CN1568008A - Motion Image Compression Bit Rate Control Method and Device Using Initial Quantization Scale Estimation - Google Patents

Abstract

A method and apparatus for controlling bit rate of motion picture compression, which performs bit allocation at a picture level using an estimated initial quantization level. According to the present invention, a target bit limit is allocated to the current picture to be coded according to the relationship between the pre-analyzed activity of the current picture and the actual complexity of the previously coded picture of the same type, and once the target bit limit is determined, the average quantization level initial value for coding the current picture is determined according to the total activity of the current picture, the target bit limit and the activity-to-complexity ratio of the previously coded picture of the same type. Under the condition of a predetermined bit quota, by using the above average quantization level initial value, a higher picture quality can be obtained.

Description

Motion Image Compression Bit Rate Control Method and Device Using Initial Quantization Scale Estimation

technical field

The present invention relates to the field of data compression, in particular to a method and device suitable for performing bit rate control with an estimated initial quantization level in a video coding system.

Background technique

As we all know, the electronic communication world is undergoing a digital revolution. The main advantage of representing information digitally is the data bit stream (bitstream) that can be stored, reproduced, transmitted, processed and utilized almost without error. For example, NTSC color video images have 29.97 frames per second, and each frame has about 480 visible scan lines, and each scan line requires about 480 pixels in red, green, and blue colors. , but if each color component is encoded with 8 bits, the resulting bit rate (bitrate) is about 168 megabits per second (Mbits/s), so the uncompressed bit rate of various video formats is very high and Uneconomical and therefore unsuitable for most applications.

The integration of digital audio and video with computers, telecommunication networks, and consumer products has further stimulated the progress of the information revolution, and the core of this revolution is the digital compression technology of audio and video. Many compression standards include algorithms based on the common essence of compression technology, such as: ITU-T (formerly known as CCITT) recommendations H.261 and H.263, and ISO/IEC MPEG-1, MPEG-2 and MPEG-4 standard. The algorithm of MPEG is developed by the Moving Picture Experts Group (MPEG), which is the International Standards Organization (International Standards Organization, ISO) and the International Electrotechnical Commission (International Electrotechnical Commission, IEC). ) joint technical committee dedicated to the development of video and audio compression and multi-channel representation. These standards stipulate the syntax (syntax) and decoding method of the compressed bit stream, but the algorithm used by the encoder is different However, it retains a considerable degree of freedom for the novelty and diversity of coding.

According to MPEG, a series of video pictures (pictures) are divided into a sequence of picture groups (group of picture, GOP), wherein each group of GOP starts with I-picture, followed by the arrangement of P-picture and B-picture, as shown in Figure 1 The sequence of shows a typical set of GOPs. The encoding of I-pictures does not need to refer to previous or future pictures. P-pictures are coded with reference to the closest I-picture or P-picture in time in a series of video pictures, and B-pictures are interspersed with I-pictures. and between P-pictures. And coded with immediately adjacent I-pictures and P-pictures, previous, future, or both. Although several B-pictures can appear in succession, B-pictures should never predict other pictures.

Each picture has three components: luminance value (luminance, represented by Y), red color difference (red color difference, represented by Cr), and blue color difference (blue color difference, represented by Cb). For the MPEG 4:2:0 format, the sampling points of each Cr and Cb component are only half of the Y component in the horizontal and vertical directions. As shown in FIG. 2 , the basic blocks of an MPEG picture are macroblocks (denoted by MB). Taking 4:2:0 video as an example, each MB contains an array of 16×16 sampling points for the Y component, and two blocks of 8×8 sampling points for the Cr and Cb components, of which the Y component is 16×16 sampling points The array of is actually composed of 4 blocks of 8×8 sampling points.

The role of the encoder is to determine which picture coding type and which prediction mode is the best. For each I-picture, each 8×8 block in the MB undergoes a discrete cosine transform (discrete cosine transform, DCT) to form an 8×8 transformation coefficient array. The transformation coefficients are then quantized with a quantization matrix, and then zigzag (zig-zag) scans the quantized result of the DCT coefficients to obtain a series of DCT coefficients, and the DCT coefficient sequence is coded with a variable length code (variable length code, VLC). A P-picture must decide whether to encode each MB as an I-type MB or a P-type MB. The encoding of the I-type MB is done in the above-mentioned way, and for each P-type MB, it is necessary to obtain the MB that was done in the previous picture. The prediction is obtained by a motion vector, which represents the change between the MB to be coded in the current picture and the predicted MB in the previous picture, and the difference between the predicted MB and the current MB Prediction errors are coded in DCT, quantization, zigzag, and VLC.

When processing B-pictures, one must decide in which of the following MB modes to encode each MB: I mode, F mode, B mode, and FB mode. I mode is coded by MB itself without motion compensation (like I-type MB); F mode is unidirectional forward predictive coding, and the prediction of motion compensation is obtained from the previous picture (like P-type MB); otherwise, The B mode is unidirectional backward predictive coding, and the motion compensation estimation is obtained from the subsequent pictures. In particular, the FB mode is a bi-directional predictive coding, using both forward and backward motion compensation predictions for interpolation to obtain the motion compensation prediction of the FB mode. For F, B and FB modes, the prediction error can be coded using DCT, quantization, zigzag and VLC.

A very important aspect of a video codec is bit rate control. The main purpose of bit rate control is to intelligently allocate the number of bits used to encode each picture and each MB in it, so that when the encoded video is decoded by the decoder, the efficiency of the encoded video can be improved as much as possible. visual quality. The encoder must choose a quantization level for the entire picture in order to control visible distortion at a given bit rate, however, the actual number of bits used to encode a picture at the chosen quantization level must be after the actual encoding Only then will we know that there is no inversion function in reality, which can determine the actual number of bits used by a picture under the given desired quantization level. The key feature of MPEG is the adaptive (or variable) quantization method, which allows different areas in each picture to be coded to different degrees, so that the average and consistent quality can be achieved within each picture and between different pictures. visual quality. However, the traditional bit rate control method is quite complicated, and it usually takes several times to complete the video encoding process. In addition, the previous technology lacks a simple mechanism that can be used by adaptive quantization methods to set The initial quantization level makes the picture quality more average.

Contents of the invention

The present invention provides a dynamic image compression bit rate control technology using initial quantization level estimation, which can be used in a single, real-time video encoder. Furthermore, it is also expected to provide a bit rate control method and device, The estimated initial quantization scale is used to bit-allocate and compress the motion picture at the frame level.

According to the present invention, the allocation of target bits for each picture is based on the encoding results of previous pictures and the pre-analyzed activity of the current picture to be encoded. Once the target bits are allocated, the average quantization level The initial value is also determined accordingly. From the relationship between the pre-analyzed activity of the current picture and the actual complexity of the previously encoded picture, the complexity of the current picture can be estimated. This estimated current picture complexity is very useful for video encoders , can more accurately allocate the target bit quota to each picture. Furthermore, by using the above-mentioned initial value of the average quantization scale, the video encoder can output better picture quality under a given bit quota condition.

The present invention is directed to a motion image compression bit rate control method using initial quantization scale estimation. Firstly, the total motion is calculated for a current frame to be encoded, wherein the current frame is in a group of consecutive frames of a video sequence; then based on the total motion of the current frame and the same type of frames in the group of consecutive frames The ratio of the amount of activity to the complexity of the previous coded picture estimates the complexity of the current picture, and updates the instantaneous complexity of this group of continuous pictures with the estimated complexity of the current picture; the allocation of the target bit quota of the current picture is based on the expected The estimated complexity and instantaneous complexity are determined; the initial value of the average quantization level used to encode the current picture can be determined according to the total activity of the current picture, the target bit amount, and the activity of the previous encoded picture of the same type. The ratio of complexity; therefore, the current frame in the video sequence is encoded based on the average quantization scale initial value. After the current picture is encoded, the ratio of the current picture's activity to complexity is calculated according to the current picture's total activity, the current picture's actual consumption of bits, and the current picture's average quantization level. In this regard, the current The ratio of activity to complexity of a picture is directly proportional to the total activity of the current picture, and inversely proportional to the actual number of consumed bits and the average quantization level of the current picture.

On the other hand, the present invention also discloses a motion image compression bit rate control device using initial quantization level estimation, including a motion estimation unit, a complexity estimator, a quota allocator, and a quantization level estimator and a parameter updater. The motion estimation unit receives a current frame to be encoded in a group of consecutive frames in the video sequence, and is used to calculate the total motion of the current frame when performing block matching for motion estimation. The complexity estimator can thus be used to estimate the complexity of the current picture based on the total activity of the current picture and the activity-to-complexity ratio of previously encoded pictures of the same type in the group of consecutive pictures. Based on the estimated complexity of the current frame, the quota allocator updates the instantaneous complexity of the group of continuous frames, and allocates the target bit quota to the current frame according to the estimated complexity and the instantaneous complexity. The initial value of the average quantization scale used to encode the current picture is based on the total activity of the current picture, the target bit quota, and the activity-to-complexity ratio of previously encoded pictures of the same type by the quantization scale estimator to decide. The parameter updater calculates the ratio of the current picture's activity to the complexity based on the current picture's total activity, the current picture's actual consumption of bits, and the current picture's average quantization level. The ratio of speed is directly proportional to the total activity of the current picture, and inversely proportional to the actual number of consumed bits and the average quantization level of the current picture.

Description of drawings

Figure 1 presents typical groups of pictures (GOPs) in the order shown;

Fig. 2 is the big block of MPEG;

FIG. 3 is a schematic block diagram of a bit rate control device according to a preferred embodiment of the present invention; and

Fig. 4 is an operation flowchart of the present invention.

Description of figure number

300 Motion Image Video Encoder 310 Motion Estimation Unit

312 Signal Line 320 Complexity Estimator

330 Quota Allocator 340 Parameter Updater

350 Quantization Scale Estimator 360 Image Encoder

P ⁽ⁱ⁾ screen A ⁽ⁱ⁾ full activity

C _est ⁽ⁱ⁾ Estimated complexity ACR ⁽ⁱ⁾ Ratio of activity to complexity

EB Overused Bits TB ⁽ⁱ⁾ Target Bit Quota

AQ ⁽ⁱ⁾ Average quantization level UB ⁽ⁱ⁾ Actual number of bits consumed

Q _est ⁽ⁱ⁾ Average quantization scale initial value CD coded data

Detailed ways

In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and understandable, a preferred embodiment is specifically cited below, and in conjunction with the accompanying drawings, the detailed description is as follows:

At the beginning, the I-picture and P-picture used for prediction must be decoded in the MPEG encoder first, and the root mean square (root mean square) error between the original picture and the decoded picture can be calculated to obtain the other parameters of the decoded picture. An objective measure of quality. Taking the root mean square error as the distortion degree of the decoded picture, the complexity (complex)-C of the decoded picture is defined as follows:

C=r×d

Wherein, r is the bit rate required for encoding the picture, and d is the distortion degree of the decoded picture. Theoretically, there is a proportional relationship between the average quantization step size and the degree of distortion. Therefore, the complexity of the picture can be defined as:

$\mathrm{<span\; class="notranslate">\; C</span>}<span\; class="notranslate">\; \&ap;</span>\frac{\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; q</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}$

Wherein, q is the average quantization level of the picture. After the picture is encoded, the bit rate (actual number of bits consumed) and the average quantization level will be known, and the complexity of the picture will be obtained accordingly.

The complexity of the picture depends on the nature of the picture and its encoding form. The prior art uses the complexity of the most recently coded picture to represent the complexity of the current picture, and maintains its complexity for I-, P- and B-pictures respectively to mitigate the impact of different encoding forms, provided that the content of the picture in the video sequence is smooth changes, this method can achieve quite good results. However, due to the discrepancy between the target bit quota of frame coding and the actual consumed bits, if the content of the video sequence changes greatly, the above method is not suitable for achieving consistent visual quality.

According to the present invention, the relationship between the activity of the current picture and the actual complexity of the previously coded picture is pre-analyzed, and the complexity of the current picture can be estimated accordingly. The amount of activity is an attempt to measure the bit rate and quality of a picture after encoding. Assuming that the total amount of activity A of a whole picture is proportional to the complexity C, then

$\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; C</span>}<span\; class="notranslate">\; \&ap;</span>\frac{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; q</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; k</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; q</span>}$

Among them, k is a constant of proportionality. If k' is the activity-to-complexity ratio (activity-to-complexity ratio, ACR), then the total activity A can also be expressed by the following formula:

A=ACR×r×q

According to the present invention, ACR can be estimated from previously coded pictures of the same type. Conceptually, before encoding a picture, its total activity A is calculated, and then the complexity C _est ^{(i ) of the picture P (i)} to be encoded is estimated as follows:

${\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; est</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; A</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; ACR</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}}$

Wherein, A ⁽ⁱ⁾ is the total activity of the picture P ⁽ⁱ⁾ , and ACR ^(il) is the ratio of the activity to the complexity of the previously coded picture of the same type. Hereinafter, a superscript ⁽ⁱ⁾ in a symbol or variable indicates that the symbol or variable is related to the current picture to be encoded; similarly, a superscript ^(il) in a symbol or variable is related to a previously coded picture.

The estimated complexity C _est ⁽ⁱ⁾ can be used to update the complexity of the appropriate type of frame, so the complexities C _I , _CP and C _B of the three types of frames are stored for subsequent processing. The instantaneous complexity of a sequence of pictures is also updated as follows:

INST_C＝N _I ×C _I +N _P ×C _P +N _B ×C _B

Wherein, N _I , N _P and N _B are respectively the numbers of I-, P- and B-pictures in the group of consecutive pictures, and the group of consecutive pictures referred to here includes at least one group of GOP. Once the instantaneous complexity INST_C has been determined, the target bit amount TB ^{(i ) of the current picture P (i)} is

${\mathrm{<span\; class="notranslate">\; TB</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; est</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}}{\mathrm{<span\; class="notranslate">\; INST</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; C</span>}}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; R</span>}}_{\mathrm{<span\; class="notranslate">\; eff</span>}}$

Wherein, R _eff is the effective bit usage of a group of continuous pictures. It can be seen from the above formula that the target bit quota TB ⁽ⁱ⁾ is directly proportional to the complexity C _est ⁽ⁱ⁾ , and inversely proportional to the instantaneous complexity INST_C.

All MPEG-2 data bit streams must follow the Video Buffer Verifier (VBV) rules of the MPEG-2 standard, and the allocated target quota must be limited so that the VBV buffer will not overflow (overflow) or lack (underflow). Originally, the average quantization level of the current picture P ⁽ⁱ⁾ can only be obtained after encoding, and it can be estimated after the target bit amount TB ⁽ⁱ⁾ is determined. According to the present invention, the estimated average of the current picture P ⁽ⁱ⁾ The quantization scale is as follows:

${\mathrm{<span\; class="notranslate">\; Q</span>}}_{\mathrm{<span\; class="notranslate">\; est</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; A</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; TB</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; ACR</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}}$

Wherein, Q _est ⁽ⁱ⁾ is used as the initial value of the average quantization scale. With the initial value Q _est ⁽ⁱ⁾ of the average quantization scale, the video encoder can output better picture quality under the given bit quota TB ⁽ⁱ⁾ . After the target bit quota and the initial value of the quantization level of the current picture are determined, many different types of image encoders can be used, such as Test Model 5 described in the MPEG-2 standard, according to the target bit The amount of yuan is used to complete the compression of the picture.

After the encoding of the current picture is completed, its actual complexity will be known, and the relationship between the total activity of the current picture and the actual complexity, ACR ⁽ⁱ⁾ can be calculated from the following

${\mathrm{<span\; class="notranslate">\; ACR</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; A</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; AQ</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; UB</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}}$

Wherein, ACR ⁽ⁱ⁾ is directly proportional to the total activity A ⁽ⁱ⁾ , and inversely proportional to the actual consumed bit number UB ⁽ⁱ⁾ and the average quantization level AQ ⁽ⁱ⁾ of the current frame. This ACR ⁽ⁱ⁾ can be used to predict the complexity of the next picture of the same type. ACR ⁽ⁱ⁾ can be linearly combined with ACR ^(il) to avoid the influence of those images rich in noise.

The single-pass video encoding of the present invention can be explained by the preferred embodiment in FIG. 3 and the operation flowchart in FIG. 4 . As shown in FIG. 3, the dynamic image video encoder 300 includes a motion estimation unit 310, a complexity estimator 320, a quota allocator 330, a parameter updater 340, a quantization scale estimator 350 and an image encoder 360. The motion estimation unit 310 receives the current frame P ⁽ⁱ⁾ to be encoded in a group of consecutive frames in the video sequence, and is used to calculate the total activity A of the current frame P ⁽ⁱ⁾ when performing block matching for motion estimation ⁽ⁱ⁾ (step S410). The complexity estimator 320 can estimate the complexity C _est ⁽ⁱ⁾ of the current picture based on the total activity A ⁽ⁱ⁾ and the ACR ^(il) of the same type of previously coded pictures in the group of consecutive pictures (step S420 ). The quota allocator 330 updates the instantaneous complexity INST_C of this group of continuous pictures with the estimated complexity C _est ⁽ⁱ⁾ , and allocates the target bit quota TB ⁽ⁱ⁾ to the current picture P ⁽ⁱ⁾ according to C _est ⁽ⁱ⁾ and INST_C (step S430). The average quantization scale initial value Q _est ⁽ⁱ⁾ used to encode the current picture P ⁽ⁱ⁾ is calculated by the quantization scale estimator 350 with the total activity A ⁽ⁱ⁾ , the target bit quota TB ⁽ⁱ⁾ and the activity It is determined based on the ratio ACR ^(il) to the complexity (step S440). In this embodiment, the video encoder 360 encodes the current picture P ⁽ⁱ ) in an adaptive quantization manner based on the initial value Q _est ⁽ⁱ⁾ , and upon completion, reports the actual consumed bit number UB ⁽ⁱ⁾ and the average quantization bit The order AQ ⁽ⁱ⁾ is given to the parameter updater 340 (step S450). Based on the full activity A ⁽ⁱ⁾ , the actual consumed bits UB ⁽ⁱ⁾ and the actual average quantization scale AQ ⁽ⁱ⁾ , the parameter updater 340 calculates the activity pair for the current picture P ⁽ⁱ⁾ Degree ratio ACR ⁽ⁱ⁾ (step S460).

Equivalently, the preferred embodiment in FIG. 3 can be considered to be implemented by hardware and/or software. According to the present invention, the motion estimation unit 310 and other components in FIG. 3 can operate in a pipeline mode. Before the complexity estimator 320 starts to calculate the complexity of the current picture, the motion estimation unit 310 must first complete the current picture to be encoded. The calculation of the full motion of the frame and the estimation of the motion vector, and when the motion estimation unit 310 is preparing for the next frame, the complexity estimator 320 and other components are still busy completing all operations related to the current frame. Next, each component in the preferred embodiment will be described in detail.

One of the main purposes of the motion estimation unit 310 is to decide which prediction mode to use to encode each MB in a picture, and if necessary, to perform forward and backward motion prediction, and also from the block matching operation Information about the amount of screen activity is extracted. Firstly, the intra-activity of each MB is calculated, and the pixel intensities of four 8×8 luminance value blocks in one MB are expressed as Y _m,n ,m=0,...,7,n= 0, ..., 7, and the average value of each 8×8 block is Y, then the amount of activity IntraAct within each MB:

$\mathrm{<span\; class="notranslate">\; IntraAct</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 3</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}$

in

${\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\sqrt{\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 7</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 7</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; Y</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}$

If lower computational complexity is required, it can be replaced by the absolute difference relative to Y:

$\mathrm{<span\; class="notranslate">\; IntraAct</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 3</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; \&Delta;</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}$

in

${\mathrm{<span\; class="notranslate">\; \&Delta;</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 7</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 7</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; Y</span>}}<span\; class="notranslate">\; |</span>$

Since the MB in the I-picture has only one mode: I-mode, IntraAct is the activity amount of each MB of the I-picture.

If the picture is a P- or B-picture, motion estimation is required. The technique most commonly used to find the best movement vector is block matching. For nonintra coding (e.g. P- and B-pictures), selection of forward, backward, bidirectional prediction or no motion compensation. Once the MB mode of each MB in the P- or B-picture is determined, the variance of the 4 8×8 blocks in each motion-compensated difference MB can also be obtained, and the motion-compensated difference MB is to be processed The pixel-by-pixel difference between the MB and estimated MB; the absolute sum of the distortions is usually more computationally efficient and can therefore be used instead of the variance. Add the variance or absolute value of distortion of 4 8×8 blocks to obtain the inter-activity of each MB in the non-intra-coded picture, InterAct, and then calculate the inter-activity of each MB in the non-intra-coded picture The IntraAct and its InterAct are used for comparison to determine whether the InterAct is smaller. If so, use the InterAct as the activity of the MB and encode the MB in inter-mode; otherwise use the IntraAct as the activity of the MB amount, and encode the MB in intra-mode. Finally, for the current I-, P- or B-picture, the total activity A ⁽ⁱ⁾ is obtained by adding the activity of all MBs, and the motion estimation unit 310 sends the full activity A ⁽ⁱ⁾ to the complexity An estimator 320 , a parameter updater 340 and a quantization scale estimator 350 .

Next, the complexity estimator 320 estimates the complexity of a certain type of current picture P ⁽ⁱ⁾ , and introduces weighting coefficients to the estimated complexity C _est ⁽ⁱ⁾ according to the three picture types. Because B-pictures can never be used to predict other pictures, the weighting coefficients of B-pictures can be reduced to allocate less bits to B-pictures and reserve more bits to I- and P-pictures; generally In other words, encoding I-pictures will generate the most bits, so the weighting coefficients of P-pictures are smaller than those of I-pictures. The complexity estimator 320 updates one of the complexity C _I , C _P or C _B according to the type of the current picture, and the complexity C _est ^{(i ) of the current picture P (i} ) is estimated as follows:

if(I-picture)

${\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; est</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; I</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; K</span>}}_{\mathrm{<span\; class="notranslate">\; I</span>}}<span\; class="notranslate">\; \&times;</span>\frac{{\mathrm{<span\; class="notranslate">\; A</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; ACR</span>}}_{\mathrm{<span\; class="notranslate">\; I</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}}$

else if(P-picture)

${\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; est</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; P</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; K</span>}}_{\mathrm{<span\; class="notranslate">\; P</span>}}<span\; class="notranslate">\; \&times;</span>\frac{{\mathrm{<span\; class="notranslate">\; A</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; ACR</span>}}_{\mathrm{<span\; class="notranslate">\; P</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}}$

else if(B-picture)

${\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; est</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; K</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}}<span\; class="notranslate">\; \&times;</span>\frac{{\mathrm{<span\; class="notranslate">\; A</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; ACR</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}}$

Among them, ACR _I ^(il) , ACR _P ^(il) and ACR _B ^(il) are the respective activity-to-complexity ratios of previously coded pictures of I, P and B types in a group of consecutive pictures. These ACR _I ^(il) , ACR _P ^(il) and ACR _B ^(il) are stored in the parameter updater 340, and the complexity estimator 320 reads the corresponding activity versus complexity for the appropriate type of previous frame Ratio. K _I , K _P and K _B are weighting coefficients of I-, P- and B-pictures respectively, and their ranges are generally between 0 and 1.0. As for I-pictures, a preferred embodiment may use K _I =1.0.

When C _est ⁽ⁱ⁾ is received, the quota allocator 330 updates the instantaneous complexity INST_C for the group of continuous pictures and allocates the target bit quota TB ⁽ⁱ⁾ to the current picture P ⁽ⁱ⁾ , the update of the instantaneous complexity INST_C as follows:

INST_C=N _I ×C _I +N _P ×C _P +N _B ×C _B where N _I , N _P and N _B are respectively the numbers of I-, P- and B-pictures in the group of consecutive pictures. Moreover, the target bit quota TB ⁽ⁱ⁾ of the current picture P ⁽ⁱ⁾ is

${\mathrm{<span\; class="notranslate">\; TB</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; est</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}}{\mathrm{<span\; class="notranslate">\; INST</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; C</span>}}<span\; class="notranslate">\; \&times;</span>\frac{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; f</span>}}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; R</span>}$

Among them, n is the number of frames in this group of continuous frames, f is the number of frames per second, that is, frame rate, and R is the expected average bit rate of each group of consecutive frames. However, the actual number of consumed bits will not be exactly equal to the expected bit quota, so a feedback strategy is developed to make the actual consumed bits of the picture close to the target bit quota. In a preferred embodiment, The excess bits encoded so far can be amortized by the current picture being encoded:

TB ⁽ⁱ⁾ = TB ⁽ⁱ⁾ −AR×EB where EB is the amount of overused bits sent by the parameter updater 340 , and AR is the predetermined amortization rate, which generally ranges from 0.05 to 0.2. The quota allocator 330 has to adjust the target bit quota TB ⁽ⁱ⁾ to comply with the VBV specification, so an upper limit (U-bound) and a lower limit (L_bound) of the quota are also set. For constant bit rate (constant bitrate, CBR) operation, the target bit quota allocated to a picture must be such that the VBV buffer will not overflow or be starved, so the target bit quota TB ⁽ⁱ⁾ is limited to Within the lower limit:

if (TB ⁽ⁱ⁾ > U_bound) then TB ⁽ⁱ⁾ = U_bound

if (TB ⁽ⁱ⁾ < L_bound) then TB ⁽ⁱ⁾ = L_bound If it is a variable bit rate (variable bitrate, VBR) operation, it only needs to prevent the lack of VBV, so:

if (TB ⁽ⁱ⁾ > U_bound) then TB ⁽ⁱ⁾ = U_bound Then, the target bit amount TB ⁽ⁱ⁾ will be sent to the parameter updater 340 , the quantization scale estimator 350 and the image encoder 360 .

Once the target bit quota TB ⁽ⁱ⁾ is determined, the quantization scale estimator 350 can estimate the average quantization scale initial value Q _est ⁽ⁱ⁾ required for coding the current picture P ⁽ⁱ⁾ . The initial value Q _est ⁽ⁱ ) used to quantify the current frame P ⁽ⁱ⁾ of a particular type is estimated as follows:

if(I-picture)

${\mathrm{<span\; class="notranslate">\; Q</span>}}_{\mathrm{<span\; class="notranslate">\; est</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; A</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; TB</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; ACR</span>}}_{\mathrm{<span\; class="notranslate">\; I</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}}$

else if(P-picture)

${\mathrm{<span\; class="notranslate">\; Q</span>}}_{\mathrm{<span\; class="notranslate">\; est</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; A</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; TB</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; ACR</span>}}_{\mathrm{<span\; class="notranslate">\; P</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}}$

else if(B-picture)

${\mathrm{<span\; class="notranslate">\; Q</span>}}_{\mathrm{<span\; class="notranslate">\; est</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; A</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; TB</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; ACR</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}}$

Then the average quantization scale initial value Q _est ⁽ⁱ⁾ is sent to the video encoder 360 . Due to the adaptive change of the quantization level of each spatial region in the picture, the video encoder 300 of this embodiment uses a virtual buffer (virtual buffer) to provide a feedback mechanism for bit rate control. Due to the virtual buffer and adaptive Sexual quantification is a known technology for those familiar with this art, so it will not be repeated here. The occupancy of the virtual buffer controls the quantization level, thereby controlling the bit rate. However, the virtual buffer must specify an initial occupancy to operate, and the occupancy of the virtual buffer also reflects the used The quantization level of each spatial region is encoded, so the utilization of the virtual buffer not only controls the initial value of the quantization level, but also controls the visual quality at a given bit rate.

With the help of the initial value Q _est ⁽ⁱ⁾ , the video encoder 360 can make the picture quality more consistent in the video sequence. Via signal line 312 , video encoder 360 receives frame data from motion estimation unit 310 along with motion vectors and MB patterns for each MB. In order to reduce the visible distortion as much as possible under the given bit quota TB ⁽ⁱ⁾ , the image encoder 360 determines the quantization used to encode each MB of the current picture according to the average quantization scale initial value Q _est ⁽ⁱ⁾ Based on these data, the video encoder 360 starts to encode the current picture P ⁽ⁱ⁾ and outputs the encoded data CD. After the current picture P ⁽ⁱ⁾ is encoded, the video encoder 360 averages the quantization level and calculates the actual consumed bit number UB ⁽ⁱ⁾ , and then returns the current picture UB ⁽ⁱ⁾ and the average quantization level AQ ⁽ⁱ⁾ to the parameter updater 340.

The difference between the target bit quota TB ⁽ⁱ⁾ of the picture P ⁽ⁱ⁾ and the actual consumed bit number UB ⁽ⁱ⁾ will be accumulated by the parameter updater 340 so as to obtain the excess bits after encoding the picture P ⁽ⁱ⁾ Quantity EB:

EB=EB×(1-AR)+UB ⁽ⁱ⁾ -TB ⁽ⁱ⁾ Among them, AR is the established amortization rate. Therefore, the relationship between the amount of activity and the actual complexity, ACR ⁽ⁱ⁾ , can be calculated from

${\mathrm{<span\; class="notranslate">\; ACR</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; A</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}}{{\mathrm{<span\; class="notranslate">\; AQ</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; UB</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}}$

This ACR ⁽ⁱ⁾ can be used to estimate the complexity of the next frame of the same type. In order to make ACR ⁽ⁱ⁾ unaffected by noisy pictures, a preferred embodiment utilizes a linear combination of ACR ⁽ⁱ⁾ and ACR ^(il) , for example:

if(I-picture)

${\mathrm{<span\; class="notranslate">\; ACR</span>}}_{\mathrm{<span\; class="notranslate">\; I</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; ACR</span>}}_{\mathrm{<span\; class="notranslate">\; I</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; CW</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; ACR</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; CW</span>}$

else if (P-picture)

${\mathrm{<span\; class="notranslate">\; ACR</span>}}_{\mathrm{<span\; class="notranslate">\; P</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; ACR</span>}}_{\mathrm{<span\; class="notranslate">\; P</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; CW</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; ACR</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; CW</span>}$

else if (B-picture)

${\mathrm{<span\; class="notranslate">\; ACR</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; ACR</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; CW</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; ACR</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; CW</span>}$

Among them, CW is a linear combination of predetermined weighting coefficients. The updated ACR _I ⁽ⁱ⁾ , ACR _P ⁽ⁱ⁾ and ACR _B ⁽ⁱ⁾ will be sent to the complexity estimator 320 and the quantization scale estimator 350, so as to respectively estimate their The initial values of the complexity and the quantization level, in addition, the excess bit amount EB is sent to the quota allocator 330 for bit allocation.

Although the present invention has been disclosed as above with a specific embodiment, it is only for easy description of the technical content of the present invention, and the present invention is not limited to this embodiment in a narrow sense. Anyone skilled in this art will not depart from the spirit of the present invention Within the scope and scope, some changes and modifications can be made, so the protection scope of the present invention should be defined by the scope of the appended patent application as the criterion.

Claims (10)

1, a kind of dynamic image compressed bits rate control method of utilizing the initial quantization bit-level prediction is characterized in that, comprises the following step at least:

Be that a present picture that is about to be encoded calculates a full activity, this present picture is among one group of continuous pictures of a video signal sequence;

Based on this full activity of this present picture and should the group continuous pictures in one of with its activity of previous coding picture of kenel ratio to complexity, estimate one of this present picture complexity;

This complexity of estimating with this present picture is upgraded the complexity in a flash of this group continuous pictures;

According to this estimate complexity and should moment complexity, distribute a target bit amount to give this present picture;

Based on this target bit amount of this full activity of this present picture, this present picture and should be with its activity of previous coding picture of kenel ratio to complexity, decision is used for one of this present picture of coding rank, average quantization position initial value;

According to this rank, average quantization position initial value, this present picture in this video signal sequence is encoded; And

After this present picture of coding,, calculate the ratio of this its activity of present picture to complexity according to this full activity of this present picture, the actual consumption bit number of this present picture and the rank, average quantization position of this present picture;

Wherein, this activity of this present picture is directly proportional with this full activity of this present picture to the ratio of complexity, and is inversely proportional to the actual consumption bit number of this present picture and the rank, average quantization position of this present picture.

2, dynamic image compressed bits rate control method as claimed in claim 1 is characterized in that, comprises the following step more at least:

Carry out above-mentioned its activity of present picture to the ratio of complexity and above-mentioned its activity of previous coding picture with kenel to the ratio linear combination computing between the two of complexity.

3, dynamic image compressed bits rate control method as claimed in claim 1 is characterized in that, the above-mentioned full activity to above-mentioned present picture when above-mentioned present picture moves estimation calculates.

4, dynamic image compressed bits rate control method as claimed in claim 1 is characterized in that, the step of above-mentioned estimation complexity is calculated according to an equation:

$C_{\mathrm{est}}=K\&times;\frac{A}{\mathrm{ACR}}$

Wherein, C _EstThe above-mentioned complexity of estimating for above-mentioned present picture, K is a set weighted value, its scope is between 0 to 1, and A is the above-mentioned full activity of the above-mentioned present picture in above-mentioned one group of continuous pictures, and ACR is the ratio of above-mentioned its activity of previous coding picture with kenel to complexity.

5, dynamic image compressed bits rate control method as claimed in claim 1 is characterized in that, the target bit amount of above-mentioned distribution is directly proportional with the above-mentioned complexity of above-mentioned present picture, and is inversely proportional to above-mentioned moment complexity of above-mentioned one group of continuous pictures.

6, dynamic image compressed bits rate control method as claimed in claim 1 is characterized in that, rank, above-mentioned average quantization position initial value is determined by following array function:

${\mathrm{MQ}}_{\mathrm{est}}=\frac{A}{\mathrm{TB}\&times;\mathrm{ACR}}$

Wherein, MQ _EstRepresent rank, above-mentioned average quantization position initial value, A is the above-mentioned full activity of the above-mentioned present picture in above-mentioned one group of continuous pictures, TB is the above-mentioned target bit amount of the above-mentioned present picture of above-mentioned present picture, and ACR is the ratio of above-mentioned its activity of previous coding picture with kenel to complexity.

7, dynamic image compressed bits rate control method as claimed in claim 1 is characterized in that, above-mentioned one group of continuous pictures comprises a picture group at least, and this picture group meets MPEG video signal standard.

8, a kind of dynamic image compressed bits rate control device that utilizes the initial quantization bit-level prediction is characterized in that, comprises at least:

One moves evaluation unit, receives one of to be about to be encoded picture at present in one group of continuous pictures in the video signal sequence, is used for calculating one of this present picture activity entirely when carrying out the block coupling that moves estimation;

One complexity estimator, based on this full activity of this present picture and should the group continuous pictures in together its activity of previous coding picture of kenel the ratio of complexity is used for estimating one of this present picture complexity;

One quota allocator, this complexity of estimating with this present picture is upgraded the complexity in a flash of this group continuous pictures, be used for according to this estimation complexity and should moment complexity, distribute a target bit amount to give this present picture;

One quantization scale estimator, based on this target bit amount of this full activity of this present picture, this present picture and should be with its activity of previous coding picture of kenel ratio to complexity, decision is used for one of this present picture rank, the average quantization position initial value of encoding; And

One parameter update device based on this full activity of this present picture, the actual consumption bit number of this present picture and the rank, average quantization position of this present picture, calculates the ratio of this its activity of present picture to complexity;

Wherein, this activity of this present picture is directly proportional with this full activity of this present picture to the ratio of complexity, and is inversely proportional to the bit number of this present picture consumption and the rank, average quantization position of this present picture.

9, dynamic image compressed bits rate control device as claimed in claim 8, it is characterized in that, the above-mentioned parameter renovator also carry out above-mentioned its activity of present picture to the ratio of complexity and above-mentioned its activity of previous coding picture with kenel to the ratio linear combination computing between the two of complexity.

10, dynamic image compressed bits rate control device as claimed in claim 8, it is characterized in that, more at least comprise an image coder, come above-mentioned present picture is encoded according to rank, above-mentioned average quantization position initial value, and the actual consumption bit number and the used rank, average quantization position of above-mentioned present picture coding of above-mentioned present picture are repaid to the above-mentioned parameter renovator.

Images