RU2325783C2 - Adaptive weighing of reference images during video signal coding - Google Patents

Abstract

FIELD: physics.

SUBSTANCE: said utility invention relates to video encoders and, in particular, to the use of adaptive weighing of reference images in video encoders. A video encoder and a method of video signal data processing for an image block and the specific reference image index for predicting this image block are proposed, which use the adaptive weighing of reference images to increase the video signal compression, the encoder having a reference image weighting factor assigning module for the assignment of the weighting factor corresponding to the said specific reference image index.

EFFECT: increased efficiency of reference image predicting.

8 cl, 7 dwg

Description

Cross reference to related applications

This application claims the priority of a provisional US patent application with registration number 60 / 395.843 (attorney registry number PU020340) entitled "Adaptive Weighting Of Reference Pictures In Video CODEC" and filed July 15, 2002, which is fully incorporated by reference in this application links. Additionally, this application claims the priority of a provisional US patent application with registration number 60 / 395.874 (attorney's registry number PU020339), entitled "Motion Estimation With Weighting Prediction" and filed July 15, 2002, which is fully incorporated in the materials of this application by links.

FIELD OF THE INVENTION

The present invention relates to video encoders and, in particular, to the use of adaptive weighting of reference images in video encoders.

State of the art

Basically, video information is processed and transmitted in the form of bit streams. Conventional decoders and video compression encoders ("CODEC" and) achieve the greatest efficiency of their compression by generating a prediction of the reference image for the image to be encoded and encoding the difference between the current image and the prediction. The stronger the correlation of the prediction with the current image, the fewer bits are required to compress this image, thereby increasing the efficiency of the process. Accordingly, it is preferable to generate the best possible prediction of the reference image.

Many video compression standards, including the standards of the MPEG-1, MPEG-2, and MPEG-4 Moving Image Expert Group, use the motion-compensated version of the previous reference image as the prediction for the current image and only the difference between the current image and the prediction is encoded. When a single image prediction (“P” image) is used, the reference image is not scaled to form a motion-compensated prediction. When bidirectional image predictions ("B" images) are used, intermediate predictions are formed from two different images, and then two intermediate predictions are averaged using equal weights in (1/2, 1/2) for each, to generate one averaged prediction . In the mentioned MPEG standards, of the two reference images for B-images, one always corresponds to the forward direction and one to the backward direction.

SUMMARY OF THE INVENTION

Mentioned and other negative aspects and disadvantages of the prior art are eliminated by means of a system and method for adaptive weighting of reference images in video encoders and video decoders.

A video encoder and corresponding methods for processing video signal data for an image block and an index of a specific reference image for predicting this image block are disclosed, which use adaptive weighting of the reference images to enhance video compression. The encoder comprises a module for assigning a weight coefficient to a reference image for assigning a weight coefficient to said index of a specific reference image.

A suitable encoding method for video signal data for an image block includes receiving a substantially uncompressed image block and assigning a weight coefficient to the image block corresponding to a particular reference image having a corresponding index. The motion vectors corresponding to the difference between the image block and the specific reference image are calculated. In accordance with the motion vectors, the motion compensation of the specific reference image is performed, and the motion-compensated reference image is changed by the assigned weight coefficient to form a weighted motion-compensated reference image. A substantially uncompressed image block is compared with a weighted motion-compensated reference image, and a signal indicating the difference between a substantially uncompressed image block and a weighted motion-compensated reference image, together with the corresponding index of the specific reference image, is encoded.

LIST OF DRAWINGS

Adaptive weighting of reference images in video encoders and video decoders in accordance with the principles of the present invention is illustrated by the following illustrative drawings:

Figure 1 is a block diagram for a standard video decoder.

2 is a block diagram for an adaptive dual prediction video decoder.

Figure 3 is a block diagram for a video decoder with weighting a reference image in accordance with the principles of the present invention.

4 is a block diagram for a standard video encoder.

5 is a block diagram for a video encoder with weighting a reference image in accordance with the principles of the present invention.

6 is a flowchart of a decoding process in accordance with the principles of the present invention.

7 is a flowchart of an encoding process in accordance with the principles of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

The present invention provides an apparatus and method for estimating a motion vector and adaptive weighting of a reference image. In some video sequences, in particular fading video sequences, the current image or image block to be encoded is more strongly correlated with the reference image, the scaled weight coefficient, than directly with the reference image. Video codecs without weighting reference images for coding fading sequences are very inefficient. When using weighting coefficients in coding, a video encoder should determine both weighting coefficients and motion vectors, but the best choice for each of them depends on the other, with motion estimation, which is usually the most intensive part of a digital video compression encoder in computational terms.

In the proposed Video Compression Standard of the Joint Video Signal Group (JVT), each P-image can use several reference images to generate image prediction, but each individual motion block, or 8 × 8 macroblock area, uses only one reference image for prediction. In addition to encoding and transmitting motion vectors, for each motion block, or 8 × 8 region, a reference image index is transmitted indicating which reference image is being used. A limited set of possible reference images is stored in the encoder and decoder and the number of valid reference images is transmitted.

In the JVT standard for dual-predicted images (also called B-images), two predictors are formed for each motion block, or 8 × 8 region, each of which can be from a separate reference image, and these two predictors are averaged to form one averaged the predictor. For motion blocks encoded with double prediction, both reference images may correspond to the forward direction, both may correspond to the backward direction or one to the forward direction and one to the backward direction. Two lists of available reference images that can be used for prediction are supported. Two reference images are defined as predictors of list 0 and list 1. For each reference image, the index ref_idx_I0 and ref_idx_I1 are encoded and transmitted for the reference images of list 0 and list 1, respectively. Dual Prediction Images, or B-Images, of the Combined Video Signal Group (JVT) enable adaptive weighting of two predictions, i.e.

Pred = [(P0) (Pred0)] + [(P1) (Pred1)] + D,

where P0 and P1 are weights, Pred0 and Pred1 are the predictions of the reference image for list 0 and list 1, respectively, and D is the offset.

Two methods have been proposed for indicating weights. In the first, weights are determined by the directions that are used for the reference images. In this method, if the index ref_idx_I0 does not exceed ref_idx_I1, then the weighting coefficients (1/2, 1/2) are used, otherwise the coefficients (2, -1) are used.

In the second proposed method, an arbitrary number of weights for each section is transmitted. A weighting index is then transmitted for each motion block, or 8 × 8 macroblock region, which uses bidirectional prediction. The decoder uses the received weighting index to select the appropriate weighting factor from the transmitted set, for use in decoding a motion block, or 8 × 8 area. For example, if three weights were transmitted at the partition level, they must correspond to indices 0, 1, and 2 weights, respectively.

The following description only illustrates the principles of the invention. Accordingly, it will be clear that those skilled in the art will be able to invent various means that, although not described and not explicitly presented here, embody the principles of the invention and are included in its scope. In addition, all possible options and a conditional language set forth here are mainly intended specifically for educational purposes, in order to facilitate understanding of the principles of the invention and concepts introduced by the inventor in the development of technology, and should not be construed as possible options and conditions imposing restrictions. In addition, all descriptions of the principles, aspects, and embodiments of the invention set forth herein, as well as their specific possible options, are intended to cover their structural and functional equivalents. Additionally, it is understood that such equivalents include those currently known and those that will be developed in the future, i.e. any designed elements that perform an identical function regardless of structure.

So, for example, it should be understood by those skilled in the art that block diagrams here represent conceptual views of illustrative diagrams embodying the principles of the invention. Similarly, it is clear that any flowcharts, algorithms, state transition diagrams, pseudo-code, etc. represent various processes that, in essence, can be represented in a computer-readable medium, as well as executed by a computer or processor, regardless of whether such a computer or processor is explicitly depicted.

The functions of the various elements depicted in the drawings can be provided through the use of specialized hardware, as well as hardware that can run the software, together with the corresponding software. When provided with a processor, functions can be provided by one specialized processor, one shared processor, or several separate processors, some of which can be shared. In addition, the explicit use of the term “processor” or “controller” should not be construed as a reference solely to hardware that can run software, and may implicitly include, but not in a limiting sense, the hardware of a digital signal processor (DSP, DSP), read-only memory (ROM) for storing software, random access memory (RAM, RAM) and non-volatile memory. Other hardware, standard and / or custom, may also be included. Similarly, any switches shown in the drawings are only conceptual. Their function can be performed through the operation of software logic tools, specialized logic tools, the interaction of software control and specialized logic tools, or even manually, a specific method can be selected by the means of the implementing party, as is more specifically understood from the context.

In the claims, any element that acts as a means of performing a specific function is intended to encompass any means of performing that function, including, for example, a) a combination of circuit elements performing the specified function or b) any form of software to perform the function, including, therefore, firmware, microcode, etc., combined with appropriate circuits for executing this software. The invention, as defined by the aforementioned claims, is in fact intrinsic to the extent that the functionality provided by the various described means is combined and connected together as required by the claims. Accordingly, the applicant considers any means that can provide the mentioned functionality in the form of an equivalent to the functionality presented here.

Referring to FIG. 1, a standard video decoder is denoted by a reference numeral 100. Video decoder 100 comprises a variable length field (VLD) decoder 110 coupled to exchange signals with inverse quantizer 120. Inverse quantizer 120 is coupled to exchange signals with inverse converter 130. Inverse converter 130 is connected with the possibility of exchanging signals with the first input of the adder, or summing connection 140, while the output of the summing connection 140 provides the output of the video decoder 100. The output of the summing with Units 140 are coupled to a repository of 150 reference images. The repository 150 of reference images is connected with the possibility of exchanging signals with a compensator 160 movement, which is connected with the possibility of exchanging signals with the second input of the summing connection 140.

2, an adaptive dual-predicted video decoder is generally designated 200. Video decoder 200 comprises a VLD 210 coupled to a signal quantizer 220. A quantizer 220 is coupled to a signal converter 230. A converter 230 is coupled to the ability to exchange signals with the first input of the summing connection 240, while the output of the summing connection 240 provides the output of the video decoder 200. The output of the summing connection 240 is connected to zmozhnostyu signal communication with reference picture store 250. The storage 250 of reference images is connected with the possibility of exchanging signals with a compensator 260 movement, which is connected with the possibility of exchanging signals with the first input of the multiplier 270.

The VLD 210 is further coupled with a signal exchange means to a weighting means search tool 280 for providing a reference image to provide an adaptive dual prediction coefficient (ADP) index search means 280. The first output of the search tool 280 is designed to provide a weighting factor and is connected with the possibility of exchanging signals with the second input of the multiplier 270. The output of the multiplier 270 is connected with the possibility of exchanging signals with the first input of the summing connection 290. The second output of the search tool 280 is designed to provide bias and is connected with the possibility of signal exchange with the second input of the summing connection 290. The output of the summing connection 290 is connected with the possibility of exchanging signals with the second input of the summing connection 24 0.

3, a video decoder with reference image weighting is generally designated 300. Video decoder 300 includes a VLD 310 that is coupled to the inverse quantizer 320. The inverse quantizer 320 is coupled to the inverter 330. The inverter 330 is connected to the ability to exchange signals with the first input of the summing connection 340, while the output of the summing connection 340 provides the output of the video decoder 300. The output of the summing connection 340 is connected with the ability to exchange signals with a repository of 350 reference images. The storage 350 of reference images is connected with the possibility of exchanging signals with a compensator 360 motion, which is connected with the possibility of exchanging signals with the first input of the multiplier 370.

The VLD 310 is further coupled to a signal exchange with a weighting means search tool 380 for providing a reference image to provide a means for index search tool 380. The first output of the search means 380 is designed to provide a weighting factor and is connected with the possibility of exchanging signals with the second input of the multiplier 370. The output of the multiplier 370 is connected with the possibility of exchanging signals with the first input of the summing connection 390. The second output of the search means 380 is designed to provide bias and is connected with the possibility of signal exchange with the second input of the summing connection 390. The output of the summing connection 390 is connected with the possibility of exchanging signals with the second input of the summing connection 390 340.

4, a standard video encoder is generally designated 400. The input to the encoder 400 is interconnected with a non-inverting input of the summing connection 410. The output of the summing connection 410 is signal-exchanged with the block converter 420. A converter 420 is coupled to a quantizer 430. The output of a quantizer 430 is coupled to a variable length field encoder (VLC) 440, where the output of the VLC 440 is an encoder 400 that is accessible from the outside.

The output of the quantizer 430 is additionally connected with the possibility of exchanging signals with the inverse quantizer 450. The inverse quantizer 450 is connected with the possibility of exchanging signals with the inverse transformer 460 of the blocks, which, in turn, is connected with the possibility of exchanging signals with the storage 470 of reference images. The first output of the storage 470 reference images is connected with the possibility of exchanging signals with the first input of the module 480 motion estimation. The input to the encoder 400 is further coupled to the second input of the motion estimation module 480. The output of motion estimation module 480 is coupled to a signal exchange with a first input of motion compensator 490. The second output of the storage 470 reference images is connected with the possibility of exchanging signals with the second input of the compensator 490 motion. The output of the motion compensator 490 is connected to exchange signals with an inverting input of the summing connection 410.

5, a video encoder with reference image weighting is indicated generally at 500. The input to the encoder 500 is interconnected with a non-inverting input of the summing connection 510. The output of the summing connection 510 is signal-exchanged with the block converter 520. A converter 520 is coupled to a quantizer 530. The output of a quantizer 530 is coupled to a VLC 540, where the output of the VLC 440 is an output of the encoder 500, accessible from the outside.

The output of the quantizer 530 is additionally connected with the possibility of exchanging signals with the inverse quantizer 550. The inverse quantizer 550 is connected with the possibility of exchanging signals with the inverse converter 560 of the blocks, which, in turn, is connected with the possibility of exchanging signals with the storage 570 of reference images. The first output of the storage 570 reference images is connected with the possibility of exchanging signals with the first input of module 572 for assigning the weight coefficient of the reference image. The input to the encoder 500 is additionally connected with the possibility of exchanging signals with the second input of the module 572 assigning the weight coefficient of the reference image. The output of the reference image weight assignment module 572 indicating the weight is connected to exchange signals with a first input of the motion estimation module 580. The second output of the storage 570 reference images is connected with the possibility of exchanging signals with the second input of the module 580 motion estimation.

The input to the encoder 500 is additionally connected with the possibility of exchanging signals with the third input of the motion estimation module 580. The output of the motion estimation module 580 indicating the motion vectors is coupled to the signal exchange with the first input of the motion compensator 590. The third output of the storage 570 reference images is connected with the possibility of exchanging signals with the second input of the compensator 590 motion. The output of the motion compensator 590 indicating a motion-compensated reference image is connected for signal exchange with the first input of the multiplier 592. The output of the reference image weight assignment module 572 indicating the weight coefficient is connected for signal exchange with the second input of the multiplier 592. Multiplier output 592 is interconnected with an inverting input of a summing connection 510.

6, an exemplary video signal decoding process for an image block is generally designated 600. The process includes a start block 610, where control is passed to the input block 612. At input step 612, data of the compressed image block is received and control is transferred to input step 614. At input step 614, at least one reference image index with data for the image block is received, with each reference image index corresponding to a specific reference image. At input step 614, control is transferred to functional step 616, in which a weight coefficient corresponding to each of the received reference image indices is determined, and control is transferred to optional functional step 617. At optional functional step 617, an offset corresponding to each of the received reference image indices is determined, and transmitted control of functional step 618. At functional step 618, a reference image corresponding to each of the received indices of reference images, and transfer control to the functional step 620. At the functional step 620, in turn, the motion compensation of the extracted reference image is performed and control is transferred to the functional step 622. At the functional step 622, the motion-compensated reference image is multiplied by the corresponding weight coefficient and the control is transferred to the optional functional step 623. At an optional functional step 623, a motion-compensated reference image is summed with the corresponding offset it and passes control to a function block 624. The function block 624, in turn, form a weighted motion compensated reference picture, and passes control to step 626 is complete.

7, an illustrative encoding process of video signal data for an image block is generally designated 700. The process includes a start block 710, where control is passed to the input block 712. At input step 712, essentially, the data of the uncompressed image block is received and control is transferred to functional step 714. At functional step 714, a weight coefficient is assigned to the image block corresponding to a particular reference image having a corresponding index. At a functional step 714, control is transferred to an optional functional step 715. At an optional functional step 715, an offset for an image block corresponding to a particular reference image having a corresponding index is assigned. In optional functional step 715, control is transferred to functional step 716, in which motion vectors corresponding to the difference between the image block and the specific reference image are calculated, and control is transferred to functional step 718. At functional step 718, in accordance with the motion vectors, the motion compensation of the specific reference images and transfer control to the functional step 720. At the functional step 720, in turn, the motion-compensated reference image is multiplied click on the assigned weight coefficient to form a weighted motion-compensated reference image and control is transferred to the optional functional step 721. At the optional functional step 721, in turn, the motion-compensated reference image with the assigned offset is summed to form the weighted motion-compensated reference image and transmitted control of functional step 722. At functional step 722, the weighted compensated by move the reference image from the essentially uncompressed image block and transfer control to functional step 724. At functional step 724, in turn, a signal is encoded with the difference between the essentially uncompressed image block and the weighted motion-compensated reference image along with the corresponding index specific reference image and transfer control to step 726 completion.

In the presented possible embodiment, for each encoded image or section, for each valid reference image, in relation to which blocks of the current image can be encoded, a weight coefficient is compared. When encoding or decoding each individual block in the current image, to form a weighted predictor, the weighting coefficient (s) and the offset (s) corresponding to its reference image indices are applied to the reference prediction. All blocks in a section that are encoded with respect to the same reference image apply the same weight to the prediction of the reference image.

Whether adaptive weighting should be used when encoding an image or not can be indicated in the image parameter set or sequence parameter set or in the image or section header. For each section or image using adaptive weighting, a weight coefficient can be transmitted for each of the valid reference images that can be used to encode that section or image. The number of valid reference images is transmitted in the section header. For example, if three reference images can be used to encode the current section, then up to three weights are transmitted, and they are matched against the reference image with the same index.

If no weights are transmitted, then the default weights are used. In one embodiment of the present invention, when no weight is transmitted, the default weights are used (1/2, 1/2). Weights can be transmitted using fixed or variable length codes.

Unlike standard systems, each weighting factor that is transmitted with each section, block or image corresponds to a specific index of the reference image. Previously, any set of weights transmitted with each section or image has not been matched to any particular reference image. Instead, an adaptive double prediction index was transmitted for each motion block, or 8 × 8 region, to select which of the weights from the transmitted set should be applied to that particular motion block, or 8 × 8 region.

In the present embodiment, the weighting index for each motion block, or 8 × 8 area, is not explicitly transmitted. Instead, a weighting coefficient associated with the transmitted index of the reference image is used. This significantly reduces the amount of unproductive overhead in the transmitted bit stream to enable adaptive weighting of reference images.

The aforementioned system and methodology can be applied to prediction P-pictures that are encoded by one predictor, or to dual-predicted B-pictures that are encoded by two predictors. The decoding processes for the cases of P- and B-images, which take place both in the encoder and in the decoders, are described below. Alternatively, the technique may also be applied to coding systems using concepts like I, B and P-images.

For unidirectional prediction in B-images and for bi-directional prediction in B-images, identical weights can be used. When a single predictor is used for a macroblock, in P-images or for unidirectional prediction in B-images, one index of the reference image is transmitted for this block. After the predictor is created in the motion compensation step in the decoding process, a weight factor is applied to this predictor. Then, the weighted predictor is summed with the encoded remainder and, with respect to the summation result, clipping along the boundaries of the region is performed to form a decoded image. When used for blocks in P-images or for blocks in B-images that use only list 0 prediction, a weighted predictor is formed as

Pred = W0 * Pred0 + D0 (1),

where W0 is the weighting coefficient mapped to the reference image of list 0, D0 is the offset mapped to the reference image of list 0, and Pred0 is the motion-compensated prediction block from the reference image of list 0.

For use for blocks in B-images that use only list 0 prediction, a weighted predictor is formed as

Pred = W1 * Pred1 + D1 (2),

where W1 is the weighting coefficient mapped to the reference image of list 1, D0 is the offset mapped to the reference image of list 1, and Pred1 is the motion-compensated prediction block from the reference image of list 1.

To ensure that the resulting values are within an acceptable range of pixel values, typically from 0 to 255, clipping along the boundaries of the region with respect to weighted predictors can be performed. The accuracy of the multiplication in the weighting formulas can be limited by any predetermined number of resolution bits.

In the case of dual prediction, reference image indices are transmitted for each of the two predictors. To form two predictors, motion compensation is performed. To form two weighted predictors, each predictor uses a weight coefficient mapped to its reference image index. Then, two weighted predictors are averaged together to form an averaged predictor, which is then added to the encoded remainder.

For use for blocks in B-images that use the predictors of list 0 and list 1, a weighted predictor is formed as

Pred = (P0 * Pred0 + D0 + P1 * Pred1 + D1) / 2 (3)

To ensure that the resulting values are within an acceptable range of pixel values, typically from 0 to 255, clipping along the boundaries of the region can be applied to the weighted predictor or any of the intermediate values in the calculation of the weighted predictor.

Accordingly, a weighting factor is applied to the prediction of the reference image of the decoder and the video compression encoder that use several reference images. The weight coefficient is adapted for individual motion blocks within the image based on the index of the reference image that is used for this motion block. Since the index of the reference image is already transmitted in the bit stream of the compressed video signal, the amount of additional overhead service information for adapting the weight coefficient based on the motion block is significantly reduced. All motion blocks encoded with respect to the same reference image apply the same weight to the prediction of the reference image.

Those skilled in the art, based on the principles set forth herein, can readily discover these and other features and advantages of the present invention. It is understood that the principles of the present invention can be implemented in various types of hardware, software, firmware, specialized processors, or combinations thereof.

Most preferred is the implementation of the principles of the present invention as a combination of hardware and software. In addition, the software is preferably implemented as an application program physically implemented in a memory unit for storing programs. An application program may be downloaded to and executed on a computing device having any appropriate architecture. Preferably, the computing device is implemented on a computer platform having hardware such as one or more central processing units (CPUs), random access memory (RAM), and input / output (I / O) interfaces. A computer platform may also include micro-command code and an operating system. The various processes and functions described herein may be part of the micro-command code or part of an application program, or any combination of them that can be executed by the CPU. In addition to the computer platform, various other peripheral devices can be connected, such as an additional memory unit for storing data and a printing device.

It should also be understood that, since some of the components of the system and methods depicted in the attached drawings are preferably implemented in software, the actual relationships between system components or process function blocks may vary depending on how the present invention is programmed. In the presence of the principles set forth herein, one skilled in the art may envision these and similar implementations or configurations of the present invention.

Although illustrative embodiments have been described herein, according to the attached drawings, it is understood that the present invention is not limited to just these embodiments and that a person skilled in the art can make various changes without departing from the gist and without departing from the scope of the present invention and modifications. It is intended that all such changes and modifications be included within the scope of the present invention as defined by the appended claims.

Claims (8)

1. A video encoder for encoding video signal data for an image block and an index of a specific reference image, comprising a receiving module for receiving a substantially uncompressed image block, a weighting coefficient assigning unit for assigning a weighting factor for this image block corresponding to the first specific reference image having a corresponding index, calculation means for calculating motion vectors corresponding to a difference between said image block and a first specific reference image, a motion compensation module for performing motion compensation with respect to the first specific reference image in accordance with said motion vectors, a multiplication module for multiplying a motion-compensated first specific reference image by an assigned weight coefficient in order to generate a weighted motion-compensated first reference image, a second weight assignment module for assigning, for said image block, a weight corresponding to a second specific reference image having a second corresponding index, calculation means for calculating motion vectors corresponding to the difference between said image block and a second specific reference image, a motion compensation module for performing motion compensation with respect to the second specific reference image in accordance with said motion vectors, a multiplication module for multiplying a motion-compensated second specific reference image by an assigned weight coefficient in order to generate a weighted motion-compensated reference image, combining means for combining a weighted motion-compensated first reference image with a weighted motion-compensated second reference image to form a motion-compensated resulting reference image, a subtraction module for subtracting the motion-compensated resulting reference image from said substantially uncompressed image block, and an encoding module for encoding a signal indicating a difference between said substantially uncompressed image block and a motion-compensated resulting reference image, together with corresponding indices of said two specific reference images used to form a motion-compensated resulting reference image. 2. The video encoder according to claim 1, further comprising storages of reference images associated with the ability to exchange signals with said modules for assigning a weight coefficient of a reference image to provide reference images corresponding to the indices of said specific reference images. 3. The video encoder according to claim 1, used with dual predictive image predictors and further comprising prediction means for generating the first and second predictors from two different reference images. 4. The video encoder according to claim 3, in which both of said two different reference images correspond to one direction relative to said image block. 5. A method of encoding video data for an image block, comprising the steps of: receive a substantially uncompressed image block, assign a weighting factor for this image block corresponding to the first particular reference image having a corresponding index, calculating motion vectors corresponding to the difference between said image block and the first specific reference image, performing motion compensation with respect to the first specific reference image in accordance with said motion vectors, multiplying the motion-compensated first specific reference image by an assigned weight coefficient to form a weighted motion-compensated first reference image, a weighting factor corresponding to the second specific reference image having a corresponding index is assigned to said image block, calculating motion vectors corresponding to the difference between said image block and a second specific reference image, performing motion compensation with respect to the second specific reference image in accordance with said motion vectors, multiplying the motion-compensated second specific reference image by an assigned weight coefficient to form a weighted motion-compensated second reference image, combining a weighted motion-compensated first reference image with a weighted motion-compensated second reference image to form a motion-compensated resulting reference image, subtracting the motion-compensated resulting reference image from said substantially uncompressed image block, and encode a signal indicating the difference between said substantially uncompressed image block and a motion-compensated resulting reference image, together with corresponding indices of said two specific reference images used to form a motion-compensated resulting reference image. 6. The method according to claim 5, in which the calculation of the motion vectors includes the steps of checking inside the search area for any movement within a predetermined range of displacements relative to the image block, at least one of the sum of the absolute differences and the mean square errors of each pixel in said image block with a motion-compensated reference image, and the offset with the smallest sum of absolute differences or rms is selected by mistake as a motion vector. 7. The method according to claim 5, in which both of said two different reference images correspond to one direction relative to said image block. 8. The method according to claim 5, in which the calculation of the motion vectors includes the steps of checking inside the search area for any movement within a predetermined range of displacements relative to the image block, at least one of the sum of the absolute differences and the mean square the errors of each pixel in the said image block with the first motion-compensated reference image corresponding to the first predictor select the offset with the smallest sum of the absolute differences or the standard error as the motion vector for the first predictor, calculate at least one of the sum of the absolute differences and the standard error of each pixel in the image block with the second motion-compensated reference image corresponding to the second predictor, and select the offset with the smallest sum of absolute differences or standard error as the motion vector for the second predictor.

Images