Calculation method for predicted motion vector

BACKGROUND OF THE INVENTION

Field of the Invention

0001The present invention relates to a moving picture coding system, and more particularly to a prediction motion vector calculation method by defining a motion vector to be used in a median operation when a neighboring block around a block to be coded has a plurality of motion vectors , so as to obtain a prediction motion vector (PMV) of the block to be coded, using motion vector information of neighboring blocks, and improve a coding efficiency.

Description of the Related Art

0002Generally, in order to reduce the amount of bits to be used for the transfer of motion information, an encoder, instead of sending a motion vector MV directly to a decoder, selects a median value of motion vectors of three neighboring blocks through a median operation, determines the selected median value as a prediction motion vector PMV, obtains a difference MVD between the MV and the PMV (i.e., MVD = MV-PMV), and sends the obtained difference MVD to the decoder.

0003Then, the decoder obtains the motion vector MV by obtaining the prediction motion vector PMV in the same manner as the encoder and adding the sent MVD to the obtained PMV. In <figref idref="f0001">Fig. 1</figref>, a block E is a block to be coded (or decoded), and blocks A, B and C are neighboring blocks of the block E. Defining motion vectors of the neighboring blocks A, B and C, respectively, as MV_A, MV_B and MV_C, a prediction motion vector PMV of the block E can be obtained through a median operation as follows: <maths id="math0001" num=""><math display="block"><mi>PMV</mi><mo mathvariant="normal">=</mo><mi>median</mi><mfenced open="{" close="}" separators=""><msub><mi>MV</mi><mi mathvariant="normal">A</mi></msub><mo>⁢</mo><msub><mi>MV</mi><mi mathvariant="normal">B</mi></msub><mo>⁢</mo><msub><mi>MV</mi><mi mathvariant="normal">C</mi></msub></mfenced></math><img file="EP2003902A2_D0001.tif" /></maths>

0004A block D in <figref idref="f0001">Fig. 1</figref> is a block which is used instead of the block C when the block C exists outside of a picture. Provided that only one of the three blocks A, B and C, or A, B and D refers to the same reference picture as that referred to by the block E, a motion vector MV of that block will be used as the prediction motion vector PMV. This motion information sending method is applied to all pictures irrespective of their types.

0005On the other hand, a B picture has five types of predictive modes such as forward mode, backward mode, bi-predictive mode, direct mode and intra mode. Generally, a neighboring block in the forward mode has one motion vector MVFW obtained from a forward reference picture with an index ref_idx_fwd, and a neighboring block in the backward mode has one motion vector MVBW obtained from a backward reference picture with an index ref_idx_bwd.

0006In the bi-predictive mode of the B picture, the prediction is allowed from different directions and the same directions, such as forward/forward, backward/backward, and forward/backward. Each reference picture uses the index ref_idx_fwd or ref_idx_bwd regardless of its direction (forward or backward), and each motion vector is also represented as MVFW or MVBW regardless of its direction (The reason is that the predefined 'syntaxes' are used as they are. For expression of the syntaxes, 'ref_idx_10' or 'ref_idx_11' may be used for each index and 'mv_list0' or 'mv_list1' may be used for each motion vector.). The direct mode of the B picture is a predictive mode where motion information is not sent to the decoder and motion vectors MVf and MVb and reference pictures are derived from the inside of the decoder. The fact that the derived motion vectors are represented as MVf and MVb irrespective of their directions is the same as that in the bi-predictive mode.

0007In a conventional method for calculating a prediction motion vector PMV of the B picture, a forward prediction motion vector of the block E is obtained by extracting only forward motion vectors of the neighboring blocks and performing a median operation with respect to the extracted forward motion vectors. If one of the neighboring blocks has no forward motion vector, its motion vector is set to 0 and the median operation is performed under such a condition. This method is similarly applied to a backward prediction motion vector of the block E, so as to use only backward motion vectors of the neighboring blocks. If one of the neighboring blocks is in the intra mode, its motion vector is set to 0, the neighboring block is considered to refer to a reference picture different from that referred to by the block E, and the prediction motion vector PMV is obtained under such a condition.

0008However, as stated above, in the bi-predictive mode of the B picture, the prediction is allowed from different directions and the same directions, such as forward/forward, backward/backward, and forward/backward, each reference picture uses the index ref_idx_fwd or ref_idx_bwd regardless of its direction (forward or backward), and each motion vector is also represented as MVFW or MVBW regardless of its direction. As a result, there is a need to define a method for calculating a prediction motion vector PMV when a neighboring block having two motion vectors exists.

0009Provided that a neighboring block is in the bi-predictive mode (or the direct mode), motion vectors MVFW and MVBW (or MVf and MVb) thereof may have the same directions such as forward/forward or backward/backward, or different directions such as forward/backward. This direction information of the motion vectors cannot be determined from only the motion vector syntaxes 'MVFW' and 'MVBW' or the reference picture indexes 'ref_idx_fwd' and 'ref_idx_bwd'. The conventional method for calculating the PMV of the B picture gives no accurate description of such a problem, resulting in great confusion.

0010For example, in the case where a neighboring block is in the bi-predictive mode having two motion vectors in the forward/forward directions, the conventional PMV calculation method gives no clearly defined determination as to whether both or any one of the two motion vectors must be used for the calculation of the forward prediction motion vector PMV of the block E.

SUMMARY OF THE INVENTION

0011Therefore, the present invention has been made in view of the above problems, and it is an object of the present invention to provide a method for assigning direction information to reference pictures and a method for determining the directions of the reference pictures, wherein unique information enabling the acquisition of direction information of motion vectors is assigned to each reference picture, so that information regarding a direction from each neighboring block to each reference picture can be acquired.

0012It is another object of the present invention to provide a prediction motion vector calculation method by defining a motion vector to be used in a median operation when a neighboring block of a block to be coded has a plurality of motion vectors, so as to obtain a prediction motion vector (PMV) of the block to be coded, using motion vector information of neighboring blocks, and improve a coding efficiency.

0013In accordance with one aspect of the present invention, the above and other objects can be accomplished by assigning direction information to a reference picture as a feature of the reference picture, so as to give the direction information of the reference picture pointed by reference picture index.

0014Preferably, the direction information indicative of the display order of each reference picture may be represented by a picture order count (POC) value.

0015In accordance with another aspect of the present invention, there is provided a method for determining directions of reference pictures pointed to, respectively, by reference picture indexes, comprising the step of acquiring display order information of each reference picture, comparing the acquired display order information with display order information of a block to be currently coded, and determining a direction (forward or backward) of each reference picture against the block to be currently coded.

0016Preferably, the display order information of each reference picture may be acquired from a POC value.

0017In accordance with yet another aspect of the present invention, there is provided a method for calculating a prediction motion vector (PMV) of a block to be coded, by performing a median operation using motion vectors of neighboring blocks, comprising the steps of a), if the neighboring blocks have the motion vectors, acquiring direction information of reference pictures pointed by the motion vectors of the neighboring blocks; and b) selecting ones of the motion vectors of the neighboring blocks with reference to the acquired direction information and performing the median operation including the selected motion vectors to obtain the prediction motion vector of the block to be coded.

0018Preferably, the step a) may include the step of determining the direction information of the motion vectors by comparing display order information of the reference pictures pointed by the motion vectors of the neighboring blocks with display order information of the block to be coded.

0019Preferably, the step b) may include the step of, if one of the neighboring blocks has two motion vectors with different directions, selecting one of the two motion vectors having the same direction as that of the prediction motion vector and performing the median operation including the selected motion vector to obtain the prediction motion vector.

0020Alternatively, the step b) may include the step of, if one of the neighboring blocks has two motion vectors with the same directions, which are different from that of the prediction motion vector, setting the two motion vectors to 0, considering the neighboring block to refer to a reference picture different from that referred to by the block to be coded, and performing the median operation including the zero motion to obtain the prediction motion vector.

0021Alternatively, the step b) may include the step of b-1), if one of the neighboring blocks has two motion vectors MV1 and MV2 with the same directions, which are the same as that of the prediction motion vector, and both the two motion vectors MV1 and MV2 refer to the same reference picture, selecting one of the two motion vectors MV1 and MV2 and performing the median operation including the selected motion vector to obtain the prediction motion vector.

0022More preferably, the step b-1) may include the step of b-2) selecting one of the two motion vectors MV1 and MV2 to be earlier decoded or having the same mode (MV1 mode or MV2 mode) as that of the prediction motion vector, and performing the median operation including the selected motion vector to obtain the prediction motion vector. Here, the motion vector having the same mode signifies a motion vector having the same transcription as that indicative of the prediction motion vector.

0023Alternatively, the step b) may include the step of, if one of the neighboring blocks has two motion vectors MV1 and MV2 with the same directions, which are the same as that of the prediction motion vector, and only one of the motion vectors MV1 and MV2 refers to a reference picture referred to by the block to be coded, selecting one of the motion vectors MV1 and MV2 referring to the reference picture referred to by the block to be coded, and performing the median operation including the selected motion vector to obtain the prediction motion vector.

0024Alternatively, the step b) may include the step of, if one of the neighboring blocks has two motion vectors MV1 and MV2 with the same directions, which are the same as that of the prediction motion vector, neither of the motion vectors MV1 and MV2 refers to a reference picture referred to by the block to be coded and they refer to different reference pictures, selecting one of the motion vectors MV1 and MV2 referring to a reference picture closest to the reference picture referred to by the block to be coded, or a reference picture closest to a picture to be currently coded, and performing the median operation including the selected motion vector to obtain the prediction motion vector.

0025Alternatively, the step b) may include the step of, if one of the neighboring blocks has one motion vector with a direction different from that of the prediction motion vector, setting the motion vector of the neighboring block to 0, considering the neighboring block to refer to a reference picture different from that referred to by the block to be coded, and performing the median operation including the zero motion of the neighboring block to obtain the prediction motion vector.

0026Alternatively, the step b) may include the step of, if one of the neighboring blocks has one motion vector with the same direction as that of the prediction motion vector, performing the median operation including the motion vector of the neighboring block to obtain the prediction motion vector.

BRIEF DESCRIPTION OF THE DRAWINGS

0027The above and other objects, features and other advantages of the present invention will be more clearly understood from the following detailed description taken in conjunction with the accompanying drawings, in which: <ul id="ul0001" list-style="none" compact="compact"><li><figref idref="f0001">Fig. 1</figref> is a view illustrating the calculation of a prediction motion vector of a block E using motion vectors of neighboring blocks A, B and C.</li></ul>

DESCRIPTION OF THE PREFERRED EMBODIMENTS

0028If a neighboring block of a block to be coded is in a bi-predictive mode (or a direct mode), motion vectors MVFW and MVBW (or MVf and MVb) thereof may have the same directions such as forward/forward and backward/backward, or different directions such as forward/backward. This direction information of the motion vectors cannot be determined from only motion vector syntaxes 'MVFW' and 'MVBW' or reference picture indexes 'ref_idx_fwd' and 'ref_idx_bwd'. For this reason, there is a need to acquire the direction information by referring to different unique information held by reference pictures.

0029This invention proposes a method for acquiring direction information of motion vectors by comparing the display orders of reference pictures, and calculating a prediction motion vector PMV on the basis of the acquired direction information.

0030Now, preferred embodiments of the present invention will be described in detail with reference to the annexed drawing.

1. RECOGNITION OF MOTION VECTOR DIRECTIONS THROUGH DISPLAY ORDER COMPARISON

0031Direction information of motion vectors of neighboring blocks must be acquired before calculation of a prediction motion vector PMV of a block to be coded. Then, through the direction information of the motion vectors of the respective neighboring blocks, determinations are made as to whether the motion vectors of the neighboring blocks must be included in the median operation.

0032In general, if a neighboring block is in a forward mode or backward mode having one motion vector, the direction of that motion vector can be determined from a reference picture index. However, in the case where a neighboring block is in a bi-predictive mode or direct mode having two motion vectors, it is impossible to recognize actual directions of reference pictures pointed to by two reference picture indexes.

0033The reason is as follows. A decoder of a moving picture coding system which allows multiple reference pictures and a B picture to be used as references cannot estimate direction information only with reference picture indexes because it cannot accurately acquire the number of forward reference pictures and backward reference pictures of the B picture to be currently decoded, even though it can recognize the relation between a default forward/backward indexing order and a relative forward/backward indexing order from re-mapping information sent from an encoder.

0034In the present invention, it is proposed that a reference picture pointed to by a reference picture index will include unique information indicative of its display order for recognition of its direction. This display order is represented by a picture order count (POC) value. As a result, the direction of each motion vector can be easily recognized by comparing a display order of each reference picture pointed to by each reference picture index with a display order of the B picture to be currently coded.

2. MEDIAN OPERATION FOR CALCULATION OF PREDICTION MOTION VECTOR PMV WHEN NEIGHBORING BLOCK IS IN BI-PREDICTIVE MODE OR DIRECT MODE

0035If a neighboring block of a block to be coded is in the bi-predictive mode or direct mode, it has two motion vectors. Of these motion vectors, one having the same direction as that of a prediction motion vector PMV of the block to be coded will have to be used for the median operation. A detailed description will hereinafter be given of the efficient prediction motion vector PMV calculation method proposed by the present invention. For the convenience of description, assume that a block to be coded is E, neighboring blocks are A, B, C and D, and two motion vectors of each neighboring block are MV1 and MV2, as shown in <figref idref="f0001">Fig. 1</figref>.

2.1 CASE WHERE TWO MOTION VECTORS OF NEIGHBORING BLOCK HAVE DIFFERENT DIRECTIONS

0036A motion vector having the same direction as that of a prediction motion vector PMV of a block to be coded is selected and the prediction motion vector PMV is then calculated through the median operation.

0037In other words, a forward motion vector is selected for calculation of the PMV with respect to a forward prediction motion vector of the block E, and a backward motion vector is selected for calculation of the PMV with respect to a backward prediction motion vector of the block E. Then, the prediction motion vector PMV of each direction is obtained through the median operation.

2.2 CASE WHERE TWO MOTION VECTORS OF NEIGHBORING BLOCK HAVE THE SAME DIRECTIONS, WHICH ARE THE SAME AS THAT OF PREDICTION MOTION VECTOR PMV

0038First, a determination is made as to whether two motion vectors MV1 and MV2 of a neighboring block refer to a reference picture referred to by the block E to be coded.

0039If both the motion vectors MV1 and MV2 refer to the same reference picture, one of them (for example, a motion vector to be earlier decoded, or a motion vector having the same mode (MV1 mode or MV2 mode) as that of a prediction motion vector) is selected and included in the median operation for calculation of the prediction motion vector PMV. Here, the motion vector having the same mode signifies a motion vector having the same transcription as that indicative of the prediction motion vector. Also, the reference picture referred to by the motion vectors MV1 and MV2 may be the same as or different from the reference picture referred to by the block E to be coded.

0040If only one of the motion vectors MV1 and MV2 refers to the reference picture referred to by the block E, it is included in the median operation for calculation of the prediction motion vector PMV.

0041If neither of the motion vectors MV1 and MV2 refers to the reference picture of the block E and they refer to different reference pictures, one thereof referring to a reference picture closest to the reference picture referred to by the block E, or a reference picture closest to a picture to be currently coded is selected and included in the median operation for calculation of the prediction motion vector PMV.

2.3 CASE WHERE TWO MOTION VECTORS OF NEIGHBORING BLOCK HAVE THE SAME DIRECTIONS, WHICH ARE DIFFERENT FROM THAT OF PREDICTION MOTION VECTOR PMV

0042Two motion vectors MV1 and MV2 of a neighboring block are set to 0, the neighboring block is considered to refer to a reference picture different from the reference picture referred to by the block E, and the prediction motion vector PMV of the block to be coded is obtained through the median operation including the zero motion.

0043In this manner, when a neighboring block has two motion vectors, the directions of the motion vectors are recognized from display orders of associated reference pictures and determinations are made as to whether they are the same as the direction of the prediction motion vector PMV, thereby making it possible to obtain a PMV more approximating an MV. This results in a reduction in the magnitude of a motion vector difference MVD (= MV-PMV) to be sent to the decoder and, in turn, a reduction in the amount of bits to be sent to the decoder. Accordingly, the entire coding efficiency can be raised.

0044On the other hand, in the case where a neighboring block has one motion vector, direction information of the motion vector is acquired from a display order of an associated reference picture. If the direction of the motion vector is not the same as that of the prediction motion vector, the motion vector is set to 0, the neighboring block is considered to refer to a reference picture different from the reference picture referred to by the block to be coded, and the prediction motion vector is obtained through the median operation including zero motion.

0045Further, in the case where a neighboring block has one motion vector, direction information of the motion vector is acquired from a display order of an associated reference picture. If the direction of the motion vector is the same as that of the prediction motion vector, the motion vector is included in the median operation for calculation of the prediction motion vector.

0046As apparent from the above description, the present invention provides a method for assigning direction information to reference pictures and a method for determining the directions of the reference pictures, wherein display order information enabling the acquisition of direction information of motion vectors is assigned to each reference picture. Therefore, information regarding a direction from a block to be currently coded to each reference picture can be acquired.

0047Further, the present invention provides a prediction motion vector calculation method by defining a motion vector to be used in a median operation when a neighboring block around a block to be coded has two motion vectors due toa bi-predictive mode or direct mode of a B picture. As a result, a prediction motion vector (PMV) of the block to be coded can be predicted using motion vector information of neighboring blocks, and a coding efficiency can be improved.

0048Although the preferred embodiments of the present invention have been disclosed for illustrative purposes, those skilled in the art will appreciate that various modifications, additions and substitutions are possible, without departing from the scope and spirit of the invention as disclosed in the accompanying claims. Summarized the present invention provides a method for calculating a prediction motion vector of a block to be coded, by performing a median operation using motion vectors of neighboring blocks. Herein, if one of the neighboring blocks has motion vectors MV1 and MV2, direction (forward or backward) information of reference pictures referred to, respectively, by the motion vectors are acquired and one of the motion vectors is selected with reference to the acquired direction information. Then, the median operation including the selected motion vector is performed to obtain the prediction motion vector of the block to be coded.