RU2340115C1 - Method of coding video signals, supporting fast algorithm of precise scalability on quality - Google Patents

Abstract

FIELD: information technology.

SUBSTANCE: video decoding method supports an algorithm of precise scalability on quality (FGS). The method involves: obtaining a predicted image for the current frame, using a motion vector, evaluated with a predetermined accuracy, quantisation of the difference between the current frame and the predicted image, inverse quantisation of the difference and formation of a reconstructed image of the current frame, compensation of motion on the reference frame of the FGS level and on the reference frame of the main level, using an estimated motion vector, calculation of the difference between the FGS level reference frame with compensated motion and the reference frame of the main level with compensated motion, subtraction of the reconstructed image from the current frame and the calculated remainder from the current frame and coding of the subtraction result.

EFFECT: reduced volume of calculations, required for a multi-level algorithm of progressive precise scalability on quality (PFGS).

49 cl, 12 dwg

Description

FIELD OF THE INVENTION

The methods and devices of the present invention relate to encoding video signals and, more specifically, to encoding video information, which reduces the amount of computation required for the Progressive Fine Granular Scalability (PFGS) algorithm based on the multilevel algorithm.

State of the art

With the development of information transmission technology, including the Internet, the volume of multimedia services containing various types of information, such as text, video, audio information and so on, has increased. Multimedia data requires a large capacity of storage media and a wide bandwidth for transmission, since the volume of multimedia data is usually large. Accordingly, for transmitting multimedia data, including text, video and audio information, a compression encoding method is a necessity.

The basic principle of data compression is to eliminate data redundancy. Data can be compressed by removing spatial redundancy, in which the same color or object is repeated in the image, temporary redundancy, in which there are only small changes between adjacent frames of the moving image, or the same sound is repeated in the audio information, or mental visual redundancy taking into account the peculiarities of human vision and the limited perception of high frequency. In conventional video coding, temporal redundancy is removed by temporal filtering based on motion compensation, and spatial redundancy is removed by spatial transformation.

Communication media is required to transmit multimedia data created after data redundancy removal. Different types of media for multimedia have different characteristics. Currently used communication tools have different transmission speeds. For example, an ultra-fast communication network can transmit data at a speed of several tens of megabits per second, while a mobile communication network has a transmission speed of 384 kilobits per second. Scalable data encoding methods may be suitable for supporting communication media having different transmission rates or for transmitting multimedia data in a multimedia environment.

Scalability indicates the ability to partially decode a single compressed bit stream. Scalability includes spatial scalability indicating video resolution, signal-to-noise scalability (SNR), indicating the level of video quality, and temporal scalability, indicating the frame rate.

Standardization work to implement multi-level scalability based on Scalable Extension H.264 technology (hereinafter referred to as “H.264 SE”) is currently being carried out by the joint video group (JVT) MPEG (cinematography expert group) and ITU ( International Telecommunication Society). To support signal-to-noise ratio (SNR) scalability, JVT implements existing Fine Granular Scalability (FGS) technologies.

1 is a diagram for explaining a conventional FGS method. The FGS-based codec performs encoding by dividing the video bitstream into the main layer and the FGS layer. In the present description, a dash ()) is used to indicate a reconstructed image obtained after quantization / inverse quantization. More specifically, a block PB predicted in advance from block MB ′ in the reconstructed left frame of the main layer 11 and block NB ′ in the reconstructed right frame 12 of the main level using the motion vector is subtracted from block O in the original current frame 12 to obtain a difference block RB. Thus, the difference block RB can be defined by equation (1):

R _B = O - P _B = O - (M _B '+ N _B ') / 2 (one)

The difference block R _B is quantized using the quantization step QP _B (RB ^Q ) of the main layer and then is inverted quantized to obtain the reconstructed difference block R _B ′. The difference between the non-quantized difference block R _B and the restored difference block R _B ', the block Δ corresponding to the difference is quantized with a quantization step size QP _F smaller than the quantization step of the main level QP _B (the compression ratio decreases as the size of the quantization step decreases ) The quantized block is denoted as Δ Δ ^Q. The quantized difference block R _B ^{Q of the} difference at the fundamental level and the quantized block Δ ^Q at the FGS level are ultimately transmitted to the decoder.

Figure 2 shows a diagram for explaining the operation of the traditional method of progressive precision scalability in quality (PFGS). The conventional FGS method uses the reconstructed difference R _B ′ of the quantized base layer to reduce the amount of data at the FGS level. With reference to FIG. 2, the PFGS method exploits the fact that the quality of the left and right reference frames at the FGS level is also improved by the FGS method. That is, the PFGS method comprises computing a new difference R _F using the newly updated left reference frame 21 and the right reference frame 23 and quantizing the difference between the new difference block R _F and the quantized base layer block R _F ', thereby improving coding characteristics. The new difference block RF is defined by equation (2):

RF = O - P _F = O - (M _F '+ N _F ') / 2 (2)

where M _F 'and N _F ' respectively denote the areas in the restored left reference frame 21 and the right reference frame 23 at the FGS level according to the corresponding motion vectors.

The PFGS method has the advantage over the FGS method in that the data volume at the FGS level can be reduced due to the high quality of the left and right reference frames. Since the FGS level also requires separate motion compensation, the amount of computation increases. That is, although the PFGS method has improved performance compared to the conventional FGS method, it requires a lot of computation because motion compensation is performed for each FGS level to create a predicted signal and a difference signal between the predicted signal and the original signal. Recently developed video codecs interpolate an image signal to compensate for motion with an accuracy of 1/2 or 1/4 pixel. When motion compensation is performed with an accuracy of 1/4 pixel, an image should be created with a size corresponding to four times the resolution of the original image.

SUMMARY OF THE INVENTION

Technical problem

The H.264 SE method uses a six-pole filter as a 1/2 pixel interpolation filter, which has significant computational complexity, requiring a large amount of computation to compensate for motion. This complicates the encoding and decoding processes, thus requiring increased system resources. In particular, this drawback can be most problematic in the field, requiring real-time encoding and decoding, such as live broadcasting or video conferencing.

Technical solution

The present invention provides a method and apparatus for reducing the amount of computation required to compensate for motion while maintaining the characteristics of the Progressive Fine Quality Scalability Algorithm (PFGS).

According to an aspect of the present invention, there is provided a video encoding method supporting FGS, a video encoding method comprising the steps of obtaining a predicted image for the current frame using a motion vector estimated with predetermined accuracy, quantizing the difference between the current frame and the predicted image, inverting quantization quantized difference and create a reconstructed image for the current frame, performing motion compensation on the reference frame level The FGS and the reference frame of the main level, using the estimated motion vector, calculate the difference between the reference frame of the FGS level with compensated movement and the reference frame of the main level with compensated movement, subtract the reconstructed image for the current frame and the calculated differences from the current frame and encode the result of subtraction.

According to another aspect of the present invention, there is provided a video coding method supporting FGS, a video coding method comprising the steps of obtaining a predicted image of a current frame using a motion vector estimated with predetermined accuracy, quantizing the difference between the current frame and the predicted image, performing inverse quantization quantized difference and create a reconstructed image for the current frame, perform motion compensation on the reference frame the FGS and the reference frame of the main level, using the estimated motion vector, and create a predicted frame for the FGS level and the predicted frame for the main level, respectively, calculate the difference between the predicted frame for the FGS level and the predicted frame for the main level, subtract the reconstructed image and the difference from the current frame and encode the result of subtraction.

According to another aspect of the present invention, there is provided a video coding method supporting FGS, a video coding method comprising obtaining a predicted image for a current frame using a motion vector estimated with predetermined accuracy, quantizing the difference between the current frame and the predicted image, performing inverse quantization of the quantized difference and create a reconstructed image for the current frame, calculate the difference between the reference frame for the FGS level and nym frame for the base layer, performing motion compensation on the difference from the estimated motion vector, subtracting the reconstructed image and the motion-compensated result from the current frame, and encoding the result of subtraction.

According to yet another aspect of the present invention, there is provided a video coding method supporting an accurate quality scalability algorithm (FGS), a video coding method comprising the steps of obtaining a predicted image for a current frame using a motion vector estimated with predetermined accuracy, performing motion compensation on a reference frame the FGS level and on the reference frame of the main level, using the motion vector with lower accuracy than the accuracy of the estimated motion vector, calculate the value between the reference frame of the FGS level with compensated movement and the reference frame of the main level, subtract the predicted image from the current frame and encode the result of the subtraction.

According to yet another aspect of the present invention, there is provided a video encoding method supporting FGS, a video encoding method comprising the steps of obtaining a predicted image for the current frame using a motion vector estimated with predetermined accuracy, performing motion compensation on the reference frame of the FGS level and on the reference frame of the main level using a motion vector with lower accuracy than the accuracy of the estimated motion vector, and create a predicted frame for the FGS level and the predicted frame for the main level, respectively, the difference between the predicted frame for the FGS level and the predicted frame for the main level is calculated, the predicted image and the calculated difference are subtracted from the current frame, and the subtraction result is encoded.

According to yet another aspect of the present invention, there is provided a video encoding method supporting FGS, a video encoding method comprising the steps of obtaining a predicted image for the current frame using a motion vector estimated with predetermined accuracy, calculating the difference between the reference frame of the FGS level and the reference frame ground level, perform motion compensation on the difference using a motion vector with lower accuracy than the accuracy of the estimated motion vector, subtract The tanned image and the result of the compensated movement of the current frame also encode the result of subtraction.

According to another aspect of the present invention, there is provided a video decoding method supporting FGS, a video decoding method comprising the steps of extracting texture layer data of the main layer and texture data of the FGS layer and motion vectors from the input bit stream, restoring the frame of the basic layer from the texture layer data of the main layer, motion compensation on the reference frame of the FGS level and on the reference frame of the main level, using the motion vectors, calculate the difference between the reference frame of the FGS level with compensation motion and the reference frame of the main level with compensated movement and add together the frame of the main level, texture data of the FGS level and the difference.

According to another aspect of the present invention, there is provided an FGS-based video encoder comprising an element obtaining a predicted image for the current frame using a motion vector estimated with predetermined accuracy, an element that quantizes the difference between the current frame and the predicted image, inverts quantization of the quantized difference and creates of the reconstructed image for the current frame, an element that performs motion compensation on the reference frame of the FGS level and the reference frame of the main level using the estimated motion vector, an element calculating the difference between the reference frame of the FGS level with compensated movement and the reference frame of the main level with compensated movement, an element subtracting the reconstructed image and the difference from the current frame, and an element encoding the result of subtraction.

According to another further aspect of the present invention, there is provided an FGS-based video decoder, a video decoder comprising an element extracting ground layer texture data, FGS layer texture data and motion vectors from an input bit stream, an element restoring a main layer frame from the main layer texture data, element performing motion compensation on the reference frame of the FGS level and on the reference frame of the main level using the motion vector, and creating a predicted frame of the FGS level and predicted th frame of the core layer, element, calculating the difference between the predicted FGS layer frame and the predicted frame of the core layer, and an element adding together the texture data, the reconstructed base layer frame and the difference.

Description of drawings

The above and other aspects of the present invention will become clearer with a detailed description of examples of variants of its implementation with reference to the accompanying drawings, in which:

1 is a diagram for explaining a conventional FGS method.

2 is a diagram for explaining a conventional progressive PFGS method.

FIG. 3 is a diagram for explaining fast progressive accurate quality scalability (PFGS) in accordance with an example embodiment of the present invention.

4 is a block diagram of a video encoder in accordance with an example embodiment of the present invention.

5 is a block diagram of a video encoder according to another example of an embodiment of the present invention.

6 and 7 are block diagrams of video encoders in accordance with a further example embodiment of the present invention.

8 is a block diagram of a video encoder in accordance with an example embodiment of the present invention.

9 is a block diagram of a video decoder according to another example of an embodiment of the present invention.

10 and 11 are block diagrams of video decoders in accordance with a further example embodiment of the present invention.

12 is a block diagram of a system for performing an encoding or decoding process in accordance with an example embodiment of the present invention.

Description of preferred embodiments of the invention

The present invention will now be described in more detail with reference to the accompanying drawings, in which examples of embodiments of the invention are shown.

The advantages and features of the present invention and methods for its implementation may become more apparent when considering the following detailed description of examples of embodiments and accompanying drawings. The present invention may, however, be embodied in many different forms and should not be construed as being limited to the exemplary embodiments set forth herein. Rather, these examples of embodiments are presented in this description so that this description is comprehensive and complete and fully discloses the concept of the invention to those who are specialists in this field of technology, and the present invention will be defined only by the attached claims. The same numbers refer to the same elements throughout the description.

FIG. 3 shows a PFGS diagram corresponding to a first example of an embodiment of the present invention.

With reference to figure 3, as in figure 2, Δ at the FGS level will be quantized in accordance with the PFGS algorithm and is determined by simple equation (3):

Δ = R _F - R _B ' (3)

R _{F is} determined by the above equation (2), and R _B 'is determined by the equation (4):

R _B '= O' - P _B = O '- (M _B ' + N _B ') / 2 (four)

where O 'is an image reconstructed by quantizing an original image O with the core layer quantization step size QP _B and subsequent inverse quantization of the quantized image.

Substitution of equations (2) and (4) into equation (3) gives equation (5):

Δ = O - (M _F '+ N _F ') / 2 - [O '- (M _B ' + N _B ') / 2] (5)

With reference to FIG. 3, Δ _M and Δ _N denote the difference between the left reference frames M _F 'and M _B ' at the main level and at the FGS level and the difference between the right reference frames N _F 'and N _B ' at the main level and at level FGS, respectively, and are determined by equations (6):

Δ _M = M _F '- M _B ' Δ _N = N _F '- N _B ' (6)

By substituting equation (6) into equation (5), Δ can be determined by equation (7):

Δ = O - O '- (Δ _M + Δ _N ) / 2 (7)

As can be seen from equation (7), the encoder can obtain Δ by subtracting the reconstructed image of the main level O 'obtained by quantizing the original image O with a quantization step of the main level of size QP _B and then inverting the quantized image and the average value of the differences between each of the reference frames of the main level and reference frames of the FGS level and the original image O, that is (Δ _M + Δ _N ) / 2. The decoder restores the original image O by adding the image O 'of the main level, Δ and the average value of the differences between the reference frame of the main level and the reference frame of the FGS level.

In the conventional PFGS algorithm, motion compensation is performed using a motion vector with an accuracy of one pixel or part of a pixel (1/2 pixel or 1/4 pixel) obtained from motion estimation. Recently, in order to increase compression efficiency, motion estimation and compensation are usually performed with various precision expressed in pixels, such as half pixel accuracy or quarter pixel accuracy. In the conventional PFGS algorithm, a predicted image created by motion compensation with an accuracy of, for example, 1/4 pixel is packed into integer pixels. Then, quantization is performed on the difference between the original image and the predicted image. In this case, packaging is the process of reconstructing a 4-x interpolated reference image into an image of its original size, performing motion estimation with an accuracy of 1/4 pixel. For example, during the packaging process, one out of every four pixels can be selected.

However, the Δ data at the FGS level, which must be quantized for fast PFGS according to the present invention, as defined by equation (7), which lead to low compression efficiency, should not be subject to motion estimation with high pixel accuracy. Motion estimation and compensation apply only to the third term (Δ _M + Δ _N ) / 2 on the right side of equation (7). However, since the third term is represented as the difference between the reference frames at the intermediate level, performing motion estimation and compensation with high pixel accuracy does not give high efficiency. That is, since the resulting difference image between the image at the basic level with compensated motion with a predetermined pixel accuracy and the image at the expansion level with compensated motion with pixel accuracy is insensitive to pixel accuracy, the fast PFGS algorithm allows motion estimation and compensation with lower pixel accuracy than regular PFGS.

According to the second example of the embodiment, Δ in equation (5) in the first example of the embodiment can also be represented as the difference between the predicted signals P _F and P _B , as shown in equation (8). P _F and P _B are equal (M _F '+ N _F ') / 2 and (M _B '+ N

Δ = O - O '- (P _F - P _B ) (8)

The first and second examples of the embodiment differ from each other as follows. In the first example of the embodiment, the differences Δ _M and Δ _N between the reference images of the FGS level and the reference images of the main level are first calculated and then divided by 2. In the second example of the embodiment, the difference P _F -P _B between the predicted image P _{F of} the FGS level and the predicted image P _{In the} main level is calculated after calculating the predicted images P _F and P _In these two levels. That is, although fast PFGS algorithms corresponding to the first and second examples of embodiments are implemented in different ways, the same calculation result (Δ) can be obtained.

In both the first and second examples of embodiments, motion compensation is first performed, and then the difference between the images is calculated. In a third example of an embodiment of the present invention, the difference between the reference images at various levels can be calculated first, and then motion compensation can be performed. Thus, in accordance with a third example of an embodiment of the present invention, since motion compensation is performed for the difference, filling at the borders has little effect on the resulting image. Thus, the filling process at the borders can be skipped. Filling at borders is a process of duplicating pixels located in close proximity to borders, given that during motion estimation, coincidence of blocks at the frame boundary is limited.

In a third example of an embodiment of the present invention, the difference Δ can be determined by equation (9):

Δ = O - O '- [(mc (M _F ' - M _B ') + mc (N _F ' - N _B ')) / 2 (9)

where mc (.) means a function to perform motion compensation.

While conventional PFGS is used to perform direct prediction (motion estimation and compensation) to calculate R _F or R _B defined by equation (3), fast PFGS algorithms corresponding to Examples 1 through 3 of the present invention are used to calculate the difference between the predicted images or predicting the difference between the reference images. Thus, the fast PFGS characteristics of the present invention are only slightly affected or insensitive to the interpolation used to improve the pixel accuracy of the motion vector.

Thus, a quarter or half pixel interpolation can be skipped. In addition, instead of the half-pixel interpolation filter used in the H.264 standard and requiring a large amount of computation, a bilinear filter can be used that requires less computation. For example, a bilinear filter can be applied to the third terms in the right-hand sides of equations (7) - (9). This can reduce performance degradation compared to when a bilinear filter is directly applied to the predicted signal to obtain R _F and R _B as in the normal PFGS algorithm.

The principle of operation from first to third examples of embodiments of the present invention is based on equation (3). In other words, the implementation of these examples of embodiments begins with the assumption that the difference between the difference R _{F of} the FGS level and the difference R _{B of the} main level must be encoded. However, when the difference obtained from the FGS level is very small, that is, when the time correlation is very close, the aforementioned fast PFGS algorithms corresponding to the first three examples of embodiments can significantly degrade the encoding performance. In this case, encoding only the difference obtained from the FGS level, that is, R _F in equation (3), can offer better encoding characteristics. That is, according to a fourth example of an embodiment of the present invention, equations (7) to (9) can be converted to equations (10) to (12), respectively:

Δ = O - P _B - (Δ _M + Δ _N ) / 2 (10) Δ = O - P _B - (P _F -P _B ) (eleven) Δ = O - P _B - [(mc (M _F '- M _B ') + mc (N _F '- N _B ')) / 2 (12)

In equations (10) - (12), the reconstructed main-level image O 'is replaced by the predicted image P _B for the main-level image. Of course, interpolation may not apply to the third terms on the right-hand side of equations (10) - (12), or a bilinear filter that requires less computation may be used for interpolation.

The predicted image P _B occurring twice in equation (11) is not necessarily the same. An estimated motion vector can be used during motion compensation to create a predicted image P _B in the second term. On the other hand, to create P _B and P _F in the third term during motion compensation, a motion vector can be used with an accuracy lower than that of the estimated motion vector, or a filter that requires less computation (for example, a bilinear filter).

The PFGS algorithm, in which the current frame is restored using both the restored left and right reference frames, suffers from an error caused by drift when image quality deterioration in both the left and right reference frames is cumulatively reflected in the current frame. The drift error can be reduced using transmission prediction technology using a predicted image created by a weighted sum of the predicted image obtained from both reference frames and the predicted image obtained from the main level.

According to the transmission prediction technology used in conventional PFGS, encoded at the FGS level, is expressed by equation (13):

Δ = O - [αP _F + (l-α) P _B ] (13)

Equation (13) can be converted to equation (14) in accordance with a fifth example of an embodiment of the present invention:

Δ = O - P _B - α (P _F - P _B ) (fourteen)

To obtain equation (14), the weight coefficient α can only be applied to the difference (P _F - P _B ) between the predicted images in equation (11). Thus, the present invention can also be applied to transmission prediction technology. That is, interpolation can be skipped or interpolation can be applied to the difference (P _F - P _B ) using a bilinear filter that requires less computation. In the latter case, the result of the interpolation is multiplied by the weight coefficient α.

4 is a block diagram of a video encoder 100 corresponding to a first example of an embodiment of the present invention.

Although the invention is described with respect to each block as a basic motion estimation module with reference to FIGS. 1-3, a quick PFGS algorithm will now be described with respect to each frame containing the block. For the sequence of presentation, the block identifier is indicated for “F” by a subscript indicating the frame. For example, a frame containing a block labeled R _B is denoted as F _RB . And of course, the top stroke (') is used to indicate the recovered data obtained after quantization / inverse quantization.

The current frame F _o is supplied to the motion estimation device 105, the subtracting device 115 and the difference calculating device 170.

The motion estimation device 105 performs a motion estimation for the current frame F _o using neighboring frames to obtain motion vectors MV. The adjacent frames that are mentioned during motion estimation are hereinafter referred to as “reference frames”. A block matching algorithm (BMA) is typically used to evaluate the motion of a given block. In BMA, a given block moves within the search zone in a reference frame accurate to a pixel or a fraction of a pixel, and the offset with a minimum error is defined as a motion vector. Although a motion block of a fixed size is used to estimate the motion, motion estimation can be done using the hierarchical matching of variable size blocks (HVSBM).

When motion estimation is performed to within a fraction of a pixel, reference frames should be enlarged or interpolated to a predetermined resolution. For example, when motion estimation is performed with 1/2 and 1/4 pixel precision, the reference frames should be modified or interpolated with a factor of two and four, respectively.

When the encoder 100 has an open-loop codec structure, frames F _M and F _N adjacent to the original are used as reference frames. When the encoder 100 has a closed-loop codec structure, the reconstructed adjacent frames F _MB 'and F _NB ' at the main level are used as reference frames. Although it is assumed here that the encoder 100 has a closed-loop codec structure, the encoder 100 may have an open-loop codec structure.

The motion vectors MV calculated by the motion estimation device 105 are transmitted to the motion compensator 110. The motion compensator 110 performs motion compensation on the reference frames F _MB 'and F _NB ' using the motion vectors MV, and creates a predicted frame F _PB for the current frame. When bidirectional prediction is used, the predicted image can be calculated as the average value of the reference frames with compensated movement. When using unidirectional prediction, the predicted image may be the same as the reference frame with compensated movement. Although it is further assumed here that motion estimation and compensation use bidirectional reference frames, it should be apparent to those skilled in the art that the present invention can use a unidirectional reference frame.

The subtractor 115 calculates the difference F _RB between the predicted image and the current image for transmission to the converter 120.

Transducer 120 performs a spatial transform of the difference F _RB to create a transform coefficient F _RB ^T. The spatial transform method may comprise a discrete cosine transform (DCT), or a pulse transform. Specifically, DCT coefficients can be created when DCT is used, and pulse coefficients can be created when pulsed conversion is enabled.

The quantizer 125 applies quantization to the transform coefficient F _RB ^T. Quantization means the process of expressing transform coefficients generated among arbitrary real values using discrete values, and matching discrete values with indices corresponding to a predetermined quantization table. For example, quantization device 125 may divide the actually estimated transform coefficient into a quantization step with a predetermined size and round the resulting value to the nearest integer. In general, the quantization step size of the base layer is larger than the quantization step size of the FGS level.

The quantization result, that is, the quantization coefficient F _RB ^Q obtained by the quantization device 125, is supplied to the statistical encoding unit 150 and the inverse quantization device 130.

The inverse quantization device 130 quantizes the quantization coefficient F _RB ^Q. Inverse quantization means the inverse quantization process to restore values that match the indices created during quantization using the same quantization step that was used in the quantization.

The inverse converter 135 receives the inverse quantization result and performs an inverse transformation of the received result. The inverse spatial transform may be, for example, an inverse DCT or an inverse pulse transform performed in the reverse order to that in which the transform was performed by the transducer 120. The adder 140 adds the inverted transformed result to the predicted image F _PB obtained from the motion compensator 110 to create restored image.

Buffer 145 stores the addition result received from adder 140. Buffer 145 stores the reconstructed image F _o ′ for the current frame, as well as the previously reconstructed reference frames F _MB ′ and F _NB ′ of the main layer.

The motion vector transducer 155 changes the accuracy of the received motion vector MV. For example, the motion vector MV with an accuracy of 1/4 pixel may have a value of 0, 0.25, 0.5 or 0.75. As described above, in accordance with examples of embodiments of the present invention, there is a slight difference in coding characteristics when motion compensation at the FGS level is performed for the motion vector MV with lower pixel accuracy than at the main level. Thus, the motion vector converter 155 converts the motion vector MV with an accuracy of 1/4 pixel to the motion vector MV ₁ with a pixel accuracy lower than the accuracy of 1/4 pixel, such as 1/2 pixel or 1 pixel. Such a conversion procedure can be performed by simply dropping or rounding the decimal part of the pixel accuracy for the initial motion vector.

Buffer 165 temporarily stores reference frames F _MF ' _, F _NF ' of the FGS level. Although not shown in the drawing, the reconstructed FGS frames F _MF 'and F _NF ' or the original frame adjacent to the current frame can be used as reference frames of the FGS level.

Motion compensator 160 uses the converted motion vector MV ₁ to perform motion compensation on the reconstructed reference layer frames F _MB ′ and F _NB ′ received from the buffer 145 and the restored reference frames F _MF ′ and F _NF ′ of the FGS level received from the buffer 165 , and provides the motion-compensated frame transmission mc (F _MB '), mc (F _NB '), mc (F _MF ') and mc (F _NF ') to the difference calculator 170. F _MF 'and F _NF ' mean reference previous and subsequent frames at the FGS level, respectively. F _MB 'and F _NB ' mean reference previous and subsequent frames at the main level, respectively.

When interpolation is required for motion compensation, motion compensator 160 may use a filter with a different type of interpolation than the filter used for motion estimation device 105 or motion compensator 110. When, for example, a motion vector MV ₁ with an accuracy of 1/2 pixel is used, a bilinear filter that requires little computation can be used for interpolation instead of the six-pole filter used in the H.264 standard. Since the difference between the frame of the main level with compensated motion and the frame of the FGS level with compensated motion is calculated after interpolation, the interpolation process has little effect on the compression efficiency.

The difference calculator 170 calculates the difference between the reference frame mc (F _MF '), mc (F _NF ') of the compensated motion level FGS and the reference frame mc (F _MB '), mc (F _NB ') of the main level with compensated motion. That is, the difference calculator 170 calculates Δ _M = mc (F _MF ') - mc (F _MB ') and Δ _N = mc (F _NF ') - mc (F _NB '). Of course, when a unidirectional reference frame is used, only one difference can be calculated.

Then, the difference calculator 170 calculates the average of the differences Δ _M and Δ _N and subtracts the reconstructed image F _o 'and the average of the differences Δ _M and Δ for the current frame F _about . When a unidirectional reference frame is used, an average value calculation process is not required.

The subtraction result F _Δ obtained by the difference calculating device 170 is subjected to spatial transformation by the transducer 175 and then quantized by the quantization device 180. The quantized result F _Δ ^Q is transmitted to the statistical coding unit 150. The quantization step size used by the quantizer 180 is typically smaller than that used by the quantizer 125.

The lossless statistical encoding unit 150 encodes the motion vector MV estimated by the motion estimation device 105, the quantization coefficient F _RB ^Q received from the quantizer 125, and the quantized result F _Δ ^Q received from the quantizer 180 into a bit stream. There are many lossless coding methods, including arithmetic coding, variable length coding, etc.

Alternatively, although not shown, the video encoder corresponding to the second example of the embodiment of the present invention may have the same configuration and operate as the video encoder 100 shown in FIG. 4, with the exception of the difference calculator.

That is, a difference calculating device according to a second example of an embodiment of the present invention creates a predicted frame for each level before calculating the difference between frames at different levels. In other words, the difference calculating device creates a predicted frame of the FGS layer and a predicted frame of the main level using the reference frame of the FGS level with compensated motion and the reference frame of the main level with compensated motion. The predicted frame can be calculated by simply averaging two reference frames with compensated movement. Of course, when unidirectional prediction is used, the frame with compensated motion may be the predicted frame itself.

The difference calculator then calculates the difference between the predicted frames and subtracts the reconstructed image and the calculated difference from the current frame.

5 is a block diagram of a video encoder 300 according to a third example of an embodiment of the present invention. Referring to FIG. 5, while in the first and second examples of embodiments, the difference between the reference image of the main layer and the reference image of the FGS level is calculated after performing motion compensation, the video encoder 300 shown performs motion compensation after calculating the difference between the reference frames on these two levels. In order to avoid re-explanation, the following description will focus on the distinguishing features between the first and second examples of embodiments.

Subtractor 390 subtracts the reconstructed basic layer reference frames F _MB 'and F _NB ', which are received from buffer 345, from reference frames F _MF 'and F _NF ' of FGS level, which are received from buffer 365, and provides transmission of subtraction results F _MF ' - F _MB 'and F _NF ' - F _NB 'to the motion compensator 360. When a unidirectional reference frame is used, there is only one difference.

The motion compensator 360 uses the modified motion vector MV ₁ received from the motion vector converter 355 to perform motion compensation on the differences F _MF '- F _MB ' and F _NF '- F _NB ' between the reference frame of the FGS level and the reference frame of the main level received from subtractor 390. When a motion vector MV ₁ with 1/2 pixel accuracy is used during motion compensation, a bilinear filter that requires a small amount of computation can be used for interpolation instead of the six-pole filter used in the H.264 standard. As described above, interpolation has little effect on compression efficiency.

The difference calculator 370 calculates an average value between the differences (F _MF '-F _MB ') and (F _NF '-F _NB ') of the compensated movement and subtracts the reconstructed image F _o 'and the average value from the current frame F _o . When a unidirectional reference frame is used, an averaging process is not required.

6 and 7 are block diagrams of examples of video encoders 400 and 600 corresponding to a fourth example of an embodiment of the present invention. Referring first to FIG. 6, in contrast to the first example of the embodiment shown in FIG. 4, the difference calculator 470 in the video encoder 400 of the example embodiment subtracts the predicted main layer frame F _PB instead of the reconstructed main level frame F _o ′ from the current frame F _about .

The video encoders 400 and 600 of the fourth example embodiment shown in FIGS. 6 and 7 correspond to FIGS. 4 and 5 showing the video encoders 100 and 300 of the first and third example embodiments. Turning first to FIG. 6, the difference calculator 470 subtracts the prediction image F _{PB of the} main layer received from the motion compensator 410 instead of the reconstructed image F _o ′ of the main level from the current frame. Thus, the difference calculator 470 subtracts the predicted image F _PB and the average value of the differences Δ _M and Δ _N from the current frame F _o to obtain a subtraction result F _Δ .

Similarly, with reference to FIG. 7, the difference calculator 670 subtracts the predicted image F _PB and the average value of the differences of the compensated movement mc (F _MF '- F _MB ') and mc (F _NF '- F _NB ') from the current frame F _about to get the result of subtraction F _Δ .

An example of a video encoder corresponding to a fourth example of an embodiment matching the second example of an embodiment (not shown) may have the same configuration and perform the same operation as shown in FIG. 6, except for the operation performed by the difference calculator 470. In the video encoder of the fourth exemplary embodiment corresponding to the second exemplary embodiment, the difference calculator 470 creates the predicted FGS level frame F _PB and the main level predicted frame F _BF using the reference frame mc (F _MF ′), mc (F _NF ′) with compensated the movement of the FGS level and the reference frame mc (F _MB '), mc (F _NB ') with compensated movement of the main level, respectively. The difference calculator 470 also calculates the difference F _PF - F _PB between the predicted frames F _PF and F _PB and subtracts the reconstructed image F _o 'and the difference F _PF - F _PB from the current frame F to obtain a subtraction result F _Δ .

If transmission prediction is used, the difference calculator 470 multiplies the weight coefficient α by the difference F _PF - F _PB and subtracts the reconstructed image F _о 'and the product (α × (F _PF - F _PB ) from the current frame F _o to obtain the subtraction result F _Δ .

FIG. 8 shows a block diagram of a video decoder 700 according to a first example of an embodiment of the present invention. With reference to FIG. 8, a lossless statistical decoding unit 701 decodes an input bitstream to extract main layer texture data F _PB ^Q , FGS layer texture data F _Δ ^Q , and motion vectors MV. Lossless decoding is an inverse lossless coding process.

The ground level texture data F _PB ^Q and the FGS level texture data F _Δ ^Q are supplied to inverse quantization devices 705 and 745, respectively, and the motion vectors MV are supplied to the motion compensator 720 and the motion vector converter 730.

The inverse quantization device 705 applies inverse quantization to the main layer texture data F _PB ^Q received from the statistical decoding unit 701. Inverse quantization is performed in the reverse order of quantization performed by the transducer to restore values that match the indices created during quantization in accordance with the predetermined quantization step used in the quantization.

Inverse transformer 710 performs an inverse transform of the inverse quantized result. The inverse transform is performed in the reverse order of the transform performed by the converter. Specifically, an inverse DCT transform or an inverse pulse transform can be used.

The reconstructed difference F _RB 'is supplied to the adder 715.

The motion compensator 720 performs motion compensation on the previously reconstructed reference frames F _MB ′ and F _NB ′ stored in the buffer 725 using the extracted motion vectors MV to create a predicted image F _PB , which is then sent to the adder 715.

When bidirectional prediction is used, the predicted image F _{PB is} calculated by averaging reference frames with compensated motion. When using unidirectional prediction, the predicted image F _{PB is} obtained as a reference frame with compensated movement.

An adder 715 adds the input frames F _RB and F _PB to output the reconstructed image F _o ′, which is then stored in the buffer 725.

The inverse quantizer 745 applies inverse quantization to FGS level texture data F _Δ ^Q , and the inverse transformer 750 inverts the inverse quantized result F _Δ ^T to obtain a reconstructed frame F _Δ (F _Δ '), which is then supplied to the frame recovery device 755 .

Motion vector converter 730 lowers the accuracy of the extracted MV motion vector. For example, a motion vector MV with an accuracy of 1/4 pixel may have an accuracy of 0, 0.25, 0.5, or 0.75 pixels. The motion vector converter 730 changes the motion vector MV with an accuracy of 1/4 pixel to the motion vector MV ₁ with an accuracy of less than 1/4 pixel, such as 1/2 pixel or 1 pixel.

The motion compensator 735 uses the modified motion vector MV ₁ to perform motion compensation on the reconstructed basic layer frames F _MB ′ and F _NB received from the buffer 725 and on the reconstructed reference frames F _MF ′ and F _NF ′ of the FGS level received from the buffer 740, and feeds the compensated motion frame mc (F _MB '), mc (F _NB ') and the motion compensated frame mc (F _MF '), mc (F _NF ') to the frame recovery device 755.

When the MV motion vector with an accuracy of, for example, 1/2 pixel is used during motion compensation, a bilinear filter that requires little computation can be used for interpolation instead of the six-pole filter used in the H.264 standard. The interpolation process has little effect on compression efficiency.

The frame recovery device 755 calculates the difference Δ _M between the reference frames with the compensated movement of the FGS level and the main level mc (F _MF ') and mc (F _MB '), that is, Δ _M = mc (F _MF ') - mc (F _MB ' ) and the difference between the reference frames with the compensated movement of the FGS level and the main level mc (F _NF ') and mc (F _NB '), that is, Δ _N = mc (F _NF ') - mc (F _NB '). Of course, when a unidirectional reference frame is used, only one difference can be calculated.

The frame recovery device 755 also calculates the average value of the differences Δ _M and Δ _N and adds the average value, F _Δ ', and the reconstructed main level image F _o ' to create a reconstructed image F _OF 'of the FGS level. When a unidirectional reference frame is used, an averaging process is not required.

Buffer 740 then stores the reconstructed image F _OF '. Of course, pre-reconstructed images F _MF 'and F _BF ' can be stored in buffer 740.

Alternatively, the video decoder corresponding to the second example of an embodiment of the present invention may have the same configuration and perform the same operation as shown in FIG. 8, except for the operation performed by the frame recovery apparatus. That is, the frame recovery device according to the second example of the embodiment creates a predicted frame for each level before calculating the difference between frames at these two levels. That is, we can say that the frame recovery device creates a predicted frame at the FGS level and a predicted frame at the main level using reference frames of the FGS level with compensated motion and frames of the main level with compensated motion. Predicted frames can be created by simply averaging two reference frames with compensated movement. Of course, when using unidirectional prediction, the predicted frame is the frame itself with the compensated movement.

The frame recovery device then calculates the difference between the predicted frames and adds together the texture data, the reconstructed main level frame, and the difference.

FIG. 9 shows a block diagram of a video decoder 900 corresponding to a third example of an embodiment of the present invention. With reference to Fig. 9, in contrast to the first and second examples of embodiments in which motion compensation is performed before calculating the difference between the reference image of the FGS level and the reference image of the main level, video decoder 900 performs motion compensation after calculating the difference between the reference frames on these two levels. In order to avoid re-explanation, the following description will focus on the distinguishing features of the first example embodiment shown in FIG.

Subtractor 960 subtracts the reconstructed reference frames F _MB 'and F _NB ' received from the buffer 925 from the reference frames F _MF ', F _NF ' of the FGS level, and provides subtraction results F _MF '- F _MB ' and F _NF '- F _NB 'on the compensator 935 movement. When a unidirectional reference frame is used, there is only one difference.

The motion compensator 935 uses the modified motion vector MV ₁ received from the motion vector transducer 930 to perform motion compensation on the differences F _MF '- F _MB ' and F _NF '- F _NB ' between the reference frames at the FGS level and at the main level received from a subtractor 960. When a motion vector MV ₁ with 1/2 accuracy is used during motion compensation, a bilinear filter that requires a small amount of computation can be used for interpolation instead of the six-pole filter used in the H.264 standard. As described above, interpolation has little effect on compression efficiency.

The frame recovery device 955 calculates the average value of the differences of the compensated movement, that is, the average value between mc (F _MF '- F _MB ') and mc (F _NF '- F _NB ') and adds to the calculated average value, F _Δ ', taken from the inverse transformer 950, and the reconstructed image F _o 'of the main level. When using a unidirectional reference frame, the averaging process is not required.

10 and 11 are block diagrams of examples of video decoders 1000 and 1200 corresponding to a fourth example of an embodiment of the present invention.

With reference to FIGS. 10 and 11, corresponding to FIGS. 8 and 9, showing video decoders 700 and 900 corresponding to the first and third exemplary embodiments, frame recovery devices 1055 and 1255 add a predicted base layer frame F _PB instead of a reconstructed base frame F _o ′ level.

The video decoders 1000 and 1200 of the fourth example embodiment shown in FIGS. 10 and 11 correspond to decoders of the first and third example embodiments shown in FIGS. 8 and 9, respectively.

With reference first to FIG. 10 corresponding to FIG. 8, the motion compensator 1020 supplies the reference layer image F _PB to the frame recovery device 1055 instead of the reconstructed image F _o ′. Thus, the frame reconstructor 470 adds together the device F _Δ ', received from inverse converter 1050, the predicted image F _PB core layer and the average value of the differences Δ _M and Δ _N to obtain the reconstructed image F _OF' of the core layer.

Similarly, with reference to FIG. 11, the file recovery device 1255 adds together F _Δ 'received from the inverter 1250, the predicted main level image F _PB received from the motion compensator 1220, and the average difference value mc (F _MF ' - F _MB ') and mc (F _NF ' - F _NB ') with a compensated movement to obtain a reconstructed image of the basic level F _OF '.

Meanwhile, the video decoder of the fourth example embodiment, corresponding to the video decoder of the second example of the embodiment (not shown), can have the same configuration and perform the same operation as shown in Fig. 8, except for the operation performed by the frame recovery device 1255. In the video decoder of the fourth example of the embodiment corresponding to the second example of the embodiment, the frame recovery device 1255 creates the predicted frame F _{PF of} the FGS layer and the predicted frame F _{BF of the} main layer using the reference frames mc (F _MF ') and mc (F _NF ') of the FGS level with compensated movement and reference frames mc (F _MB '), mc (F _NB ') of the main level with compensated movement. The frame recovery device 1255 also calculates the difference F _PF -F _PB between the predicted FGS level frame F _P and the main level predicted frame F _PB and adds to F _o ′ received from the inverter 1250, the predicted image F _PB received from the motion compensator 1220, and a difference F _PF -F _PB to obtain a reconstructed image F _OF '.

When applying transmission prediction (fifth example of an embodiment), the frame recovery device 1255 multiplies the weight coefficient α by the difference F _PF -F _{PB of the} intermediate layer and adds together F _Δ 'F _o ' the product α x (F _PF -F _PB ) to obtain F _OF '.

12 is a block diagram of a system for performing an encoding or decoding process using a video encoder 100, 300, 400, 600 or a video decoder 700, 900, 1000, 1200 in accordance with an example embodiment of the present invention. The system may be a television, a television set-top box (STB), a desktop computer, laptop or portable computer, a specialized handheld computer (PDA), a storage device for video information or images (for example, a VCR or digital video recorder). The system may be a combination of the devices listed above, or another device containing them. In addition, the system may be a combination of the above devices or one of the devices, which contains part of another device from the above. The system comprises at least one video source 1310, at least one input / output module 1320, a processor 1340, a storage device 1350, and a display unit 1330.

The video source 1310 may be a television receiver, VCR, or other device that stores video information. Video source 1310 may mean at least one network connection for receiving video information or images from a server using the Internet, wide area network (WAN), local area network (LAN), territorial broadcasting system, cable network, satellite communications network, radio network, telephone network or the like. In addition, the video source 1310 may be a combination of networks or one network containing a portion of another network among networks.

An input / output device 1320, a processor 1340, and a storage device 1350 communicate with each other through a communication medium 1360. The communication medium 1360 may be a communication bus, a communication network, or at least one internal communication circuit. The input video data received from the video source 1310 may be processed by the processor 1340 using at least one program that is stored in the storage device 1350 and may be executed by the processor 1340 to create output video data supplied to the display unit 1330.

In particular, the program stored in the storage device 1350 includes a pulse-scalable codec that implements the method of the present invention. The codec may be stored in memory 1350, read from a storage medium, such as read-only memory on a compact disc (CD-ROM) or floppy disk, or may be downloaded from a predetermined server via multiple networks.

Industrial applicability

As described above, the present invention provides video coding, which can significantly reduce the amount of computation required to implement the PFGS algorithm. Since, according to the present invention, the decoding process is changed in accordance with the video encoding process, the present invention can be applied to a standardized H.264 SE document.

Although the present invention has been shown and described, in particular with reference to examples of its embodiments, it will be understood by those skilled in the art that various changes can be made in the form and details without departing from the scope and spirit of the present invention, as they are defined by the claims below.

Claims (49)

1. A video coding method supporting accurate scalability in quality (FGS), the method comprises the steps of obtaining a predicted image for the current frame using a first motion vector estimated with a predetermined accuracy; calculating the difference between the current frame and the predicted image; performing quantization of the difference between the current frame and the predicted image, performing inverse quantization of the quantized difference between the current frame and the predicted image, and creating a reconstructed image for the current frame by adding the inverted transformed result of the quantized difference with the predicted image for the current frame; performing motion compensation on the reference frame at the FGS level and on the reference frame at the main level, using a second motion vector with lower accuracy than the first motion vector; calculating the difference between the reference frame at the FGS level with compensated movement and the reference frame at the main level with compensated movement; subtracting the reconstructed image for the current frame and the calculated difference from the current frame and encode the result of the subtraction. 2. The method according to claim 1, in which the execution of the motion compensation comprises creating a second motion vector by changing the accuracy of the first motion vector, and the accuracy of the second motion vector used in performing the motion compensation is lower than the accuracy of the first motion vector used in obtaining predicted image for the current frame. 3. The method according to claim 1, in which the calculated difference is the average value of the first difference between the subsequent reference frame of the FGS level and the subsequent reference frame of the main level and the second difference between the previous reference frame of the FGS level and the previous reference frame of the main level. 4. The method according to claim 2, in which if interpolation is performed to compensate for the movement, then an interpolation filter of a different type is used for interpolation than the filter used to obtain the predicted image of the current frame. 5. The method according to claim 1, in which the encoding of the subtraction result comprises the steps of converting the subtraction result to create a conversion coefficient; quantization of the transform coefficient is performed to create a quantization coefficient, and a lossless quantization coefficient is encoded. 6. The method according to claim 1, in which obtaining a predicted image for the current frame comprises the steps of evaluating the first motion vector using the current frame and at least one reconstructed frame of the main level as a reference frame; performing motion compensation on the reference frames using the first motion vector; get the predicted image by averaging the reference frames with compensated movement. 7. The method according to claim 1, in which obtaining a predicted image for the current frame comprises the steps of evaluating the first motion vector using the current frame and the original frame adjacent to the current frame as a reference frame; performing motion compensation on the reference frame using the first motion vector, and a predicted image is obtained by averaging the reference frames with compensated movement. 8. The method according to claim 1, in which the reference frame at the FGS level is the source frame adjacent to the reference frame of the FGS level, and the reference frame of the main level is an adjacent frame restored from the main level. 9. The method according to claim 1, in which the reference frame at the FGS level is an adjacent frame recovered from the FGS level, and the reference frame of the main level is an adjacent frame recovered from the main level. 10. The method according to claim 5, in which the quantization step size used in the quantization of the transform coefficient is smaller than that used in the quantization of the difference. 11. A video coding method supporting an accurate quality scalability algorithm (FGS), comprising the steps of obtaining a predicted image for the current frame using a first motion vector estimated with a predetermined accuracy; calculating the difference between the current frame and the predicted image; performing quantization of the difference between the current frame and the predicted image, performing inverse quantization of the quantized difference between the current frame and the predicted image, and creating a reconstructed image for the current frame by adding the inverted transformed result of the quantized difference with the predicted image for the current frame; performing motion compensation on the reference frame of the FGS level and the reference frame of the main level using the second motion vector with lower accuracy than the first motion vector, and create a predicted frame for the FGS level and a predicted frame for the main level, respectively; calculating the difference between the predicted frame for the FGS level and the predicted frame for the main level; subtracting the reconstructed image and the difference from the current frame; and encode the result of the subtraction. 12. The method according to claim 11, in which the execution of the motion compensation comprises creating a second motion vector by changing the accuracy of the first motion vector, and the accuracy of the second motion vector used in performing the motion compensation is lower than the accuracy of the first motion vector used in obtaining the predicted images for the current frame. 13. The method according to claim 11, in which the predicted frame of the FGS level is the average value of the reference frames of the F6S level with compensated movement, and the predicted frame of the main level is the average value of the reference frames of the main level with compensated movement. 14. The method according to item 12, in which if the interpolation is performed to compensate for the movement, then the filter type is used for interpolation, different from the interpolation filter, which is used to obtain the predicted image for the current interpolation frame. 15. The method according to claim 11, in which the encoding of the subtraction result comprises the steps of converting the subtraction result to create a conversion coefficient; quantization of the transform coefficient is performed to create a quantization coefficient, and a lossless quantization coefficient is encoded. 16. The method of claim 15, wherein the quantization step size used in quantizing the transform coefficient is smaller than the step size used in quantizing the difference. 17. A video coding method supporting an accurate quality scalability algorithm (FGS), comprising the steps of obtaining a predicted image for the current frame using a first motion vector estimated with a predetermined accuracy; calculating the difference between the current frame and the predicted image; performing quantization of the difference between the current frame and the predicted image, inverse quantization of the quantized difference between the current frame and the predicted image, and creating a reconstructed image for the current frame by adding the inverted transformed result of the quantized difference with the predicted image for the current frame; calculating the difference between the reference frame of the exact scalability level in quality (FGS) and the reference frame of the main level; performing motion compensation for the difference using the second motion vector with lower accuracy than the first motion vector; subtracting the reconstructed image for the current frame and the result of motion compensation from the current frame; encode the result of the subtraction. 18. The method according to 17, in which performing motion compensation for the difference comprises creating a second motion vector by changing the accuracy of the first motion vector, and the accuracy of the second motion vector used when performing motion compensation for the difference is lower than the accuracy of the first motion vector used upon receipt of the predicted image for the current frame. 19. The method according to 17, in which the result of the motion compensation for the difference to be subtracted is the average value of the differences with compensated movement. 20. The method according to p. 18, in which, if the interpolation is performed to compensate for the difference, then the filter type is used for interpolation, different from the interpolation filter, which is used to obtain the predicted image for the current frame. 21. The method according to 17, in which the encoding of the subtraction result comprises the steps of converting the subtraction result to create a conversion coefficient; quantizing a transform coefficient to create a quantization coefficient; encode lossless quantization coefficient. 22. The method according to item 21, in which the quantization step size used in the quantization of the converted coefficient is smaller than the step size used in the quantization of the difference. 23. A video coding method supporting an accurate quality scalability algorithm (FGS), comprising the steps of obtaining a predicted image for the current frame using a first motion vector estimated with a predetermined accuracy; performing motion compensation on the reference frame of the FGS level and the reference frame of the main level using the second motion vector with lower accuracy than the accuracy of the first motion vector; calculating the difference between the reference frame of the FGS level with compensated movement and the reference frame of the main level; subtracting the predicted image and the difference from the current frame; encode the result of the subtraction. 24. A video coding method supporting an accurate quality scalability algorithm (FGS), comprising the steps of obtaining a predicted image for the current frame using a first motion vector estimated with a predetermined accuracy; performing motion compensation on the reference frame of the FGS level and the reference frame of the main level using the second motion vector with lower accuracy than the accuracy of the first motion vector, and create a predicted frame for the FGS level and a predicted frame for the main level, respectively; calculating the difference between the predicted frame for the FGS layer and the reference frame for the main layer; subtracting the predicted image and the calculated difference from the current frame; encode the result of the subtraction. 25. The method according to paragraph 24, further comprising the step of multiplying the difference between the predicted frame of the FGS level and the predicted frame of the main level by the weight coefficient a, in which the calculated difference when subtracting the predicted image is the product of the weight coefficient a and the difference between the predicted frame for the FGS level and predicted frame for the main level. 26. A video coding method supporting an accurate quality scalability algorithm (FGS), comprising the steps of obtaining a predicted image for the current frame using a first motion vector estimated with a predetermined accuracy; calculating the difference between the reference frame of the FGS layer and the reference frame of the main level; performing motion compensation for a difference between the reference frame of the FGS level and the reference frame of the main level using the second motion vector with lower accuracy than the accuracy of the first motion vector; creating a reconstructed image for the current frame by adding the inverted transformed result of the quantized difference with the predicted image for the current frame; subtracting the reconstructed image and the result of motion compensation from the current frame; encode the result of the subtraction. 27. A video decoding method supporting an accurate quality scalability algorithm (FGS), comprising the steps of extracting the texture layer data of the main layer and texture data of the FGS layer and the first motion vector from the input bit stream; restoring the frame of the main level from the texture data of the main level, while restoring the frame of the main level contains the steps of performing inverse quantization of texture data of the main level; performing inverse transformation of the inverse quantization result; creating a predicted image from a previously reconstructed reference frame of the main level using the first motion vector; and add together the predicted image and the result of the inverse transform; performing motion compensation on the reference frame of the FGS level and the reference frame of the main level using the second motion vector with lower accuracy than the first motion vector; calculating the difference between the reference frame of the FGS level with compensated movement and the reference frame of the main level with compensated movement; add together the frame of the main level, texture data of the FGS level and the difference. 28. The method according to item 27, in which the second motion vector used when performing motion compensation, has a lower accuracy than the first motion vector. 29. The method according to item 27, in which the calculated difference is the average value of the first difference between the previous reference frame of the FGS level and the previous reference frame of the reference level and the second difference between the subsequent reference frame of the FGS level and the subsequent reference frame of the main level. 30. The method according to p, in which, if interpolation is used to compensate for the movement, then an interpolation filter of a different type than that used when restoring the frame of the main level is used for interpolation. 31. The method according to item 27, in which the texture data of the FGS level, used when adding to the frame of the main level, are obtained by performing inverse quantization and inverse transformation of the extracted texture data of the FGS level. 32. The method of claim 31, wherein the quantization step size used in the inverse quantization applied to the texture data of the FGS layer is smaller than the step size used in the inverse quantization performed when reconstructing the main layer frame. 33. The method of video decoding that supports the algorithm for accurate scalability in quality (FGS), the method comprises the steps of extracting texture data of the main layer and texture data of the FGS level and the first motion vector from the input bit stream; restoring the frame of the main level from the texture data of the main level, while restoring the frame of the main level contains the steps of performing inverse quantization of texture data of the main level; performing inverse transformation of the inverse quantization result; creating a predicted image from a previously reconstructed reference frame of the main level using the first motion vector; and add together the predicted image and the result of the inverse transform; performing motion compensation on the reference frame of the FGS level and the reference frame of the main level using a second motion vector with lower accuracy than the first motion vector, and creating a predicted frame of the FGS level and a predicted frame of the main level; calculating the difference between the predicted reference frame of the FGS level and the predicted reference frame of the main level; and add together the texture data, the reconstructed frame of the main level and the difference. 34. The method of claim 33, wherein the second motion vector used to perform the motion compensation has lower accuracy than the first motion vector. 35. The method according to clause 34, in which, if interpolation is used to compensate for the movement, then an interpolation filter of a different type than that used when restoring the frame of the main level is used for interpolation. 36. The method according to clause 33, in which the texture data of the FGS level used when adding the frame of the main level, texture data of the FGS level and the remainder are obtained by performing inverse quantization and inverse conversion of the extracted texture data of the FGS level. 37. A video decoding method supporting an accurate quality scalability algorithm (FGS), comprising the steps of extracting texture level data and texture level FGS data and a first motion vector from an input bit stream; restoring the frame of the main level from the texture data of the main level, while restoring the frame of the main level contains the steps of performing inverse quantization of texture data of the main level; performing inverse transformation of the inverse quantization result; creating a predicted image from a previously reconstructed reference frame of the main level using the first motion vector; and add together the predicted image and the result of the inverse transform; calculating the difference between the reference frame of the FGS layer and the reference frame of the main level; performing motion compensation on the difference using the second motion vector with lower accuracy than the first motion vector, and add together the texture data of the FGS level, the reconstructed frame of the main level, and the result of the motion compensation. 38. The method according to clause 37, in which the result of the motion compensation subjected to addition is the average value of the differences with compensated movement. 39. The method according to clause 37, in which the second motion vector used when performing motion compensation, has a lower accuracy than the first motion vector. 40. The method according to § 39, in which, if interpolation is performed to compensate for the movement, then an interpolation filter is used for interpolation, different from the interpolation filter used when restoring the frame of the main level. 41. The method according to clause 37, in which the FGS level texture data for addition is obtained by performing inverse quantization and inverse transformation on the extracted FGS level texture data. 42. A video decoding method supporting an accurate quality scalability algorithm (FGS), the method comprises the steps of extracting texture layer data of a basic layer, texture data of an FGS layer and a first motion vector from an input bit stream; restoring the predicted image for the frame of the main level from the texture data of the main level using the first motion vector; performing motion compensation on the reference frame of the FGS level and the reference frame of the main level using the second motion vector with an accuracy lower than the accuracy of the first motion vector; calculating the difference between the reference frame of the FGS level with compensated movement and the reference frame of the main level with compensated movement; add together the texture data of the FGS level, the predicted image and the difference between the reference frame of the FGS level with compensated movement and the reference frame of the main level with compensated movement. 43. A video decoding method supporting an accurate quality scalability algorithm (FGS), comprising the steps of extracting texture level data and FGS level texture data and first motion vectors from an input bit stream; restoring the predicted image for the frame of the main level from the texture data of the main level using the first motion vector; performing motion compensation on the reference frame of the FGS level and the reference frame of the main level using a second motion vector with an accuracy lower than the first motion vector, and creating a predicted frame of the FGS level and a predicted frame of the main level; calculating the difference between the predicted frame of the FGS level and the predicted frame of the main level; and add together the texture data of the FGS level, the predicted images and the calculated difference. 44. The method according to item 43, further comprising the step of multiplying the difference between the predicted frame of the FGS level and the predicted frame of the main level by the weight coefficient a, in which the calculated difference used in addition is the product of the weight coefficient a and the difference between the predicted frame of the FGS level and predicted frame of the main level. 45. The method of video decoding that supports the algorithm for accurate scalability in quality (FGS), comprising the steps of extracting texture data of the main level, texture data of the FGS level and the first motion vector from the input bit stream; restoring the predicted image for the frame of the main level from the texture data of the main level using the first motion vector; calculating the difference between the reference frame of the FGS layer and the reference frame of the main level; performing motion compensation for the difference using the second motion vector with an accuracy lower than the accuracy of the first vector; and add together the texture data of the FGS level, the predicted image and the difference. 46. A video encoder based on the Fine Quality Scalability Algorithm (FGS), comprising means for obtaining a predicted image for a current frame using a first motion vector estimated with predetermined accuracy; means for calculating the difference between the current frame and the predicted image; means for quantizing the difference between the current frame and the predicted image; means for inverting quantization of the quantized difference between the current frame and the predicted image; means for creating a reconstructed image for the current frame; motion compensation means for the reference frame of the FGS layer and the reference frame of the main layer using the second motion vector; means for calculating the difference between the reference frame of the FGS level with compensated movement and the reference frame of the main level with compensated movement; means for subtracting the reconstructed image and the difference from the current frame; means for encoding the result of the subtraction. 47. The encoder of claim 46, wherein the accuracy of the second motion vector used to perform the motion compensation is lower than the accuracy of the first motion vector used to obtain the predicted image for the current frame. 48. A video decoder based on the Fine Quality Scalability Algorithm (FGS), comprising: means for extracting core layer texture data, FGS layer texture data and a first motion vector from an input bit stream; means for reconstructing a frame of the base layer from texture data of the base layer; means of inverse quantization of texture data of the main level; means for inverting the inverse quantization result; means for performing motion compensation for the reference frame of the FGS level and the reference frame of the main level using the second motion vector with lower accuracy than the first motion vector; means for creating a predicted frame of the FGS level and a predicted frame of the main level; means for calculating the difference between the predicted frame of the FGS level and the predicted frame of the main level; and means for adding together the texture data, the reconstructed frame of the base layer and the difference. 49. The video decoder of claim 48, wherein the accuracy of the second motion vector used to perform the motion compensation is lower than the accuracy of the first motion vector extracted from the input bit stream.

Images