation is performed based on the determined motion vector. Rather than transmitting a motion vector of a certain size, it is possible to reduce the amount of data to be transmitted by determining the displacement vector 0 as a motion vector without motion compensation.

In FIG. 3, BD is a value corresponding to an average absolute error between a search block and a candidate block causing a minimum error function, and DBD is a value corresponding to an average absolute error between a candidate block having a motion vector of 0 and the search block.

Therefore, the MC area counting unit 40 determines the MC / non-MC area of each large compartment in the GOB based on the MC / non-MC decision curve of FIG. 3 and counts the number.

The quantization level determiner 50 corresponds to the total number of macrocompartments determined as the MC region among the macrocompartments within each GOB input from the MC region counting unit 40 and the fullness of the buffer input from the buffer 30. Based on the quantization level to generate a signal (Q) for adjusting the quantization level of each of the macroblocks according to whether each of the macroblocks to be currently quantized is an MC region or a non-MC region. Q is expressed as a function of QP, the output signal of the buffer, and the number of motion image areas.

[Formula 2]

[Equation 3]

[Equation 4]

[Equation 5]

[Equation 6]

[Formula 7]

Q: quantization level signal provided to the quantizer 104

QP: the previous quantization level signal provided from the buffer 30

N: Total number of large blocks in 1GOB

MC area: Giant compartment with motion compensation

Non-MC area: giant block without motion compensation

[Equation 2] means the following. QP≤12 indicates that the contents of the buffer 30 are relatively small compared to the contents of the reference buffer. The number of MC areas / N <0.5 indicates that the number of MC areas is less than l / 2 of the total number of large blocks of GOB. Accordingly, the quantizer 104 may express the secondary signal of the MC region in detail by lowering the QP by a step size with respect to the MC region. On the other hand, by increasing the QP of the non-MC region by the step size, the section having no important information is roughly represented. This method can improve the efficiency of encoding by expressing a region having important information in more detail and roughly expressing a region that is not.

[Equation 3] means the following. As QP ≤ 12 means that the content of the buffer 30 is small but the number of MC areas / N≥0.5 has a large number of MC areas, the quantization level is lowered for the MC area as shown in [Equation 2], and the quantization level for the non-MC area. Increasing may cause an unexpected phenomenon such as buffer overflow by suddenly increasing the contents of the buffer 30 significantly. Therefore, in this case, the above phenomenon can be controlled by taking the Q value as 12 for the MC region and the non-MC region.

[Equation 4] means the following. 12 <QP <20 indicates that the contents of the buffer 30 are about 40 to 60% of the contents of the reference buffer. The number of MC areas / N <0.3 indicates that the number of MC areas in the GOB is less than 30%. In this case, as shown in [Equation 2], the coding efficiency can be improved by lowering the quantization level of the MC region by the step size and increasing the step size for the non-MC region.

However, as shown in Equation 5, when the number of MC areas / N≥0.3 indicates that there are relatively many MC areas generating a large amount of information, the buffer overflow may be generated when the above method is applied. In this case, therefore, this phenomenon is controlled by keeping 16 for the MC area and 18 for the non-MC area.

[Equation 6] means the following. QP≥20 indicates that the contents of the buffer are more than 60% of the contents of the reference buffer. The number of MC areas / N <0.2 indicates that the number of MC areas in the GOB is less than 20%. In this case, since the buffer contents are relatively full, the size is increased by 20 for the number of MC areas and 3 times the size for non-MC areas. However, as shown in Equation 7, when the number of MC regions / N≥0.2 is relatively large compared to the buffer fullness, it is not appropriate to quantize the MC region and the non-MC region. Therefore, in this case, the value of Q is matched with the QP value determined according to the contents of the buffer.

The quantization level controller of the digital image encoding apparatus according to the present invention extracts the number of motion compensation regions in relation to the motion compensation prediction and the change coding method, determines the temporary quantization level according to the contents of the buffer, and extracts the motion compensation regions. By determining the actual quantization level from the quantization controller using the number and the quantization level as inputs, the efficiency of encoding can be improved by quantizing the error signal in detail in the motion compensation region.

Claims (7)

A motion estimation and compensation unit 116 for estimating and compensating for a motion amount between a current frame video signal and a previous frame video signal in a video signal stream including a group of blocks (GOBs) including a plurality of macro blocks; A subtractor (100) for detecting a difference signal between the current frame signal and a motion compensated signal provided to the motion estimation and compensation unit (116); DCT unit 102 for converting the difference signal provided from the subtractor 100 into transform coefficients of the DCT transform region: A quantizer 104 for quantizing the transform coefficient provided from the DCT unit 102 to a constant quantization level. ); A VLC unit 106 for displaying the short frequency code with a high frequency and the low frequency code with a long length code according to a frequency of occurrence of each code included in the quantized transform coefficients. A buffer (30) for storing and then outputting the code data and outputting a quantization level of the quantizer (104): an inverse quantizer (110) for inversely quantizing the quantized transform coefficients; An IDCT unit (112) for performing an Invert Discrete Cosine Tranform (IDCT) on the dequantized transform coefficients and restoring a differential signal before the DCT encoding: a differential signal provided from the IDCT unit 112 and the difference signal An adder 114 for generating a decoded signal by adding a motion compensated signal provided from the motion estimation and compensator 116: the motion estimation and compensator 116 after storing the decoded signal provided from the adder 114. A digital image encoding device comprising a frame memory 118 provided by: A plurality of huge partitions in the GOB based on an average absolute error between a current original signal and a previously reconstructed signal stored in the frame memory 118. The motion estimation and compensation unit 116 determines the large partition in which the motion compensation is performed and the large partition in which the motion compensation is not performed, A motion compensation area number counting unit 40 that counts the number of large partitions and the number of empty compensation large partitions: a previous first quantization level provided from the buffer 30 and the number of motion compensation area counting units 40. Based on the number of motion compensation giant compartments provided in one GOB, the motion compensation giant compartment is divided into the non-motion compensation giant compartment and a second quantization level is generated using a macro compartment to the quantizer 104. And a quantization level determiner (50) for outputting the second quantization level as a signal for adjusting the quantization level of the quantizer (104). The method of claim 1, wherein the quantization level determiner (50); If the content of the buffer 30 is less than 40% of the reference buffer content and the number of motion-compensating giant compartments is less than one-half of the total number of mega- compartments of GOB, the default step size for the first quantization level for the motion-compensating giant compartments. Outputting the second quantization level lowered by as much as the quantizer 104, and outputting, to the quantizer 104, the second quantization level which is increased by a basic step size with respect to the first quantization level for the empty large partition. And a quantization level controller of the digital image encoding device. The method of claim 1, wherein the quantization level determiner 50: When the content of the buffer 30 is less than 40% of the content of the reference buffer, the number of motion compensation macro compartments is at least 1/2 of the total number of mega compartments of GOB And the second quantization level for both the motion compensation giant compartment and the motion compensation giant compartment, the same as the first quantization level when the content of the buffer 30 is 40% of the reference buffer content. And a quantization level adjuster of the digital image encoding device. The method of claim 1, wherein the quantization level determiner 50: the content of the buffer 30 is about 40% to 60% of the content of the reference buffer, the number of motion compensation macro compartments 30% of the total number of mega compartments of GOB If less, the second quantization level lowered by the base step size with respect to the first quantization level for the motion compensation giant segment is output to the quantizer 104, and for the first quantization level for the non-motion compensation giant compartment. And outputting, to the quantizer (104), a second quantization level increased by a basic step size. The method of claim 1, wherein the quantization level determiner 50: the content of the buffer 30 is about 40% to 60% of the content of the reference buffer, the number of motion compensation macro compartments 30% of the total number of mega compartments of GOB In case of abnormality, the second quantization level is output to the quantizer 104 in the same manner as the first quantization level when the content of the buffer 30 is about 50% of the content of the reference buffer. For the empty motion compensation large partition, the second quantization level is output to the quantizer 104 in the same manner as the first quantization level when the contents of the buffer 30 are about 55% of the contents of the reference buffer. A quantization level controller of a digital image encoding device. The method of claim 1, wherein the quantization level determiner 50: When the content of the buffer 30 is 60% or more of the content of the reference buffer, the motion compensation macro compartment number is less than 20% of the total mega compartment number of GOB, For the motion compensation macro block, the second quantization level is output to the quantizer 104 in the same manner as the first quantization level when the content of the buffer 30 is about 60% of the content of the reference buffer, while being empty. And a quantization level controller of the digital image encoding apparatus, wherein the second quantization level, which is increased by three times the basic step size with respect to the first quantization level, is output to the quantizer (04). The method according to claim 1, wherein the quantization level determiner 50 comprises: when the content of the buffer 30 is 60% or more of the content of the reference buffer, and the motion compensation macroblocks are 20% or more of the total megablocks of GOB, And a first quantization level as the second quantization level for the motion compensation macroblock and the motion compensation macroblock, to be output to the quantizer (104).