KR20010059598A - Method of buffer control in video encoder - Google Patents

Abstract

PURPOSE: A method for controlling buffer of moving picture video encoder is provided to prevent a buffer overflow and a buffer exhaustion by creating a buffer overflow flag and a buffer exhaustion flag, by controlling a bit rate control operation, a code mode determination, and a DCT quantizer operation, and by inserting stuffing data according to the buffer exhaustion flag. CONSTITUTION: A process controls a bit generation of a buffer with a bit rate control mode, in case of a buffer overflow, and generates a buffer overflow signal according to a degree of the buffer overflow. The process prevents the buffer overflow by controlling a bit rate control operation, a macro block(MB) code mode determination, and a discrete cosine transform(DCT) quantizer operation.

Description

Method of buffer control in video encoder

The present invention provides a buffer control method and a method for controlling the overflow or depletion of a channel buffer in a video encoder of a video such as MPEG-1 (MPEG-1) and MPEG-2 (MPEG-2). A computer readable recording medium having recorded thereon a program for realization.

The MPEG video encoder outputs a variable rate of encoded bits at a fixed rate through a channel buffer. In this case, the video encoder controls encoding bit generation through bit rate control.

In fixed rate operation, bit rate control suppresses bit generation when the channel buffer is about to overflow and generates quantization parameter values to generate more bits when the channel buffer is about to be exhausted. However, bit rate control does not completely guarantee protection of channel buffer overflow or exhaustion.

When buffer overflow is about to occur, bit rate control is to maximize the quantization parameter value and to minimize the quantization parameter value when buffer exhaustion occurs. In this case, even if the quantization parameter is maximized, the encoding bit may be generated more than the rate of exiting from the channel buffer, and the buffer overflow may occur even if the minimum is less than the amount of exiting from the channel buffer. . Therefore, the video encoder must separately control the buffer to solve the buffer overflow or exhaustion problem, which is not guaranteed through the bit rate control.

There are many such buffer control methods, but the goal is to minimize image quality degradation with minimal hardware.

Nevertheless, conventional buffer control techniques do not fully guarantee buffer overflow or exhaustion, and hardware and performance are not optimized. Accordingly, there is a need for a method capable of preventing overflow and depletion of channel buffers in the MPEG video encoder with minimum hardware and minimum image quality degradation.

The present invention devised to meet the requirements as described above, the buffer control method for preventing the overflow of the channel buffer in the video encoder of the video such as MPEG-1 (MPEG-1), MPEG-2 (MPEG-2) and the like It is an object of the present invention to provide a computer-readable recording medium having recorded thereon a program for realizing the method.

In addition, the present invention provides a buffer control method for preventing the exhaustion of channel buffers in a video encoder of a video such as MPEG-1, MPEG-2, etc., and a computer for recording the program for realizing the method. Another purpose is to provide a readable recording medium.

1 is an exemplary configuration diagram of an MPEG video encoder to which the present invention is applied.

2 is an exemplary hardware operation timing diagram of an MPEG video encoder used in the present invention.

3 is a diagram illustrating an embodiment of a buffer control method according to the present invention;

* Explanation of symbols for the main parts of the drawings

11: frame memory unit 12: input processing unit

13: mode control unit 14: buffer control unit

15 Discrete Cosine Transform (DCT) Coding Unit 16 Variable Length Coding Unit

17: motion prediction / compensator 111: original image field buffer

112: encoded frame buffer 113: channel buffer

In order to achieve the above object, the present invention provides a method of controlling a buffer in a video and video encoder, wherein the buffer is overflowed when there is a risk of buffer overflow while controlling the bit in the bit rate control mode during normal operation. Generating a buffer overflow concern signal according to a degree of concern; And a second step of controlling a bit rate control operation, a macroblock (MB) coding mode, and a discrete cosine transform (DCT) quantization operation according to the buffer overflow concern signal to prevent buffer overflow. .

In addition, the present invention is a third step of encoding the buffer in a predetermined slice structure in a steady state, and generates a buffer depletion warning signal when the buffer is likely to be exhausted; And after the macroblock (MB) encoding, stuffing according to the buffer depletion warning signal, and starting the following macroblock (MB) from a slice header to prevent buffer depletion. .

According to an aspect of the present invention, there is provided a method of controlling a buffer in a video and video encoder, wherein the buffer is decoded in a predetermined slice structure in a steady state, and a buffer depletion warning signal is generated when the buffer is depleted. A first step occurring; And after macroblock (MB) encoding, stuffing according to the buffer depletion warning signal, and starting the following macroblock (MB) from a slice header to prevent buffer depletion.

In order to achieve the above object, the present invention provides a video video encoder having a processor to control bit generation in a bit rate control mode during normal operation. A first function of generating a buffer overflow concern signal; And a program for controlling a bit rate control operation, macroblock (MB) coding mode determination, and discrete cosine transform (DCT) quantization operation according to the buffer overflow concern signal to realize a second function for preventing buffer overflow. It provides a recording medium that can be read by.

In addition, the present invention provides a third function of encoding the buffer in a predetermined slice structure in a steady state and generating a buffer depletion warning signal when the buffer is depleted; And a program for further realizing a fourth function of preventing buffer depletion by causing macroblock (MB) to be stuffed according to the buffer depletion warning signal and starting a macroblock (MB) from a slice header after encoding the macroblock (MB). It provides a recording medium that can be read by.

According to an aspect of the present invention, there is provided a moving picture video encoder including a processor, which encodes a buffer in a predetermined slice structure in a normal state and generates a buffer depletion warning signal when the buffer is depleted; And after the macroblock (MB) encoding, the program for stuffing according to the buffer depletion warning signal and starting the following macroblock (MB) from the slice header to prevent the buffer depletion can be read by a computer. Provide a recording medium.

In order to solve the buffer overflow or exhaustion problem, the present invention generates a buffer overflow concern flag and a buffer exhaustion flag according to the channel buffer state, and performs bit rate control operation, encoding mode determination, and DCT quantization operation according to the buffer overflow concern flag. Controlling prevents buffer overflow and inserts stuffing data according to flags to prevent buffer exhaustion.

The above objects, features and advantages will become more apparent from the following detailed description taken in conjunction with the accompanying drawings. Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is an exemplary configuration diagram of an MPEG video encoder to which the present invention is applied.

As shown in FIG. 1, an MPEG video encoder encodes an image input in a field 720 × 240 or frame 720 × 480 in a 16 × 16 macroblock (MB) unit. A bitstream is generated and output at a fixed rate through the channel buffer 113. The data generation of encoding is variable rate, and bit generation is controlled through buffer control.

As an example, an operation of a video encoder for encoding interlaced video into a frame-picture structure will be described. First, video data inputted in units of fields is first converted into a frame memory unit 11 before being encoded. ) Is stored in the original image field buffer 111. This is because at least all of the first field of the frame and eight lines of the second field must be input to form and encode the MB of the frame structure.

The input processor 12 reads data from the original image field buffer 111 to form an MB, and transmits the MB to the mode controller 13 and the motion predictor / compensator 17. At this time, the motion predictor / compensator 17 finds the MB most similar to the current MB from the previous frame stored in the encoding frame buffer 112 and transfers the MB to the mode controller 13.

The mode controller 13 determines whether to encode the current MB in the intra mode or the inter mode, and in the intra mode, discrete cosine transform (DCT) encoding the current MB as it is. In the inter mode, the motion prediction error MB obtained by subtracting the current MB data by the motion compensation MB data is transmitted to the DCT encoder 15.

The DCT encoder 15 divides the macroblock MB into 8 × 8 block units to perform DCT, quantization, scan, inverse quantization, and inverse DCT, so that SCAN data is variable length coded (VLC). Variable length code), and the inverse DCT result data is transmitted to the mode control unit 13. In this case, the DCT coded data transmitted to the mode control unit 13 is added as it is in the intra mode, and is added to the motion compensation MB data used to generate the motion prediction error MB in the inter mode. Stored as).

The variable length coding (VLC) unit 16 performs VLC encoding on SCAN data and other data to generate an MPEG bitstream and transmits the SCAN data to the channel buffer 113. At this time, the channel buffer 113 outputs this encoded bitstream at a fixed rate.

In the encoding process, the buffer controller 14 controls the generation of encoding bits to ensure a predetermined bit rate output without overflow or depletion of the channel buffer 113.

In FIG. 1, the three types of buffers 111 to 113 are bundled into the frame memory unit 11 because these buffers are implemented on one DRAM when the video encoder is actually implemented.

2 is an exemplary hardware operation timing diagram of an MPEG video encoder used in the present invention.

In Fig. 2, "frm_ref" is a signal indicating the start of a frame of interlaced video input, and the encoding is started by a reset signal. In addition, "piclk" is a signal indicating the start of encoding of a frame. As described above, a signal is generated when the first field is input and eight lines of the second field are input.

Meanwhile, encoding is performed in MB units, and in the case of "NTSC" video, one frame includes 1,350 MB. In hardware operation, 1,350 MB operate according to a " mbclk " signal evenly divided within a " piclk " interval. However, it takes several "mbclk" sections until MB data starts to be encoded and outputs encoded data, and 1,350 MBs are pipelined. In the example of FIG. 2, the processing time of MB encoding is a 6 mbclk period from the time point at which "sydata (search area data)" for motion estimation is received to the time point at which "vstrm (coding bitstream)" is output.

So far we have looked at the configuration and timing of the video encoder.

Now, the buffer control for controlling the encoding bit generation in the video encoder will be described in more detail.

3 is a diagram illustrating an embodiment of a buffer control method according to the present invention.

As shown in Fig. 3, the buffer control is divided into a bit rate control 31, a buffer overflow control 32, and a buffer depletion control 33.

The bit rate control 31 is a function of controlling the generation of encoded bits to be an output bit rate on average, and controls bit generation through output of a quantization parameter (QP) value of the DCT encoder.

The buffer overflow control 32 and the buffer depletion control 33 serve to prevent buffer overflow and exhaustion that are not prevented by the bit rate control 31.

The performance of the bit rate control 31 is determined to the extent that the quality of the reconstructed image and the channel buffer overflow or depletion do not occur.

In general, a uniform image quality between images and images is evaluated for good image quality. Also, as the channel buffer overflows, the quantization parameter value increases to a maximum value, and the QP becomes a maximum value just before the buffer overflows. When the depletion occurs, the QP decreases gradually, and the minimum QP is shown immediately before the buffer is depleted, indicating the maximum image quality, and the buffer overflow prevention and depletion prevention control is also facilitated.

In the present embodiment, only the buffer overflow control 32 and the buffer depletion control 33 are handled in the buffer control of the video encoder. In this case, the buffer overflow control 32 and the buffer depletion control 33 of the present embodiment have a bit rate. No limitation is imposed on the control 31 method.

First, the buffer overflow prevention control 32 method will be described in more detail.

In normal operation, it controls the bit generation in the bit rate control mode. If there is a risk of overflowing the buffer, the buffer overflow flag is generated to switch the operation to the buffer overflow control mode.

In the present invention, a buffer overflow warning flag, a buffer overflow warning flag, and a buffer overflow danger flag are generated, and bit rate control operation, encoding mode (Intra or Inter mode) determination, and DCT quantization operation are performed according to these flags. Controlling prevents buffer overflow.

Thus, the buffer overflow prevention method of the present invention guarantees buffer overflow prevention even when the output bit rate exiting the channel buffer is 1 Mbps or less. That is, it can be applied to all bit rates of 1Mbps or more.

First, a process of generating a buffer overflow flag used in the present invention will be described in detail.

In order to prevent the buffer overflow, the present invention has three flags according to the degree of buffer overflow concern, and values are set as follows according to the channel buffer state for each macroblock.

if (buf_empty <Th_overnote) overflow_note <= 1;

else overflow_note <= 0;

if (buf_empty <Th_overwarn) overflow_warn <= 1;

else overflow_warn <= 0;

if (buf_empty <Th_overdanger) overflow_danger <= 1;

else overflow_danger <= 0;

Here, "buf_empty" is an empty state value of the channel buffer at the moment of setting the flag value for MB, and is obtained as "[vbv_buffer_size-(bit_rate / picture_rate)-buf_state]". In the above equation, "vbv_buffer_size", "bit_rate", and "picture_rate" are sequence header information determined before encoding starts, and "buf_state" is the value of the moment when the flag value is set in the channel buffer where the data is constantly drawn out at the "bit_rate" rate from the channel buffer. Buffer full status. Here, the buffer overflow is not defined as the "vbv_buffer_size" suggested by the MPEG standard, but the value of the picture encoding is the same as the standard, except that "bit_rate / picture_rate", which is the number of bits that escape from the channel buffer during the picture period, is set in hardware. This is because "piclk" is not instantaneously at the time of rise, but is constant during the picture period.

"Th_overnote" is "[the maximum number of bits that can be generated when encoding in a normal state without a buffer overflow from the time when the overflow_note flag value is set to the end of generation of the encoding bit of MB by this flag]." The maximum number of bits that can be generated by the overflow_note flag MB is added to the maximum number of bits that can be generated by the overflow_warn flag MB.

"Th_overwarn" is "[the maximum number of bits that can be generated when encoding in overflow_note state from the time when the overflow_warn flag value is set to the end of generation of the encoding bit of MB by this flag]." Plus the maximum number of bits that can be generated by " "

"Th_overdanger" is "[the maximum number of bits that can be generated when encoding in overflow_warn state from the time at which the overflow_danger flag value is set to the time when generation of the MB encoded bit by this flag is terminated".

In the above description, MB is generated from the time point at which the buffer value is set due to buffer overflow until the end of generation of the MB encoded bit by this flag after DCT encoding or VLC encoding according to the flag after the buffer overflow flag is set. This time interval represents the time interval from when the MB buffer overflow flag is set to when the VLC encoding of the MB ends. This is dependent on the hardware implementation.

In the example of FIG. 2, overflow flags are generated immediately before the "mbclk" index 6, and the VLC output according to this flag, that is, the "vstrm" data ends just before the "mbclk" index 8, and thus has a duration of about 2 mbclk. . Here, the flag setting time should be made after the VLC output of the MB is finished and before the VLC output of the next MB is made.

In the above, the reference value for setting the flag value for the buffer overflow is presented. This reference value ensures that the channel buffer overflow risk step goes through the buffer overflow warning step and the warning step, and that the buffer overflow does not occur under the condition. The minimum value was determined. However, we do not consider the amount of escape from the channel buffer when bits are generated by MB encoding in the channel buffer. Considering this value, in fact, it can be lower than the value given above. However, this value depends on the bit rate "bit_rate", and since it is only 25 bits out of 1Mbps, it will be ignored here.

Second, the operation of bit rate control according to the buffer overflow flag will be described.

The operation of the bit rate control 31 only appears to produce a quantization parameter (QP) value.

Although not involved in the method of the bit rate control 31 in this embodiment, when the buffer overflow warning flag becomes 1, the bit rate control 31 causes QP to be set to the maximum value "31" unconditionally. This may naturally be set to " 31 " at that moment in accordance with the bit rate control 31 method, but not all bit rate control 31 methods are. In the present embodiment, in order to reduce the number of bits generated by the QP transmission in the operation according to the buffer overflow warning and the danger flag, the QP is maintained at "31" from the buffer overflow warning step to the dangerous step so that there is no change in the QP value. to be.

Third, the operation of the mode controller (MB encoding mode determination) according to the buffer overflow flag will be described.

In the present embodiment, when the buffer overflow warning flag is 1, the mode control unit sets the encoding mode of the MB to the intra mode.

The occurrence of buffer overflow generally occurs when encoding a complex part of an image in which the effect of motion prediction coding is insignificant. In this case, the mode control unit determines the intra mode or the inter mode. . In the case of such a complex image, encoding in intra rather than inter causes less bit generation. Therefore, in the present embodiment, when the buffer overflow warning flag appears, the MB encoding mode is set to the intra mode. When the buffer overflow warning flag or the buffer overflow danger flag is displayed, the buffer overflow warning flag is displayed. Therefore, it is set to the intra mode.

Fourth, the operation of the DCT encoder (quantization of the DCT coefficients) according to the buffer overflow flag will be described in more detail.

Among the buffer overflow flags in this embodiment, the flags that affect the quantization of the DCT coefficients are the buffer overflow warning flag (overflow_warn) and the buffer overflow risk flag (overflow_danger).

When "overflow_warn" appears, only the DC (Direct Current) component of the normal quantization value of the DCT coefficient of the block is left, and the remaining AC (AC: Alternative Current) component is made 0. The DC value is then saved for the "overflow_danger" operation. At this time, if "overflow_danger" is 1, all coefficients of the DCT coefficients are set to 0, and the DC component is set to the DC value of the previous block stored in the "overflow_warn" operation. This is because VLC encoding of DCT DC encodes a difference from a previous DCT DC value.

By doing this at the buffer overflow risk level, the bit generation of the encoding is suppressed to the maximum, thus ensuring buffer overflow protection even at 1 Mbps, much lower than the recommended bit rate of 4 Mbps of MPEG-2.

Fifth, the MB maximum bit generation amount will be described in more detail.

The maximum number of MB coded bits used to generate the buffer overflow prevention flag value depends on "chroma_format", "intra_dc_prec", "q_scale_type", "alternate_scan", "intra_vlc_format", QP, MB encoding mode, quantization matrix, and the like. Here the user definable quantization matrix is the reference matrix suggested by the standard, and the maximum according to the steady state, "overflow_note", "overflow_warn", and "overflow_danger" states when "chroma_format" is 4: 2: 0 and there are 6 blocks per MB. The values that can be generated are given for reference.

In the steady state (QP = 1, q_scale_type = 1, inter MB), "154 (MB header) + 1,538 (block data) x 6 = 9,382 bits".

In the buffer overflow state (QP = 31, q_scale_type = 0, intra MB, intra_dc_prec = 3), "27 + 171 x 6 = 1,053 bits".

In the buffer overflow warning state (QP = 31, intra MB, intra_dc_prec = 3, only DCT DC), "3 (MB header) + 20 (luminance block) x 4 + 21 (chromatic block) x 2 = 125 bits".

In the overflow risk state (QP = 31, intra MB, intra_dc_prec = 3, DCT_DC = previous_DC), "3 (MB hair) + 3 (luminance block) x 4 + 2 (chromatic block) x 2 = 19 bits".

At this time, the reference value of the flag value setting for the timing of FIG. 2 is as follows.

Th_overnote: 9,382 × 2 + 1,053 + 125 = 19,942

Th_overwarn: 1,053 × 2 + 125 = 2,231

Th_overdanger: 125 × 2 = 250

On the other hand, the buffer exhaustion prevention method 33 will be described in more detail.

MPEG encoding supports both fixed bit rate (CBR) and variable bit rate (VBR) output.

In the variable bit rate mode, if there is data in the channel buffer, the rate is "bit_rate". Therefore, depletion of the channel buffer is not a problem in VBR operation. However, in case of CBR operation, data should always be output at the "bit_rate" rate. If there is no data in the channel buffer, a problem of buffer depletion occurs in the video encoder. This problem can be solved by filling stuffing data that contains no information.

Insertion of stuffing data means waste of information in that limited bit. Therefore, stuffing should be generated as little as possible. Here, the least possible stuffing is to stuff whenever a buffer exhaustion problem occurs. However, MPEG does not allow stuffing at any position during encoding.

In MPEG-1, each macroblock can be stuffed so that buffer exhaustion can be prevented by stuffing data such that buffer depletion does not occur even if a minimum bit is generated in the current macroblock before MB encoding. But. In MPEG-2, you can only stuff before "start_code", so you can't stuff between MB and MB. Therefore, in MPEG-2, slice headers starting with "start_code" must be generated for each MB in order to stuff in MB. If the entire picture has such a slice structure (slice includes only one MB), the overhead of the slice header generation increases.

The solution is to encode a fixed slice structure determined by the application of the encoder system in the steady state, and then code the MB only when there is a concern of depletion of the buffer, and start the subsequent MB from the slice header. In the present embodiment, as a method of preventing the buffer exhaustion as described above, specific contents are as follows.

First, a buffer depletion warning signal is generated according to the state of the channel buffer for each system clock sysclk. In this embodiment, the buffer depletion warning signal value is set to 1 if the buffer status value is small based on the number of bits exiting from the MB processing period in hardware, and to 0 otherwise.

The VLC that encodes MB data according to "mbclk" performs stuffing when the value is 1 according to the "underflow" flag until the next "mbclk" when the MB is encoded. At this time, when "mbclk" is met, if stuffing has not been performed since the previous MB, normal operation is performed, and if stuffing is performed, a slice header is planted and MB is encoded. In the MB encoding section, the process is performed as described above. When encoding a sequence, a group of picture (GOP), and a picture header, each header is encoded and stuffed according to an "underflow flag".

In the present embodiment, the reference value for generating the "underflow" signal may be obtained from the occurrence count between "mbclk" of "cstrm_en (which indicates the state exiting from the channel buffer)" input to obtain the buffer state in FIG. In addition, the present invention requires the VLC unit to secure an extra clock that can be stuffed after MB encoding in " mbclk " The amount of stuffing to be generated.

The present invention described above is capable of various substitutions, modifications, and changes without departing from the spirit of the present invention for those skilled in the art to which the present invention pertains, and the above-described embodiments and accompanying It is not limited to the drawing.

As described above, the present invention has an effect of ensuring channel buffer exhaustion prevention and buffer overflow prevention (up to 1Mbps or less) by minimizing deterioration in encoding quality without increasing hardware complexity in an MPEG video encoder.

Claims (8)

In the video video encoder to control the buffer, A first step of controlling bit generation in the bit rate control mode during normal operation, and generating a buffer overflow concern signal according to a buffer overflow concern when there is a fear of buffer overflow; And A second step of controlling a bit rate control operation, a macroblock (MB) coding mode, and a discrete cosine transform (DCT) quantization operation according to the buffer overflow concern signal to prevent buffer overflow Buffer control method of the video video encoder comprising a. The method of claim 1, A third step of encoding the buffer in a predetermined slice structure in a steady state and generating a buffer depletion warning signal when the buffer is depleted; And After macroblock (MB) encoding, a fourth step of preventing buffer depletion by stuffing according to the buffer depletion warning signal and starting the following macroblock (MB) from a slice header The buffer control method of the video video encoder further comprising. The method according to claim 1 or 2, The buffer overflow concern signal, Substantially, the buffer control method of a video video encoder, characterized in that it comprises a buffer overflow warning signal, a buffer overflow warning signal, and a buffer overflow danger signal. In the video video encoder to control the buffer, A first step of encoding the buffer in a predetermined slice structure in a steady state and generating a buffer depletion warning signal when the buffer is depleted; And After macroblock (MB) encoding, a second step of preventing buffer depletion by stuffing according to the buffer depletion warning signal and starting the following macroblock (MB) from a slice header Buffer control method of the video video encoder comprising a. The method of claim 4, wherein The buffer depletion warning signal is, A method for controlling a buffer of a video video encoder, characterized in that it is determined based on the number of bits exiting according to a state of a channel buffer in a macroblock (MB) processing period. A video video encoder with a processor, A first function of controlling bit generation in the bit rate control mode during normal operation, and generating a buffer overflow concern signal according to a buffer overflow concern in the case where there is a fear of buffer overflow; And A second function of controlling a bit rate control operation, a macroblock (MB) coding mode, and a discrete cosine transform (DCT) quantization operation according to the buffer overflow concern signal to prevent buffer overflow A computer-readable recording medium having recorded thereon a program for realizing this. The method of claim 6, A third function of encoding the buffer in a predetermined slice structure in a steady state and generating a buffer depletion warning signal when the buffer is depleted; And After macroblock (MB) encoding, a fourth function of stuffing according to the buffer depletion warning signal and starting the following macroblock (MB) from a slice header to prevent buffer exhaustion A computer-readable recording medium that records a program for further realization. A video video encoder with a processor, Encoding a buffer in a predetermined slice structure in a steady state and generating a buffer depletion warning signal when the buffer is depleted; And After macroblock (MB) encoding, stuffing is performed according to the buffer depletion warning signal and a subsequent macroblock (MB) starts from a slice header to prevent buffer depletion. A computer-readable recording medium having recorded thereon a program for realizing this.