KR100793781B1 - Apparatus and method for encoding of real-time moving image - Google Patents

Abstract

The present invention relates to an apparatus and method for encoding an analog broadcast signal in real time for recording an analog broadcast in a PVR device capable of recording / reproducing a broadcast signal received by mounting a storage medium in a broadcast reception device on the storage medium. . In particular, the present invention analyzes the number of non-zero DCT coefficients in real time for each GOP section of the video and allocates an appropriate bit amount to the corresponding section, thereby providing an optimal MPEG-2 video bit stream in terms of both image quality and bit efficiency. to provide. In addition, we propose an R-N model that models the relationship between the actual amount of bits and non-zero DCT coefficients and perform more accurate bit rate control using this model and Q-N histogram, resulting in excellent image quality and bit efficiency.

Real-time video coding, R-N model, non zero DCT coefficients

Description

Apparatus and method for encoding of real-time moving image}

1 is a block diagram showing an embodiment of analog and digital broadcast recording using a conventional PVR device

2 is a block diagram showing an embodiment of a general MPEG-2 encoding apparatus

3 is a detailed block diagram illustrating an embodiment of a bit rate control unit of a single pass variable bit rate MPEG-2 encoding apparatus according to the present invention;

4 is a diagram illustrating an embodiment of a function of determining a bit rate for each section based on the number of non-zero DCT coefficients generated by a reference quantization parameter in the MPEG-2 encoding apparatus according to the present invention.

5 is a flowchart illustrating an embodiment of a quantization parameter determination process in an MPEG-2 encoding apparatus according to the present invention.

Explanation of symbols for main parts of the drawings

201: subtractor 202,213: switching unit

203: DCT unit 204: quantization unit

205: VLC section 206: buffer

207: Inverse quantization unit 208: IDCT unit

209: Adder 210: Frame Memory

211: motion compensation unit 212: motion estimation unit

300: bit rate control unit 311: quantizer

312: counter 313: bit rate determining unit

The present invention relates to a digital broadcast receiver, and more particularly, to recording analog broadcasts in a personal video recording apparatus capable of recording / reproducing a broadcast signal received by mounting a storage medium in the broadcast receiving device. An apparatus and method for encoding an analog broadcast signal in real time.

A personal video recorder (hereinafter referred to as a PVR device) provides a function of storing a digital broadcast on an HDD by attaching a hard disk (HDD) to a digital TV or a set-top box. This eliminates the inconvenience that the existing VCR (Video Cassette Recorder) has to replace the video tape at any time for recording, and at the same time enables a large amount of broadcast recording. In addition, the PVR device provides not only a recording function, which is a main function, but also a time-shift, instant replay, and other trick play functions. To provide.

In such a PVR device, a real time MPEG-2 video encoding device is required to convert an NTSC or PAL analog broadcast signal into a moving picture expert group (MPEG) -2 video stream and record it on an HDD.

1 illustrates a path for storing analog and digital broadcasts in an HDD in a PVR device. An analog broadcast signal such as NTSC or PAL is tuned in the analog tuner 101 and then output to the A / D converter 102 to be digitized.

In FIG. 1, an analog broadcast signal of the NTSC transmission method is described as an embodiment.

Therefore, the analog broadcast signal digitized by the A / D converter 102 is output to the NTSC decoder 103. The NTSC decoder 103 converts an analog broadcast signal into a 4: 2: 0 format for encoding into MPEG-2 and outputs it to the MPEG-2 encoder 104. The MPEG-2 encoder 104 compresses and encodes the format-converted broadcast signal by the MPEG-2 compression algorithm as shown in FIG. 2 and outputs the encoded signal to the PVR module 108.

Meanwhile, the digital broadcast signal is tuned by the digital tuner 105 and then output to the demodulator 106. The demodulator 106 demodulates the inverse of the modulation scheme for the tuned broadcast signal. For example, if the received digital broadcast is a terrestrial broadcast, a VSB (Vestigial Side Band) demodulation is performed, and if the satellite broadcast is a Quadrature Phase Shift Keying (QPSK) demodulation, a TS decoder in the form of a transport stream (TS) is performed. Output to (107). The TS decoder 107 demultiplexes only a desired program and outputs it to the PVR module 108 when a plurality of programs are multiplexed in an input transport stream.

The PVR module 108 multiplexes the data stream including the audio stream and caption information to the analog broadcast stream compressed and encoded by the MPEG-2 encoder 104 or the digital broadcast stream demultiplexed by the TS decoder 107 together. After multiplexing) to store in the HDD (110). In this case, the audio stream and the data stream may also be encoded using a corresponding encoding device.

FIG. 2 is a block diagram illustrating a typical MPEG-2 encoder, in which encoding is performed on a macroblock basis. That is, if the image input in macroblock units is an intra type, that is, an I picture, it is output to the DCT unit 203 through the switching unit 202 without passing through the subtractor 201. On the contrary, if the input image is an inter type, that is, a P or B picture, the difference image obtained by the subtractor 201, that is, the difference image between the input image and the motion compensated image by the motion compensation MC 212. It outputs to the DCT unit 203 through this switching unit 202.

That is, the signal input to the DCT unit 203 means a residual image signal based on a DC value in the case of an I picture, and a residual image generated during motion estimation in the case of a P or B picture. Means signal.

The output terminal c of the switching unit 202 is switched to the input image terminal a or the output terminal b of the subtractor 201 according to whether the type of the input image is intra or inter.

The DCT unit 203 DCTs the input data in block units and outputs the result to the quantization unit 204. The quantization unit 204 quantizes the DCT coefficient according to the quantization step size determined by the quantization parameter. In this case, the quantization parameter affecting the overall bit rate and the image quality is determined by the bit rate controller 220.

That is, the DCT unit 203 removes the correlation of data through 2D axis transformation. To this end, the picture is divided into macroblock units, and each 16 * 16 macroblock is divided into 8 * 8 block units. After dividing, each divided block is transformed according to DCT equation. The axis-converted data are quantized to a predetermined quantization step size by the quantization unit 204 and rearranged in one dimension according to a predetermined scan method, and then output to the variable length coding (VLC) unit 205. The VLC unit 205 performs entropy coding to reduce the total number of bits by displaying a number of bits that appear frequently and a number of bits that rarely appear for the rearranged quantized DCT coefficients. In this case, not only the DCT coefficient but also picture header information and motion information are encoded.

At this time, the VLC data from the VLC unit 205 is output to the buffer 206, and the buffer 206 temporarily stores the VLC data and then passes through the PVR module 108 to the HDD 110 at a constant speed. The buffer fullness of the buffer is calculated and output to the bit rate controller 220.

The DCT coefficients quantized by the quantization unit 204 are again input to the inverse quantization unit (IQ) unit 207, dequantized, and then output to the IDCT unit 208. The IDCT unit 208 IDCTs the inverse quantized DCT coefficients and outputs them to the adder 209. When the IDCT data is an intra type, the adder 209 is added to the motion compensation unit 212 when the IDCT data is an intra type. ).

The motion compensator 212 performs motion compensation using a previous frame read from the frame memory 210 according to the estimation result of the motion estimator ME 211 and then switches with the subtractor 201. Output to the unit 213. If the IDCT data is intra type, the switching unit 213 is turned off, and the motion compensated data from the motion compensator 212 is not output to the adder 209. If the IDCT data is an inter type, the switching unit 213 is turned on, and the motion compensated data in the motion compensator 212 is output to the adder 209 through the switching unit 213.

The motion estimator 211 estimates a motion vector using the input image and the reference image stored in the frame memory 210 and outputs the motion vector to the motion compensator 212.

On the other hand, the most important thing in the real-time MPEG-2 video encoding process is the bit rate and image quality. These two characteristics are trade-off relations with each other, and the rate control is a process for achieving the best result by combining the two characteristics properly.

The bit rate control is a process of setting a buffer model based on a given bit rate and assigning a GOP (Group Of Picture) of video data and a bit amount required for the encoding process in each frame unit.

That is, the bit rate controller 220 calculates the quantization parameter Qp according to the fullness of the buffer 206 and outputs the quantization parameter Qp to the quantization unit 204.

The quantization unit 204 quantizes the DCT coefficient by obtaining the quantization step size from the input quantization parameter Qp. That is, the bit rate controller 220 adjusts the generation amount of data by varying the step size of the quantization unit 204 according to the fullness of the buffer.

For example, if the number of bits generated is greater than or equal to the reference value, the amount of data filled in the buffer 206 is increased. Therefore, the quantization step size is increased to reduce the number of bits to be generated later. The number of bits generated is increased to adjust the state of the buffer 206 to maintain a constant value.

In this case, referring to the international standardization MPEG-2 data (document number AVC-491, TEST MODEL 5), IS / IEC JTC1 / SC29 / WG11, an organization under ISO (International Organization for Standardization), the bit rate controller 220 Perform the following three steps.

First, the first step is to predict complexity and to allocate target bits. That is, a constant bit rate is allocated in units of GOP (Group Of Picture) according to the transmission bit rate, and bits to be allocated to each picture in the GOP are allocated according to the complexity of each picture (I, P, B frame).

The second step is to adjust the transmission rate (i.e., bit rate), and calculate a reference quantization parameter for each macro block according to the fullness of the virtual buffer 206. The bit rate is controlled so that each picture can be encoded according to the bit allocated in the first step.

Here, it is assumed that each picture has an arbitrary virtual buffer, and a method of adjusting the quantization parameter according to the state of the buffer is used.

The third step is an adaptive quantization step. After obtaining and normalizing the activity of the macroblock to be encoded currently, the normalized activity is multiplied by the reference quantization parameter obtained in the second step to obtain a quantization parameter to be actually used for quantization. In other words, adaptive quantization is a method of increasing the subjective picture quality and changing the reference quantization parameter according to the complexity of the current macroblock.

However, the conventional bit rate control method has a disadvantage in that image quality is not optimized satisfactorily by focusing on an effective allocation of a given bit amount without systematic analysis of video complexity. In particular, in the MPEG-2 video encoding process, the deterioration of image quality tends to be intensified for a section composed of an image having a lot of motion or a visually complicated structure. In addition, the degree of deterioration of image quality felt by the image quality imbalance is more pronounced than the difference in numerical value.

Therefore, the realization of uniform picture quality in MPEG-2 video encoding is a very important problem. In particular, it is very important to ensure satisfactory image quality in a complicated area, that is, a difficult part of encoding. Many variable bit rate schemes have been proposed for realizing such a uniform picture quality, but these methods have a problem in that quantitative and systematic measurement methods for image complexity are insufficient.

In addition, the biggest problem of the conventional variable bit rate control method is that there is almost no real-time prediction method for the amount of bits occurring.

Therefore, in the conventional method, bit amount control by the two-pass method is performed. That is, a method of controlling the amount of bits by performing encoding once and then performing encoding again using the result is used. However, there is a problem that such a method cannot be applied to an application field requiring real time encoding.

In the conventional method, the quantization parameter value is changed based on each complexity in units of macro blocks. However, frequent changes of quantization parameter values of adjacent macroblocks have a disadvantage in that performance is degraded in terms of bit efficiency.

The present invention is to solve the above problems, an object of the present invention is to quantitatively measure the complexity of each video section in real time to be applied to the real-time broadcast recording required by the PVR device in real time to determine the appropriate bit amount for the corresponding section A one-pass variable bit-rate MPEG-2 encoder and a method for allocating the same are provided.

Another object of the present invention is to provide an apparatus and method for predicting the actual amount of bits to be generated by using data obtained in the process of quantization of video data for accurate bit rate control by a single pass coding process.

In order to achieve the above object, the variable bit rate MPEG-2 coding apparatus according to the present invention comprises: a pre-quantization device for linearly quantizing a DCT coefficient of a current macro block using a predetermined reference quantization parameter value; A counter for counting and outputting a number of non-zero DCT coefficients among the pre-quantized DCT coefficients of the current macro block; And determining a target bit amount for a macroblock to be currently encoded based on the number of non-zero DCT coefficients counted by the counter, and modeling a relationship between the bit amount and the number of DCT coefficients as the target bit amount of the macroblock. A bit rate determination unit configured to determine the number of non-zero DCT coefficients for the target bit amount of the macro block by applying to an RN model, and to determine the quantization parameter of the macro block based on the number of non-zero DCT coefficients. Characterized in that the configuration.

The bit rate determining unit allocates a target bit amount for the current GOP based on the number of pre-quantized non-zero DCT coefficients, and allocates a target bit amount for each picture type (I, P, B) from the determined target bit amount. After assigning the target bit amount of the macro block to be currently encoded from the target bit amount in one picture, applying the same to the RN model equation to determine the number of non-zero DCT coefficients for the target bit amount of the macro block. It is characterized by.

The bit rate determining unit allocates a target bit amount for the current GOP by using an average of the number of non-zero DCT coefficients calculated as reference quantization parameters for a predetermined amount of video interval previously encoded.

The bit rate determining unit defines an RN model equation (R = α * N _NZC + β) that models a relationship between the target bit quantity and the number of non-zero DCT coefficients, and includes a bit quantity for previously encoded data. Model coefficients α and β are obtained using statistics of the number of non-zero DCT coefficients, and the number of non-zero DCT coefficients for the target bit amount R of the macroblock is substituted by substituting the target bit amount R of the macroblock N _NZC Characterized in determining.

The bit rate control unit updates the RN model equation by using the number of bits obtained in the actual encoding process and the number of non-zero DCT coefficients when the distance between the target bit amount and the bit amount generated during actual encoding differs by a predetermined value or more. Characterized in that.

The real-time video encoding method according to the present invention,

(a) pre-quantizing a DCT coefficient of the current macro block using a preset reference quantization parameter value;

(b) counting and outputting the number of non-zero DCT coefficients among the pre-quantized DCT coefficients of the current macroblock;

(c) Allocating a target bit amount for the current GOP based on the number of non-zero DCT coefficients counted in step (b), and targeting bits for each picture type (I, P, B) from the determined target bit amount. Allocating a quantity, and allocating a target bit quantity of a macroblock to be currently encoded from a target bit quantity in one picture;

(d) a non-zero DCT coefficient for the target bit amount of the macro block by applying the target bit amount of the macro block allocated in step (c) to an RN model equation modeling the relationship between the bit amount and the number of DCT coefficients. Determining the number of; And

(e) determining the quantization parameter of the macroblock based on the number of non-zero DCT coefficients determined in step (d).

In the step (e), a plurality of quantization parameters designated as candidates are respectively applied to perform quantization on a macroblock to be currently encoded, and then QN histogram data of the number of non-zero DCT coefficients for each candidate quantization parameter is extracted. And a quantization parameter for generating a value closest to the number of non-zero DCT coefficients determined in step (d) among the candidate quantization parameters as a quantization parameter for a macroblock to be currently encoded. It is done.

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, with reference to the accompanying drawings, preferred embodiments of the present invention that can specifically realize the above object will be described. At this time, the configuration and operation of the present invention shown in the drawings and described by it will be described as at least one embodiment, by which the technical spirit of the present invention and its core configuration and operation is not limited.

The same components as in the related art are denoted by the same names and the same reference numerals for convenience of description, and detailed description thereof will be omitted.

The present invention is to real-time encode an analog broadcast signal (e.g., NTSC transmission scheme) into an MPEG-2 video stream with optimal picture quality at a given bit rate. To do this, the complexity of each scene is calculated, and then the correlation between the complexity and the bit generation is analyzed and bit allocation is performed in an adaptive manner according to each section, thereby providing a high quality video output to the user and at the same time effectively reducing the bit rate of the video data. Can be used to effectively manage storage space.

As described above, the present invention implements the MPEG-2 encoder by using a one-pass algorithm so that it can be applied to real-time broadcast recording required by the PVR device, thereby maintaining uniform picture quality and simultaneously encoding video data. The amount of bits can be allocated by effectively analyzing the complexity of the

3 is a block diagram illustrating a structure of a single pass variable bit rate MPEG-2 video encoding apparatus according to an exemplary embodiment of the present invention, wherein the bit rate control unit 300 is different in the general MPEG-2 encoder of FIG. 2.

The bit rate control unit 300 is used to determine a bit amount for a current section or a group of picture (GOP), and receives a DCT coefficient from the DCT unit 203 to perform pre-quantization. 311, a counter 312 that counts non-zero DCT coefficients among the DCT coefficients quantized by the line quantizer 311, and the count result of the counter 312 and the fullness of the buffer 206. To determine the final quantization parameter and output it to the quantization unit 204.

The linear quantization of the DCT coefficients in the line quantizer 311 predicts the bit generation amount of video data to be encoded currently, thereby quantitatively measuring the complexity of the corresponding section, and allocating an appropriate bit for the current video section. To do this. That is, the linear quantizer 311 has a value of a preset reference quantization parameter Q _ref when the range of the quantization parameter Q value is 2 to 62 (Q has an even value) in the linear quantization process. Next, quantization is performed on the DCT coefficients of the current macroblock (2 ≦ Q _ref ≦ 62).

The counter 312 counts the number of non-zero DCT coefficients among the quantized DCT coefficients of the current macro block and outputs them to the bit rate determination unit 313. The number of non-zero DCT coefficients has a property that is proportional to the complexity and bit generation amount of the macroblock.

In the present invention, for convenience of description, a non-zero DCT coefficient will be referred to as a non-zero DCT coefficient. At this time, a large number of non-zero DCT coefficients means that the difference video signal of the macro block is large. This means that the spatial structure of the picture is very complicated for the I picture, and the motion estimation error is large for the P or B picture. That is, a large motion estimation error means that there is a complex motion or a very large motion in the video data. This means that there is video data that is very difficult to encode relatively. In addition, since the quantized DCT coefficient having a value of zero does not have a significant amount of bit generation in comparison to non-zero DCT coefficients in entropy coding, the control of the number of non-zero DCT coefficients in terms of the actual bit generation is in terms of bit rate control. Has a great meaning.

For this reason, the present invention assumes that the bit rate R is proportional to the number N _NZC of the non-zero DCT coefficients as shown in Equation 1 below.

R ∝N _NZC

In Equation 1, R denotes the amount of bits and is represented by the number of bits generated per second. The N _NZC means the number of non-zero DCT coefficients generated by the currently set quantization parameter.

In this case, the larger the quantization parameter is, the truncated coefficient of the high frequency component is truncated to 0. As a result, the larger the quantization parameter is, the larger the N _NZC is. A large quantization parameter means that image quality deterioration occurs, especially when a bit of the same size as another section is allocated in a complex section. As a result, the image quality can be maintained at a certain level by allocating more bits for this complex section.

In more detail, Equation 1 may be expressed in more detail. The relationship between the actual amount of bits and the non-zero DCT coefficient may be represented by a first-order linear equation as in Equation 2 below.

R = α * N _NZC + β

In the present invention, Equation 2 is defined as an R-N model equation.

In Equation 2, R denotes a bit amount, a unit is bit, and N _NZC denotes the number of non-zero DCT coefficients generated in one macroblock when quantization is performed using a reference quantization parameter.

The advantage of encoding by the variable bit rate is that the overall quality level can be maintained by allocating more bits to a complex image and allocating a small amount of bits in a relatively low complexity section. In this case, since the complexity of each section can also be inferred by the number of non-zero DCT coefficients, the target bit amount for the scene section to be encoded is variably allocated according to Equation 3 below according to the number of non-zero DCT coefficients. Do it.

R _local ∝ AVERAGE (N _QREF )

In Equation 3, R _local means a target bit amount for the current interval, and N _QREF means the number of non-zero DCT coefficients for the reference quantization parameter. That is, the target bit amount is allocated using an average of N _QREFs calculated for a predetermined amount of video intervals previously encoded.

In this case, the reference quantization parameter uses the same value in all video encoding processes, and in the embodiment of the present invention, the smallest value is selected from all currently available quantization parameter values.

As a result, N _QREF represents the largest value among the number of non-zero DCT coefficients that can occur in a specific coding unit such as a macroblock. For example, when linear quantization is used in MPEG-2 encoding, the reference quantization parameter value is 2.

In this case, the R _local is allocated to be proportional to the average of N _QREF or a discontinuous function is used to determine the target bit amount of the corresponding section, for example, a GOP.

As described above, the present invention can perform quantitative measurement on the complexity of the corresponding GOP interval by using the number of DCT coefficients generated through the line quantization process.

4 illustrates a function modeling a relationship between R _local and N _QREF used in the single pass variable bit rate MPEG-2 encoding apparatus according to the present invention. In FIG. 4, R _global denotes an overall target bit rate for video to be recorded. That is, in the present invention, a larger bit rate is allocated to a section having a high complexity, and an increase in the bit rate generated at this time is solved by reducing the bit rate of a section having a low complexity.

As a result, after measuring the number N _QREF of the non-zero DCT coefficients, the amount of bits to be allocated to the corresponding interval is determined using the function defined by FIG. 4.

In general, image quality deterioration occurring in an uncomplicated region is not visually significant compared to image quality deterioration occurring in a complex region. Therefore, it is possible to reduce the bit consumption in the uncomplicated region and use it for encoding the complex region, thereby improving the overall visual quality.

5 is a flowchart illustrating a quantization parameter calculation process for bit rate control in a single pass variable bit rate MPEG-2 encoding apparatus according to the present invention.

That is, in step 501, a target bit amount for the current section is allocated. The target bit amount is to first determine the bit amount R _GOP for the current interval, that is, the current GOP. At this time, as described above, the target bit amount for the current GOP is determined using a function defined in the form of FIG. 4 and Equation 3 based on the number of pre-quantized non-zero DCT coefficients. The number of non-zero DCT coefficients uses the number N _QREF of the non-zero DCT coefficients generated when the quantization parameter Q value is used as the reference quantization parameter Q _ref in the dequantization process.

Then, the bit amount to be allocated is determined according to each picture type (I, P, B). As a result, a relationship as shown in Equation 4 below is established.

R _GOP = R _I + R _P + R _B

In this case, R _I , R _P , and R _B are bits used by I, P, and B pictures in the current GOP, respectively. In other words, R _P means a sum of bits allocated to all P pictures included in the current GOP. As a result, the bit amount allocated to one P picture is obtained by dividing R _P by N _P , where N _P represents the number of P pictures in the current GOP. The same applies to I and B pictures.

In order to comply with the allocated target bit amount for each picture, an appropriate quantization parameter value should be set. For accurate bit control, the R-N model equation defined by Equation 2 is used.

Using the R-N model equation, it is possible to determine the number N of non-zero DCT coefficients for complying with a specific bit amount R. Therefore, first, it is necessary to calculate the model coefficients (α, β) of Equation 2 in which the number of bits R and the number of non-zero DCT coefficients are modeled.

The update of the R-N model equation is performed only when a change in the characteristics of the video occurs and the existing R-N model equation can no longer adequately express the relationship between R and N.

The bit amount for calculating the RN model coefficient may be measured based on a unit such as a macro block or a slice, and the N _{NZC is} also measured according to a corresponding video coding unit (step 502).

 In an embodiment, when the bit rate R and the non-zero DCT coefficients are measured in macroblock units to complete the RN model of Equation 2, the measured values of the N macroblocks already encoded are It can be developed as shown in Equation 5.

In this case, R _MBj and N _MBj are the number of bits and non-zero coefficients generated by the quantization parameter used in the encoding process for the j-th macroblock, respectively.

Equation 5 may be represented again in a matrix form as shown in Equation 6 below.

In order to complete the R-N model of Equation 2, the model coefficients (α, β) are calculated by the linear regression method.

As a result, the model coefficients α and β have the same values as in Equation 8 below.

When the model coefficients (α, β) are calculated as shown in Equation 8, the RN model equation is completed, and the number of non-zero DCT coefficients generated by the quantization parameter currently set in the bit rate control process of the MPEG-2 encoding apparatus is used. It is possible to accurately predict the amount of bits generated by using. At this time, the number of quantization parameters and DCT coefficients is determined by extracting a Q-N histogram.

The R-N model formula may perform an update when the separation between the target bit amount and the actual bit amount is 15% or more. For example, when the j-th frame is encoded, when the difference between the target bit amount and the bit amount actually used to encode the j-th frame is 15% or more, the bit amount and the DCT obtained in the process of encoding the j-th frame are different. Update the RN model equation using the number of coefficients.

In addition, since the previous data does not exist when the video is first encoded, the initial values of the model coefficients (α, β) are set to (0.789x10 ⁴ , 7.56x10 ⁴ ) using the results obtained by the experiment.

In addition, separate R-N model equations may be set according to picture types for more effective and accurate results.

When the target bit amount for each picture, i.e., a frame, is determined by the above-described process, an adaptive bit allocation operation is performed in units of macroblocks within each frame. Bit allocation in units of macroblocks is shown in Equation 9 below.

In Equation 9, MB_CNT denotes the number of remaining macroblocks including the current macroblock in the frame currently being encoded, R _frame is the amount of bits allocated to the current frame, and R _used is the current frame encoding. The bit amount used so far in the process. N _norm means the number of normalized non-zero DCT coefficients and is defined as in Equation 10 below.

In the process of obtaining the number of normalized non-zero DCT coefficients as shown in Equation 10, N _avg is an average of the number of non-zero DCT coefficients for the reference quantization parameter in the previous frame, and N _cur is the reference quantization of the current macroblock. Number of non-zero DCT coefficients for the parameter. That is, according to Equation 10, more bits are allocated to macroblocks having a relatively high complexity even in the same frame.

The number N of non-zero DCT coefficients for R is determined by substituting the finally set target bit amount R into the R-N model equation (step 503). Once N is determined, the corresponding quantization parameter Q must be determined.

To this end, the present invention performs a linear quantization by a hardware or software method to extract a histogram for the number of non-zero DCT coefficients according to each quantization parameter (hereinafter referred to as Q-N histogram) (step 504). The Q-N histogram is obtained by performing linear quantization on all candidate quantization parameters for the macroblock to be encoded and extracting the number of DCT coefficients generated for each.

By determining an appropriate quantization parameter Q value corresponding to N determined by using the QN histogram obtained in the above-described quantization process, the quantization parameter may be determined and encoded to satisfy the given bit rate control condition by encoding the corresponding video data ( Step 505).

The process of finally determining the quantization parameter in the Q-N histogram is shown in Equation 11 below.

In Equation 11, Q _final denotes a final quantization parameter for a macroblock to be currently encoded, and H (Qi) denotes a histogram of a candidate quantization parameter Qi, that is, a macroblock generated when prequantization is performed using Qi. The sum of the number of non-zero DCT coefficients And N _target means the target number of non-zero DCT coefficients for the current macroblock determined by the RN model equation.

In this case, the unit for encoding by applying the Q-N histogram basically uses a macroblock and can be extended to a slice.

In the case of the application in the slice unit, the Q-N histogram uses the sum of the number of non-zero DCT coefficients of all macroblocks in the slice, and all macroblocks in the slice apply the same quantization parameter Q value.

In addition, Equations 9 and 10 for adaptive bit allocation perform operations in units of slices.

The final quantization parameter determined in step 505 is input to the quantization unit 204. The quantization unit 204 obtains a quantization step size from an input final quantization parameter and quantizes a DCT coefficient to adjust the amount of data generation.

As described above, when the target bit amount is determined, the number N of non-zero DCT coefficients corresponding to the target bit amount is determined using the R-N model equation. Subsequent substitution of the determined N with the Q-N histogram extracted by the prequantizer determines the appropriate quantization parameter Q value, thereby enabling accurate bit control.

As described above, the present invention implements the MPEG-2 encoder by using a one-pass algorithm so that it can be applied to real-time broadcast recording required by the PVR device, thereby maintaining uniform image quality and simultaneously encoding video data. The amount of bits can be allocated by effectively analyzing the complexity of the

Accordingly, the present invention enables real-time encoding of an analog broadcast signal (e.g., NTSC transmission scheme) into an MPEG-2 video stream with optimal picture quality at a given bit rate.

Therefore, the present invention can be applied to the field of encoding and storing analog broadcasts such as NTSC and PAL as a digital stream in a PVR device equipped with a hard disk in a digital TV, and besides, an application requiring a real-time variable bit rate MPEG-2 encoding device. Applicable to the field.

The terminologies used in the present invention are terms defined in consideration of functions in the present invention and may vary according to the intention or customs of those skilled in the art, and the definitions refer to the overall contents of the present invention. It must be laid down.

The present invention is not limited to the above-described embodiments, and can be modified by those skilled in the art as can be seen from the appended claims, and such modifications are within the scope of the present invention.

The effects of the real-time video encoding apparatus and method according to the present invention described above are as follows.                     

First, in an application field for recording analog broadcasting in real time in a PVR device, by analyzing the number of non-zero DCT coefficients in real time for each section of video and allocating an appropriate bit amount for the corresponding section, In terms of aspect, it provides an optimal MPEG-2 video bit stream.

Second, we propose an R-N model that models the relationship between the actual amount of bits and non-zero DCT coefficients, and perform more accurate bit rate control using this model and Q-N histogram, resulting in excellent image quality and bit efficiency.

Third, since the model equation can be updated using the number of bits and the number of non-zero DCT coefficients, it is possible to prepare for sudden changes in the screen or other unexpected situations that may occur during bit rate tuning.

Fourth, it can be applied to the field of encoding and storing analog broadcasting such as NTSC and PAL as digital stream in personal digital video recorder equipped with hard disk in digital TV. In addition, applications requiring real-time variable bit rate MPEG-2 encoder Applicable to the field.

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.

Claims (14)

A real-time video encoding apparatus for DCT transforming an image input in macroblock units, and then quantizing a DCT coefficient according to a quantization parameter and then compressing and encoding an input image by variable length coding (VLC). A pre-quantization device for linearly quantizing a DCT coefficient of the current macro block using a preset reference quantization parameter value; A counter for counting and outputting a number of non-zero DCT coefficients among the pre-quantized DCT coefficients of the current macro block; And Based on the number of non-zero DCT coefficients counted by the counter, the target bit amount for the macroblock to be currently encoded is determined, and the target bit amount of the macroblock is modeled by the relationship between the bit amount and the number of DCT coefficients. Including a bit rate determiner for determining the number of non-zero DCT coefficients for the target bit amount of the macro block by applying to an RN model equation and then determines the quantization parameter of the macro block based on the number of non-zero DCT coefficients. Real-time video encoding apparatus characterized in that the configuration. The method of claim 1, wherein the bit rate determination unit Allocates a target bit amount for the current GOP based on the number of pre-quantized non-zero DCT coefficients, and allocates a target bit amount for each picture type (I, P, B) from the determined target bit amount, and one picture. Assigning the target bit amount of the macroblock to be currently encoded from the target bit amount, and applying the same to the RN model to determine the number of non-zero DCT coefficients for the target bit amount of the macroblock. Real time video encoding device. The method of claim 2, wherein the bit rate determination unit And a target bit amount for the current GOP is allocated by using an average of the number of non-zero DCT coefficients calculated as reference quantization parameters for a predetermined amount of video interval previously encoded. The method of claim 2, wherein the bit rate determination unit And assigning a target bit amount R _MB of a macroblock in the picture by applying the following equation when a target bit amount for a picture to be currently encoded is allocated.

The MB_CNT is the number of remaining macroblocks including the current macroblock in the picture currently being encoded, R _frame is the amount of bits allocated to the current frame, and R _used is the bits used so far in encoding the current frame. N _norm means the number of normalized non-zero DCT coefficients. The method of claim 4, wherein the bit rate determination unit Real-time video encoding apparatus characterized in that the number of non-zero DCT coefficients are normalized by applying the following equation.

N _avg is an average of the number of non-zero DCT coefficients for the reference quantization parameter in the previous picture, and N _cur is the number of non-zero DCT coefficients for the reference quantization parameter of the current macroblock. The method of claim 1, wherein the bit rate determination unit An R-N model equation that models the relationship between the target bit quantity and the number of non-zero DCT coefficients is defined by the following equation, R = α * N _NZC + β Model coefficients α and β are obtained by using statistics on the amount of bits and non-zero DCT coefficients of previously encoded data, and the target bit amount R of the macro block is substituted by substituting the target bit amount R of the macro block. And a number N _NZC of non-zero DCT coefficients for the real-time video encoding apparatus. The method of claim 6, wherein the bit rate control unit A real-time video encoding apparatus characterized by obtaining the model coefficients α, β by applying the following equation.

R _MBj and N _MBj are the number of bits and non-zero coefficients generated by the quantization parameter used in the encoding process for the j-th macroblock among the N macroblocks (0 to N-1) that are already encoded. The method of claim 6, wherein the bit rate control unit The RN model equation is updated by using the number of bits obtained in the actual encoding process and the number of non-zero DCT coefficients when the distance between the target bit amount and the bit amount generated during actual encoding differs by more than a predetermined value. Real-time video encoding device. The method of claim 1, wherein the bit rate determination unit After pre-quantization is performed on the macroblock to be currently encoded by applying a plurality of quantization parameters designated as candidates in advance, the QN histogram data for the non zero DCT coefficients generated for each is extracted, and then the predetermined among the candidate quantization parameters is determined. and a quantization parameter that generates a value closest to the number of non-zero DCT coefficients as a quantization parameter for a macroblock to be currently encoded. 10. The apparatus of claim 9, wherein the bit rate determining unit The unit for encoding and applying the QN histogram may be extended to a slice. In this case, the QN histogram uses the sum of the number of non-zero DCT coefficients of all the macroblocks in the slice, and all macroblocks in the slice are the same. Apparatus for real-time video encoding, characterized in that for applying a quantization parameter value. In the real-time video encoding method of DCT transform the input image in macroblock unit, and then quantize the DCT coefficient according to the quantization parameter and then compression-coding the input image by variable length coding (VLC), (a) pre-quantizing a DCT coefficient of the current macro block using a preset reference quantization parameter value; (b) counting and outputting the number of non-zero DCT coefficients among the pre-quantized DCT coefficients of the current macroblock; (c) Allocating a target bit amount for the current GOP based on the number of non-zero DCT coefficients counted in step (b), and targeting bits for each picture type (I, P, B) from the determined target bit amount. Allocating a quantity, and allocating a target bit quantity of a macroblock to be currently encoded from a target bit quantity in one picture; (d) a non-zero DCT coefficient for the target bit amount of the macro block by applying the target bit amount of the macro block allocated in step (c) to an RN model equation modeling the relationship between the bit amount and the number of DCT coefficients. Determining the number of; And and (e) determining the quantization parameter of the macroblock based on the number of non-zero DCT coefficients determined in step (d). The method of claim 11, wherein step (d) An R-N model equation that models the relationship between the target bit quantity and the number of non-zero DCT coefficients is defined by the following equation, R = α * N _NZC + β Model coefficients α and β are obtained by using statistics on the amount of bits and non-zero DCT coefficients of previously encoded data, and the target bit amount R of the macro block is substituted by substituting the target bit amount R of the macro block. And determining the number N _NZC of the non-zero DCT coefficients for the real-time video encoding method. The method of claim 12, wherein step (d) Real-time video encoding method characterized by obtaining the model coefficients α, β by applying the following equation.

R _MBj and N _MBj are the number of bits and non-zero coefficients generated by the quantization parameter used in the encoding process for the j-th macroblock among the N macroblocks (0 to N-1) that are already encoded. The method of claim 11, wherein step (e) Extracting Q-N histogram data for the number of non-zero DCT coefficients for each candidate quantization parameter by performing quantization on a macroblock to be currently encoded by applying a plurality of quantization parameters previously designated as candidates; A quantization parameter for generating a value closest to the number of non-zero DCT coefficients determined in step (d) among the candidate quantization parameters is determined as a quantization parameter for a macroblock to be currently encoded. Coding method.

Images