KR100494831B1 - Method of determining motion vector - Google Patents

Abstract

In the direct mode motion vector calculation method of a B picture according to the present invention, in extracting a direct mode motion vector of a B (bi-predictive) picture in a video coding system, a list 1 reference picture of a direct mode of a B picture (list 1 reference) When the picture for direct mode is located before or after the B picture in time, the mode of the motion vector, which is contained in the co-located block in the list 1 reference picture of the direct mode. The direct motion vector of the B picture is calculated by deriving the list 0 motion vector MV _F and the list 1 motion vector MV _B by scaling the determined motion vector regardless of 0 and / or list 1).

In addition, the direct mode motion vector calculation method of the B picture according to the present invention includes a list 0 reference picture in direct mode that the reference picture indicated by the motion vector of the block having the same position in the list 1 reference picture. mode) to calculate the motion vector of the direct mode of the B picture.

Description

METHOD OF DETERMINING MOTION VECTOR}

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a video coding method, and more particularly, to a method of determining a motion vector in a bi-predictive (B) picture defined in a next generation video compression technique.

Conventional B pictures have five prediction modes, such as forward, backward, bi-directional, direct, and intra modes. Here, the forward, backward, and bidirectional modes contain direction information in the mode name, so that the direction of the motion vector can be determined by looking at the mode name. In particular, the direct mode derives two bidirectional motion vectors from a motion of a block in the same position in a neighboring picture by using a temporal redundancy characteristic that motion continuity between two adjacent pictures is kept constant. Since the direct mode does not transmit motion information to the decoder, there is an advantage that the bit rate can be reduced.

On the other hand, the B picture in a next-generation video compression technology such as H.264 or MPEG-4 part 10 allows the B picture to be used as a reference picture, so that the B picture may be stored in a reference picture buffer. In addition, the prediction mode of the B picture includes list 0, list 1, bi-predictive, direct, and intra modes.

Here, the List 0 mode is a mode similar to the conventional forward mode, in which the reference picture index and the motion vector difference, which are motion information, are displayed in 'ref_idx_fwdl0' and 'mvd_l0', respectively. List 1 mode is similar to the conventional backward mode, and motion information is displayed in ref_idx_l1 and mvd_l1, respectively. The bi-prediction mode has two reference pictures, and the positions of the two reference pictures may be both ahead of time or behind each of the B pictures, or may exist back and forth around the B picture. At this time, the difference between the two reference picture indexes and the motion vector is indicated in ref_idx_l0, ref_idx_l1, mvd_l0, and mvd_l1, and each reference picture has data called POC (picture order count), which is temporal position information.

The direct mode selects one of a spatial technique and a temporal technique to derive the motion vector. The spatial technique is a technique of deriving the reference picture index and the motion vector of list 0 and list 1 from the neighboring blocks of the macroblock to be coded. On the other hand, the temporal technique is a method used in the conventional B picture, and scales the list 0 motion vector of a block having the same position in the list 1 reference picture for direct mode in direct mode. MV _F ) and list 1 motion vector (MV _B ) are derived. Here, the list 1 reference picture in the direct mode is a picture having an index of 0 for list 1 prediction, and the list 0 reference picture in the direct mode uses a list 0 reference picture of a block having the same position in the list 1 reference picture.

1 shows a default index for list 0 prediction and a default index for list 1 prediction for each B picture in an IBBBP pattern when the number of available list 0 and list 1 reference pictures (or short-term buffer size) is 6 And list 1 reference pictures in direct mode, respectively. Here, the default index for list 0 prediction and the default index for list 1 prediction are indexed by the output order, that is, the POC value, of the decoded reference picture regardless of the decoding order. In Fig. 1, a P picture followed by all B pictures in time is used as a list 1 reference picture in direct mode.

Meanwhile, FIG. 2 shows a default index for list 0 prediction, a default index for list 1 prediction, and a list 1 reference picture in direct mode for each B picture in an IBBB pattern using only B pictures, respectively. 2 (a) shows that when the B picture to be coded is B8, the temporally preceding B5 having index 0 of list 1 becomes the list 1 reference picture in direct mode, and as shown in (b), the next decoding order is B7. The list 1 reference picture in direct mode for is B8 located later in time. Finally, as shown in (c), the list 1 reference picture of the direct mode for the next decoding order B9 becomes B7 located temporally.

1 and 2, the list 1 reference picture in the direct mode may be a P or B picture located later in time than the B picture to be coded, or may be a B picture located in front of the time. have.

FIG. 3 illustrates a mode that a co-located block may have when the list 1 reference picture of the direct mode is located later in time than the B picture. In this case, since the list 1 reference picture may be a P picture or a B picture, a block at the same position of the list 1 reference picture may have one or two motion vectors or have an intra mode. Next-generation video compression techniques, such as H.264 or MPEG-4 part 10, allow reordering the reference picture index at the slice level so that index 0 for list 1 prediction can be assigned to the picture immediately after the B picture. have. That is, since the list 1 reference picture may exist immediately after the B picture, the motion vector of the block having the same position may be directed forward or backward.

4 illustrates a mode that a block having the same position may have when the list 1 reference picture of the direct mode is located ahead of the B picture in time. As before, a block of the same position has one or two motion vectors or has an intra mode. In this case, since another reference picture may exist between the list 1 reference picture and the B picture, the direction of the motion vector may be forward and backward in time.

As shown in FIG. 3 and FIG. 4, since the list 1 reference picture of the direct mode may have various prediction modes, there is a need to search for a motion vector calculation method of the direct mode in consideration of such various cases.

The present invention proposes a direct mode motion vector extraction technique in B (Bi-predictive) pictures defined in the next-generation video compression technology, thereby increasing the likelihood that the direct mode is selected as the prediction mode of the macroblock. An object of the present invention is to provide a method for determining a direct mode motion vector of a B picture that can improve efficiency.

In order to achieve the above object, the direct mode motion vector calculation method of the B picture according to the present invention derives the motion vector of the direct mode when the block having the same position in the list 1 reference picture has the list 0 and list 1 motion vectors. The motion vector to be used is characterized by selecting list 0 or list 1 motion vectors by two methods.

According to the present invention, a motion vector having a close temporal distance from a list 1 reference picture having a block having the same position among list 0 and list 1 motion vectors is selected as a motion vector for inducing a motion vector in a direct mode. If is pointing to the same reference, there is a method of selecting list 0 motion vector as a motion vector for inducing a direct motion vector, and another method is to unconditionally derive a list 0 motion vector for direct motion vector regardless of temporal distance. The feature is that it is selected as a vector.

In addition, according to the present invention, the conventional technique that the motion vector of the direct mode is derived from the list 0 motion vector of the block in the same position is applied to the case where the block at the same position of the list 1 reference picture has only the list 1 motion vector. In this case, the list 0 motion vector of the block in the same position is 0, so that the motion vectors in the direct mode are all 0. In order to solve this problem, the present invention is directed to the block regardless of the mode type of the block in the same position. The feature is that a motion vector in direct mode is derived using this motion vector.

Herein, the present invention provides a method for obtaining a direct motion vector when a block having the same position in the list 1 reference picture has only the list 1 motion vector. In the first method, the motion vector value to be used for calculating the direct mode motion vector is 0, and the list 0 reference picture for direct mode of the direct mode is defined as a decoded picture existing at the position immediately before the B picture in time. Characterized in that. In the second method, the motion vector value used for the direct mode motion vector calculation uses the list 1 motion vector of the same block as it is, and instead, the list 0 reference picture in the direct mode is decoded in the position immediately before the B picture in time. Its characteristics are that it is defined as a completed picture. In the third method, the motion vector value used for the direct mode motion vector calculation is to use the list 1 motion vector of the block in the same position as it is, and determine the reference picture pointed to by the list 1 motion vector as the list 0 reference picture in the direct mode. It is characterized by.

Also, according to the present invention, when the list 1 reference picture in the direct mode is defined as a picture having an index of 0 used in the list 1 prediction according to the conventional technique, an index 0 when another picture is decoded between the B picture and the picture having the index 0 is decoded. Although additional memory usage is essential because the motion information and reference picture information of the in-picture must be maintained, the present invention has a feature that additional memory usage can be saved by defining list 1 reference picture in direct mode as the most recently decoded picture. .

According to the present invention, when extracting a direct mode motion vector of a B (bi-predictive) picture in a video coding system, when the list 1 reference picture of the direct mode of the B picture is located ahead of the B picture in time, By scaling the motion vectors of the co-located macroblocks in the direct mode list 1 reference picture, the list 0 motion vector (MV _F ) and the list 1 motion vector (MV _B ) are derived. The characteristic is that the direct mode motion vector of the picture is calculated.

According to the present invention, the macroblock of the B picture and the macroblock at the same position of the list 1 reference picture are both in frame mode, and the list 0 reference picture in the direct mode of the B picture is smaller than the list 1 reference picture. When present in front of time, the motion vectors MV _F and MV _B in the direct mode of the B picture are characterized in that they are calculated from the following equation.

MV _F = TD _B x MV / TD _D

MV _B = (TD _B -TD _D ) x MV / TD _D

or

Z = TD _B x 256 / TD _D MV _F = (Z x MV + 128) >> 8

W = Z-256 MV _B = (W x MV + 128) >> 8

Where TD _B is the temporal distance between the current B frame and the list 0 reference frame, TD _D is the temporal distance between the list 1 reference list and the list 0 reference frame, and MV is the same as in the list 1 reference picture in the direct mode. The motion vector of the block of position.

In addition, according to the present invention, both the macroblock of the B picture and the macroblock at the same position of the list 1 reference picture are all in frame mode, and the list 0 reference picture in the direct mode of the B picture is more temporally than the list 1 reference picture. When present later, the motion vectors MV _F and MV _B in the direct mode of the B picture are characterized in that they are calculated from the following equation.

MV _F =-TD _B x MV / TD _D

MV _B =-(TD _B + TD _D ) x MV / TD _D

or

Z =-TD _B x 256 / TD _D MV _F = (Z x MV + 128) >> 8

W = Z-256 MV _B = (W x MV + 128) >> 8

Where TD _B is the temporal distance between the current B frame and the list 0 reference frame, TD _D is the temporal distance between the list 1 reference list and the list 0 reference frame, and MV is the same as in the list 1 reference picture in the direct mode. The motion vector of the block of position.

According to the present invention, the macroblock of the B picture and the macroblock at the same position of the list 1 reference picture are both in the field mode, and the list 0 reference picture in the direct mode of the B picture is the list 1 reference picture. When present earlier in time, the motion vectors MV _{F, i} and MV _{B, i} in the direct mode for each of the frames i of the B frame are characterized in that they are calculated from the following equation.

MV _{F, i} = TD _{B, i} x MV _i / TD _{D, i}

MV _{B, i} = (TD _{B, i} -TD _{D, i} ) x MV _i / TD _{D, i}

or

Z = TD _{B, i} x 256 / TD _{D, i} MV _{F, i} = (Z x MV _i + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV _i + 128) >> 8

Here, TD _{B, i} is the temporal distance between the current B feed and the list 0 reference feed, TD _{D, i} is the temporal distance between the list 1 reference list and the list 0 reference feed, and MV _i is the direct The motion vector of the same block in the mode's list 1 reference field.

According to the present invention, the macroblock of the B picture and the macroblock at the same position of the list 1 reference picture are both in the field mode, and the list 0 reference picture in the direct mode of the B picture is the list 1 reference picture. When present later in time, the motion vectors MV _{F, i} and MV _{B, i} in the direct mode for each of the frames i of the B frame are characterized in that they are calculated from the following equation.

MV _{F, i} =-TD _{B, i} x MV _i / TD _{D, i}

MV _{B, i} =-(TD _{B, i} + TD _{D, i} ) x MV _i / TD _{D, i}

or

Z = -TD _{B, i} x 256 / TD _{D, i} MV _{F, i} = (Z x MV _i + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV _i + 128) >> 8

Here, TD _{B, i} is the temporal distance between the current B feed and the list 0 reference feed, TD _{D, i} is the temporal distance between the list 1 reference list and the list 0 reference feed, and MV _i is the direct The motion vector of the same block in the mode's list 1 reference field.

Further, according to the present invention, the macroblock of the B picture is in the shield mode and the macroblock at the same position of the list 1 reference picture is the frame mode, and the list 0 reference picture in the direct mode of the B picture is less than the list 1 reference picture. When present in time, the motion vectors MV _{F, i} and MV _{B, i} in the direct mode for each of the frames i of the B frame are characterized in that they are calculated from the following equation.

MV _{F, i} = TD _{B, i} x MV / TD _D

MV _{B, i} = (TD _{B, i} -TD _D ) x MV / TD _D

or

Z = TD _{B, i} x 256 / TD _D MV _{F, i} = (Z x MV + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV + 128) >> 8

Here, TD _{B, i} is the temporal distance between the current B feed and the list 0 reference feed, TD _D is the temporal distance between the list 1 reference list and the list 0 reference frame, and MV is the list 1 reference of the direct mode. This is the motion vector of the same block in the frame.

Further, according to the present invention, the macroblock of the B picture is in the shield mode and the macroblock at the same position of the list 1 reference picture is the frame mode, and the list 0 reference picture in the direct mode of the B picture is less than the list 1 reference picture. When present later in time, the motion vectors MV _{F, i} and MV _{B, i} in the direct mode for each of the frames i of the B frame are characterized in that they are calculated from the following equation.

MV _{F, i} =-TD _{B, i} x MV / TD _D

MV _{B, i} =-(TD _{B, i} + TD _D ) x MV / TD _D

or

Z =-TD _{B, i} x 256 / TD _D MV _{F, i} = (Z x MV + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV + 128) >> 8

Here, TD _{B, i} is the temporal distance between the current B feed and the list 0 reference feed, TD _D is the temporal distance between the list 1 reference list and the list 0 reference frame, and MV is the list 1 reference of the direct mode. This is the motion vector of the same block in the frame.

According to the present invention, the macroblock of the B picture is a frame mode and the macroblock at the same position of the list 1 reference picture is a shield mode, and the list 0 reference picture of the direct mode of the B picture is less than the list 1 reference picture. When present in front of time, the motion vectors MV _F and MV _B in the direct mode of the B frame are characterized in that they are calculated from the following equation. Here, the motion information of the block of the same position in the feed 1 of the list 1 reference frame is used to obtain the motion vector of the direct mode.

MV _F = TD _B x MV ₁ / TD _{D, 1}

MV _B = (TD _B -TD _{D, 1} ) x MV ₁ / TD _{D, 1}

or

Z = TD _B x 256 / TD _{D, 1} MV _F = (Z x MV ₁ + 128) >> 8

W = Z-256 MV _B = (W x MV ₁ + 128) >> 8

Here, TD _B is a temporal distance between a current B frame and a list 0 reference frame, TD _{D, 1} is a temporal distance between a feed ₁ of the list 1 reference frame and a list 0 reference feed, and MV ₁ is the list 1. This is the motion vector of the same position block in feed 1 of the reference frame.

According to the present invention, the macroblock of the B picture is a frame mode and the macroblock at the same position of the list 1 reference picture is a shield mode, and the list 0 reference picture of the direct mode of the B picture is less than the list 1 reference picture. When present later in time, the motion vectors MV _F and MV _B in the direct mode of the B frame are calculated from the following equation. Here, the motion information of the block of the same position in the feed 1 of the list 1 reference frame is used to obtain the motion vector of the direct mode.

MV _F =-TD _B x MV ₁ / TD _{D, 1}

MV _B =-(TD _B + TD _{D, 1} ) x MV ₁ / TD _{D, 1}

or

Z =-TD _B x 256 / TD _{D, 1} MV _F = (Z x MV ₁ + 128) >> 8

W = Z-256 MV _B = (W x MV ₁ + 128) >> 8

Here, TD _B is a temporal distance between a current B frame and a list 0 reference frame, TD _{D, 1} is a temporal distance between a feed ₁ of the list 1 reference frame and a list 0 reference feed, and MV ₁ is the list 1. This is the motion vector of the same position block in feed 1 of the reference frame.

According to the present invention, when the macroblock at the same position of the list 1 reference picture in the direct mode of the B picture has two motion vectors, one motion vector (list 0) is used. Or list 1 motion vector) and derives the motion vector of the direct mode of the B picture from the selected motion vector.

In addition, in order to achieve the above object, another embodiment of the direct mode motion vector calculation method of the B picture according to the present invention,

In extracting the direct mode motion vector of the B (bi-predictive) picture in the video coding system, when both the list 0 reference picture and the list 1 reference picture of the direct mode of the B picture are present in time later than the B picture, Scales the motion vectors of the co-located blocks in the list 1 reference picture in direct mode, derives the list 0 motion vector (MV _F ) and the list 1 motion vector (MV _B ) Its characteristic is that it computes the direct mode motion vector of.

According to the present invention, the macroblock of the B picture and the macroblock at the same position of the list 1 reference picture are both in the frame mode, and the list 0 reference picture of the B picture is present later in time than the list 1 reference picture. In this case, the motion vectors MV _F and MV _B in the direct mode of the B picture are calculated from the following equation.

MV _F = TD _B x MV / TD _D

MV _B = (TD _B -TD _D ) x MV / TD _D

or

Z = TD _B x 256 / TD _D MV _F = (Z x MV + 128) >> 8

W = Z-256 MV _B = (W x MV + 128) >> 8

Where TD _B is the temporal distance between the current B frame and the list 0 reference frame, TD _D is the temporal distance between the list 1 reference list and the list 0 reference frame, and MV is the same as in the list 1 reference picture in the direct mode. The motion vector of the block of position.

In addition, according to the present invention, both the macroblock of the B picture and the macroblock at the same position of the list 1 reference picture are all in frame mode, and the list 0 reference picture of the B picture is present in time compared to the list 1 reference picture. In this case, the motion vectors MV _F and MV _B in the direct mode of the B picture are calculated from the following equation.

MV _F =-TD _B x MV / TD _D

MV _B =-(TD _B + TD _D ) x MV / TD _D

or

Z =-TD _B x 256 / TD _D MV _F = (Z x MV + 128) >> 8

W = Z-256 MV _B = (W x MV + 128) >> 8

Where TD _B is the temporal distance between the current B frame and the list 0 reference frame, TD _D is the temporal distance between the list 1 reference list and the list 0 reference frame, and MV is the same as in the list 1 reference picture in the direct mode. The motion vector of the block of position.

According to the present invention, both the macroblock of the B picture and the macroblock at the same position of the list 1 reference picture are all in the field mode, and the list 0 reference picture of the B picture is temporally compared to the list 1 reference picture. In this case, the motion vectors MV _{F, i} and MV _{B, i} in the direct mode for each of the frames i of the B frame are calculated from the following equation.

MV _{F, i} = TD _{B, i} x MV _i / TD _{D, i}

MV _{B, i} = (TD _{B, i} -TD _{D, i} ) x MV _i / TD _{D, i}

or

Z = TD _{B, i} x 256 / TD _{D, i} MV _{F, i} = (Z x MV _i + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV _i + 128) >> 8

Here, TD _{B, i} is the temporal distance between the current B feed and the list 0 reference feed, TD _{D, i} is the temporal distance between the list 1 reference list and the list 0 reference feed, and MV _i is the direct The motion vector of the same block in the mode's list 1 reference field.

According to the present invention, both the macroblock of the B picture and the macroblock at the same position of the list 1 reference picture are all in the field mode, and the list 0 reference picture of the B picture is temporally compared to the list 1 reference picture. In this case, the motion vectors MV _{F, i} and MV _{B, i} in the direct mode for each of the frames i of the B frame are calculated from the following equation.

MV _{F, i} =-TD _{B, i} x MV _i / TD _{D, i}

MV _{B, i} =-(TD _{B, i} + TD _{D, i} ) x MV _i / TD _{D, i}

or

Z = -TD _{B, i} x 256 / TD _{D, i} MV _{F, i} = (Z x MV _i + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV _i + 128) >> 8

Here, TD _{B, i} is the temporal distance between the current B feed and the list 0 reference feed, TD _{D, i} is the temporal distance between the list 1 reference list and the list 0 reference feed, and MV _i is the direct The motion vector of the same block in the mode's list 1 reference field.

Further, according to the present invention, the macroblock of the B picture is in the shield mode, the macroblock at the same position of the list 1 reference picture is in the frame mode, and the list 0 reference picture of the B picture is temporally compared to the list 1 reference picture. When present later, the motion vectors MV _{F, i} and MV _{B, i} in the direct mode for each of the frames i of the B frame are characterized in that they are calculated from the following equation.

MV _{F, i} = TD _{B, i} x MV / TD _D

MV _{B, i} = (TD _{B, i} -TD _D ) x MV / TD _D

or

Z = TD _{B, i} x 256 / TD _D MV _{F, i} = (Z x MV + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV + 128) >> 8

Here, TD _{B, i} is the temporal distance between the current B feed and the list 0 reference feed, TD _D is the temporal distance between the list 1 reference list and the list 0 reference frame, and MV is the list 1 reference of the direct mode. This is the motion vector of the same block in the frame.

Further, according to the present invention, the macroblock of the B picture is in the shield mode, the macroblock at the same position of the list 1 reference picture is in the frame mode, and the list 0 reference picture of the B picture is temporally compared to the list 1 reference picture. When present before, the motion vectors MV _{F, i} and MV _{B, i} in the direct mode for each of the frames i of the B frame are characterized in that they are calculated from the following equation.

MV _{F, i} =-TD _{B, i} x MV / TD _D

MV _{B, i} =-(TD _{B, i} + TD _D ) x MV / TD _D

or

Z =-TD _{B, i} x 256 / TD _D MV _{F, i} = (Z x MV + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV + 128) >> 8

Here, TD _{B, i} is the temporal distance between the current B feed and the list 0 reference feed, TD _D is the temporal distance between the list 1 reference list and the list 0 reference frame, and MV is the list 1 reference of the direct mode. This is the motion vector of the macroblock at the same position in the frame.

According to the present invention, the macroblock of the B picture is a frame mode, the macroblock at the same position of the list 1 reference picture is a shield mode, and the list 0 reference picture of the B picture is temporally compared to the list 1 reference picture. When present later, the motion vectors MV _F and MV _B in the direct mode of the B frame are calculated from the following equation. Here, the motion information of the same block located in the 0 of the list 1 reference frame is used to obtain the motion vector of the direct mode.

MV _F = TD _B x MV ₀ / TD _{D, 1}

MV _B = (TD _B -TD _{D, 0} ) x MV ₀ / TD _{D, 0}

or

Z = TD _B x 256 / TD _{D, 0} MV _F = (Z x MV ₀ + 128) >> 8

W = Z-256 MV _B = (W x MV ₀ + 128) >> 8

Here, TD _B is the temporal distance between the current B frame and the list 0 reference frame, TD _{D, 1} is the temporal distance between the 0 and list 0 reference frames of the list 1 reference frame, and MV ₀ is the direct mode. This is the motion vector of the block at the same position in feed 0 of the list 1 reference frame.

According to the present invention, the macroblock of the B picture is a frame mode, the macroblock at the same position of the list 1 reference picture is a shield mode, and the list 0 reference picture of the B picture is temporally compared to the list 1 reference picture. When present before, the motion vectors MV _F and MV _B in the direct mode of the B frame are characterized in that they are calculated from the following equation. Here, the motion information of the same block located in the 0 of the list 1 reference frame is used to obtain the motion vector of the direct mode.

MV _F =-TD _B x MV ₀ / TD _{D, 0}

MV _B =-(TD _B + TD _{D, 0} ) x MV ₀ / TD _{D, 0}

or

Z =-TD _B x 256 / TD _{D, 0} MV _F = (Z x MV ₀ + 128) >> 8

W = Z-256 MV _B = (W x MV ₀ + 128) >> 8

Here, TD _B is the temporal distance between the current B frame and the list 0 reference frame, TD _{D, 1} is the temporal distance between the 0 and list 0 reference frames of the list 1 reference frame, and MV ₀ is the direct mode. list 1 This is the motion vector of the same position block in feed 0 of the reference frame.

According to the present invention, when a block at the same position of the list 1 reference picture in the direct mode of the B picture has two motion vectors, one motion vector MV ₁ or MV ₂ is selected, and the selected motion is selected. Its characteristic is that it derives the direct motion vector of the B picture from the vector.

In addition, to achieve the above object another embodiment of a direct mode motion vector calculation method of a B picture according to the present invention,

In extracting the direct mode motion vector of a B (bi-predictive) picture in a video coding system, the B picture is represented by representing the temporal distance between the pictures as a signed value to simplify the equation used to obtain the direct mode motion vector. Regardless of the position of the list 0 reference picture and the list 1 reference picture in direct mode, the motion vector contained in the co-located block of the same position in the list 1 reference picture in the direct mode is scaled. The characteristic is that the direct mode motion vector of the B picture is calculated by deriving the vector MV _F and the list 1 motion vector MV _B.

According to the present invention, when the macroblock of the B picture and the macroblock at the same position of the list 1 reference picture are both in the frame mode, the motion vectors MV _F and MV _B in the direct mode of the B picture are calculated from the following equation. It is characterized by the point.

MV _F = TD _B x MV / TD _D

MV _B = (TD _B -TD _D ) x MV / TD _D

or

Z = TD _B x 256 / TD _D MV _F = (Z x MV + 128) >> 8

W = Z-256 MV _B = (W x MV + 128) >> 8

Here, TD _B is the temporal distance between the current B frame and the list 0 reference frame. The temporal distance calculated from the B frame is represented by a positive sign, and the temporal distance calculated from the list 0 reference frame is represented by a negative sign. TD _D is the temporal distance between the list 1 reference frame and the list 0 reference frame. The temporal distance calculated from the list 1 reference frame is indicated by a positive sign and the temporal distance calculated from the list 0 reference frame is negative. The MV is a motion vector of a block of the same position in the list 1 reference picture of the direct mode.

According to the present invention, when the macroblocks of the B picture and the macroblocks at the same position of the list 1 reference picture are both in the field mode, the motion vector in the direct mode for each feed i of the B frame is in the field mode. MV _{F, i} and MV _{B, i} are characterized by being calculated from the following equation.

MV _{F, i} = TD _{B, i} x MV _i / TD _{D, i}

MV _{B, i} = (TD _{B, i} -TD _{D, i} ) x MV _i / TD _{D, i}

or

Z = TD _{B, i} x 256 / TD _{D, i} MV _{F, i} = (Z x MV _i + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV _i + 128) >> 8

Where TD _{B, i} is the temporal distance between the current B and list 0 reference feeds. The time distance calculated from the B feeds is represented by a positive sign and the time distance calculated from the list 0 reference feed is negative. TD _{D, i} is the temporal distance between list 1 reference list and list 0 reference feed, and the time distance calculated from list 1 reference feed is represented by a positive sign and list The temporal distance calculated from the zero reference period is represented by a negative sign, and MV _i is a motion vector of the same position block in the list 1 reference shield of the direct mode.

Further, according to the present invention, when the macroblock of the B picture is in the shield mode and the macroblock at the same position of the list 1 reference picture is in the frame mode, the motion vector MV in the direct mode for each of the feeds i of the B frame is provided. _{F, i} , MV _{B, i} are characterized by being calculated from the following equation.

MV _{F, i} = TD _{B, i} x MV / TD _D

MV _{B, i} = (TD _{B, i} -TD _D ) x MV / TD _D

or

Z = TD _{B, i} x 256 / TD _D MV _{F, i} = (Z x MV + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV + 128) >> 8

Where TD _{B, i} is the temporal distance between the current B and list 0 reference feeds. The time distance calculated from the B feeds is represented by a positive sign and the time distance calculated from the list 0 reference feed is negative. TD _D is the temporal distance between the list 1 reference frame and the list 0 reference frame. The temporal distance calculated from the list 1 reference frame is represented by a positive sign and is calculated from the list 0 reference frame. The temporal distance is indicated by a negative sign, and MV is a motion vector of a block in the same position in the list 1 reference frame of the direct mode.

Also, according to the present invention, when the macroblock of the B picture is a frame mode and the macroblock at the same position of the list 1 reference picture is a shield mode, and the list 1 reference picture is present later in time than the B picture, The motion vectors MV _F and MV _B in the direct mode of frame _B are calculated from the following equation used to obtain the motion vector of the direct mode in the macroblock at the same position in feed 0 of the list 1 reference frame. It has its features.

MV _F = TD _B x MV ₀ / TD _{D, 0}

MV _B = (TD _B -TD _{D, 0} ) x MV ₀ / TD _{D, 0}

or

Z = TD _B x 256 / TD _{D, 0} MV _F = (Z x MV ₀ + 128) >> 8

W = Z-256 MV _B = (W x MV ₀ + 128) >> 8

Here, TD _B is the temporal distance between the current B frame and the list 0 reference frame. The temporal distance calculated from the B frame is represented by a positive sign, and the temporal distance calculated from the list 0 reference frame is represented by a negative sign. TD _{D, 0} is the temporal distance between list 0 and list 0 reference frames of the list 1 reference frame, and the calculated time distance from feed 0 of the list 1 reference frame is represented by a positive sign and The temporal distance calculated from the zero reference period is represented by a negative sign, and MV ₀ is a motion vector of a block at the same position in the zero of the list 1 reference frame in the direct mode.

According to the present invention, when the macroblock of the B picture is the frame mode and the macroblock at the same position of the list 1 reference picture is the shield mode, and the list 1 reference picture is present in time before the B picture, The motion vectors MV _F and MV _B in the direct mode of frame _B are obtained from the point that the motion information of the block at the same position in feed 1 of the list 1 reference frame is calculated from the following equation used to obtain the motion vector of the direct mode. There is a characteristic.

MV _F = TD _B x MV ₁ / TD _{D, 1}

MV _B = (TD _B -TD _{D, 1} ) x MV ₁ / TD _{D, 1}

or

Z = TD _B x 256 / TD _{D, 1} MV _F = (Z x MV ₁ + 128) >> 8

W = Z-256 MV _B = (W x MV ₁ + 128) >> 8

Here, TD _B is the temporal distance between the current B frame and the list 0 reference frame. The temporal distance calculated from the B frame is represented by a positive sign, and the temporal distance calculated from the list 0 reference frame is represented by a negative sign. TD _{D, 1} is the temporal distance between feed ₁ of the list 1 reference frame and the list 0 reference feed, and the calculated time distance from feed 1 of the list 1 reference frame is represented by a positive sign and The temporal distance calculated from the zero reference period is represented by a negative sign, and MV ₁ is a motion vector of the same position block in the first position of the list 1 reference frame of the direct mode.

In addition, to achieve the above object another embodiment of a direct mode motion vector calculation method of a B picture according to the present invention,

In extracting the direct mode motion vector of a B (bi-predictive) picture in a video coding system, when the co-located macroblock in the list 1 reference picture of the B mode direct mode is intra mode. Using spatial redundancy, predict and calculate a list 0, list 1 reference picture and a motion vector for each list from neighboring blocks of the macroblock to be coded for the B picture, and calculate a direct mode motion vector of the B picture. Its characteristics are that it operates.

According to the present invention, in selecting the list 1 reference picture, when reference pictures referenced by neighboring macroblocks A, B, and C of the macroblock to be coded are different from each other, the reference picture having the smallest index is listed. 1 The feature is that it is selected as a reference picture.

According to the present invention, in selecting the list 1 reference picture, if a reference picture referred to by two or more macroblocks among neighboring macroblocks of the macroblock to be coded has the same index, the reference picture having the index is selected. The feature is that you select it as a list 1 reference picture.

According to the present invention, in predicting the motion vectors for the respective lists, if there are macroblocks having intra modes among the neighboring macroblocks A, B, and C, the motion vectors of list 0 and list 1 of the macroblocks are respectively determined. If set to 0, the motion vector having the same direction as the temporal position of each list reference picture is selected from the neighboring macroblocks and the motion vector of each list is obtained through median operation, or if the neighboring macroblocks are the motions of the same direction If you have two vectors, you can select only one from the macroblock and include them in the median operation to find the motion vectors of each list.

According to the present invention, if all valid reference picture indices for each list cannot be derived, the reference picture indices of list 0 and list 1 become 0, and the motion vector for each list is set to 0. There is this.

According to the present invention, by presenting a direct mode motion vector extraction technique in the B (Bi-predictive) picture defined in the next generation video compression technology, B picture coding by increasing the likelihood that the direct mode is selected as the prediction mode of the macroblock. There is an advantage to improve the coding (coding efficiency).

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

The present invention provides a method of deriving a direct motion vector when a macroblock of the same position in the list 1 reference picture in the direct mode is intra mode, when the list 1 reference picture is located later in time than the B picture, and than the B picture. List 0, for direct mode, by expressing the temporal distance between pictures as a signed value to simplify the method of obtaining the direct motion vector for the case where it is located temporally, and the equation used to obtain the direct mode motion vector. list 1 A method for obtaining a direct mode motion vector regardless of the reference picture position is presented separately. In this case, since the frame mode and the field mode are switched at the picture level, the B picture and the list 1 reference picture may be coded in a frame structure or a feed structure. Therefore, the macroblock of the B picture and the macroblock at the same position in the list 1 reference picture have four cases of frame / feed coding combinations.

[1] When the macroblock at the same position in the list 1 reference picture has intra mode

As shown in FIGS. 3F and 4F, macroblocks in the same position in the list 1 reference picture for direct mode in the direct mode are independent of the temporal position of the reference picture. It may have an intra mode. Since the macroblock having this mode does not have motion information, the conventional technique simply sets the direct motion vector to 0 and defines the list 0 reference picture as the most recently decoded picture. However, since the conventional techniques cannot guarantee high coding efficiency, the present invention uses spatial redundancy to list the list 0 and list 1 reference pictures and the motion vectors for each list from the neighboring blocks of the macroblock to be coded B picture. Predict and calculate.

The reference picture index for each list is obtained by the following method. FIG. 5 is a diagram for explaining calculation of a predicted motion vector of block E using motion vectors of neighboring blocks A, B, and C in consideration of general spatial redundancy.

If the reference picture indices of neighboring macroblocks A, B, and C are different from each other, the smallest reference picture index among them is determined as the reference picture index of the direct mode.

-If two of the neighboring macroblocks have the same reference picture index, the index is determined as the reference picture index of the direct mode.

If all of the neighboring macroblocks have the same reference picture index, the index is determined as the reference picture index of the direct mode.

In addition, the motion vector for each list is obtained through the following motion vector prediction. At this time, if there is a macroblock having an intra mode among the neighboring macroblocks A, B, and C, the motion vectors of list 0 and list 1 of the macroblock are set to 0, respectively.

-Motion vectors with the same direction as the temporal position of the reference picture of each list obtained above are selected from the surrounding macroblocks, and the motion vectors of each list are obtained through median operation.

If a neighboring macroblock has two motion vectors in the same direction, select only one of the macroblocks to be included in the median operation.

On the other hand, if all valid reference picture indices of list 0 and list 1 cannot be derived from neighboring blocks, reference picture indices of list 0 and list 1 become 0, respectively, and the motion vector for each list is set to 0.

[2] The list 1 reference picture in direct mode is later in time than the B picture

Case 1: When the macroblock of the B picture and the macroblock at the same position of the list 1 reference picture are both in frame mode

As shown in (a) to (f) of FIG. 3, a block at the same position of the list 1 reference picture may have one motion vector or two motion vectors. In the present invention, when a block having the same position has two motion vectors, one motion vector (L0 MV or L1 MV) is selected, and a motion vector of direct mode is derived from the selected motion vector (hereinafter, L0). A description will be given based on the case where MV (list 0 motion vector) is selected).

Therefore, FIG. 3A and FIG. 3C are the same as FIG. 6A, and FIG. 3B and FIG. 3E are FIG. 6C and 3G, respectively. May be simplified and displayed as shown in FIG.

If the list 0 reference picture and the list 1 reference picture in the direct mode exist in front of or behind the B picture in time (FIG. 6A), or in the list 0 and list 1 reference pictures in the direct mode than the B picture. When the list 0 reference picture exists later than the list 1 reference picture while being present in time (FIG. 6B), the motion vectors MV _F and MV _B in the direct mode are calculated as follows.

MV _F = TD _B x MV / TD _D

MV _B = (TD _B -TD _D ) x MV / TD _D

Where TD _B Denotes the temporal distance between the current B frame and the list 0 reference frame, and TD _D Denotes the temporal distance between the list 1 reference frame and the list 0 reference frame, respectively.

In addition, the above expression may be expressed as follows, which represents a case in which a bit operation is applied for convenience in performing an actual operation.

Z = TD _B x 256 / TD _D MV _F = (Z x MV + 128) >> 8

W = Z-256 MV _B = (W x MV + 128) >> 8

If both list 0 and list 1 reference pictures in the direct mode exist later in time than the B picture, and the list 0 reference picture exists before the list 1 reference picture ((c) of FIG. 6), the motion vector MV _F in the direct mode , MV _B is calculated as

MV _F =-TD _B x MV / TD _D

MV _B =-(TD _B + TD _D ) x MV / TD _D

Or it may be represented by the following equation.

Z =-TD _B x 256 / TD _D MV _F = (Z x MV + 128) >> 8

W = Z-256 MV _B = (W x MV + 128) >> 8

Case 2: When both the macroblock of the B picture and the macroblock at the same position of the list 1 reference picture are both in feed mode

FIG. 7 illustrates a case in which both the macroblocks of the B picture and the macroblocks at the same position of the list 1 reference picture are all in the shield mode. The motion vector in each block of the macroblock of the B picture is derived from the list 0 motion vector at the same position in the list 1 reference feed of the same parity.

If the list 0 and list 1 reference pictures in the direct mode exist in front of and behind the B picture in time (FIG. 7 (a)), or the list 0 and list 1 reference pictures in the direct mode are more temporal than the B pictures. If the list 0 reference picture is later than the list 1 reference picture (Fig. 7 (b)), each of the frames i in frame B (i = 0 is the first and i = 1 is the second) Direct motion list 0, list 1 motion vectors MV _{F, i} and MV _{B, i} are calculated as follows.

MV _{F, i} = TD _{B, i} x MV _i / TD _{D, i}

MV _{B, i} = (TD _{B, i} -TD _{D, i} ) x MV _i / TD _{D, i}

Where MV _i is the list 0 motion vector of the same block in _i of the list 1 reference frame, TD _{B, i} is the temporal distance between the current B and list 0 reference fields, and TD _{D, i} is Means the temporal distance between the 1st reference list and the 0 list feed. The above expression can be expressed as

Z = TD _{B, i} x 256 / TD _{D, i} MV _{F, i} = (Z x MV _i + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV _i + 128) >> 8

If a block at the same position in the frame i of the list 1 reference frame has a motion vector that points to the shield of the frame that is later in time than the B picture, both the list 0 and list 1 reference pictures in direct mode are less than the B picture. When the list 0 reference picture is present in front of the list 1 reference picture (FIG. 7 (c) and (d)), the list 0 and list 1 motion vectors MV _{F, i} and MV _{B, i are in the} direct mode. Is calculated as

MV _{F, i} =-TD _{B, i} x MV _i / TD _{D, i}

MV _{B, i} =-(TD _{B, i} + TD _{D, i} ) x MV _i / TD _{D, i}

Or it may be represented by the following equation.

Z = -TD _{B, i} x 256 / TD _{D, i} MV _{F, i} = (Z x MV _i + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV _i + 128) >> 8

Case 3: When the macroblock of the B picture is in shield mode and the macroblock at the same position of the list 1 reference picture is in frame mode

FIG. 8 illustrates a case in which a macroblock of a B picture is in a shield mode and a macroblock in the same position of the list 1 reference picture is in a frame mode. Here, if the vertical coordinate of the current macroblock is y _current and the macroblock vertical coordinate of the same position of the list 1 reference picture is y _co-located , a y _co-located = 2 xy _current relationship is established between the two coordinates. The list 0 and list 1 reference feeds exist in the same parity of the list 0 and list1 reference frames.

If the list 0 and list 1 reference pictures in the direct mode exist in front of and behind the B picture temporally (Fig. 8 (a)), or the list 0 and list 1 reference pictures in the direct mode are more temporal than the B pictures. If the list 0 reference picture is later than the list 1 reference picture (Fig. 8 (b)), the list 0, list 1 motion vector MV _{F, i} , MV _{B, i} is calculated as

MV _{F, i} = TD _{B, i} x MV / TD _D

MV _{B, i} = (TD _{B, i} -TD _D ) x MV / TD _D

The above expression can be expressed as

Z = TD _{B, i} x 256 / TD _D MV _{F, i} = (Z x MV + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV + 128) >> 8

If the same block in the list 1 reference frame has a motion vector that points to a frame that is later in time than the B picture, both list 0 and list 1 reference pictures in direct mode are later in time than the B picture. When the reference picture is present before the list 1 reference picture (Fig. 8 (c)), list 0, list 1 motion vectors MV _{F, i} and MV _{B, i} in the direct mode for each of the frames i of the B frame are Is calculated as

MV _{F, i} = -TD _{B, i} x MV / TD _D

MV _{B, i} =-(TD _{B, i} + TD _D ) x MV / TD _D

Or it may be represented by the following equation.

Z =-TD _{B, i} x 256 / TD _D MV _{F, i} = (Z x MV + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV + 128) >> 8

Here, TD _{B, i} is the temporal distance between the current B feed and the list 0 reference feed, TD _D is the temporal distance between the list 1 reference list and the list 0 reference frame, and MV is the list 1 reference of the direct mode. This is the motion vector of the same block in the frame.

Case 4: When the macroblock of the B picture is in frame mode and the macroblock at the same position of the list 1 reference picture is in the feed mode

FIG. 9 shows a case in which a macroblock of a B picture is a frame mode and a macroblock at the same position of a list 1 reference picture is a shield mode. Here, if the vertical coordinate of the current macroblock is y _current and the macroblock vertical coordinate of the same position of the list 1 reference picture is y _co-located , a y _co-located = y _current / 2 relationship is established between the two coordinates. Since the 0 of the list 1 reference frame has a closer temporal distance to the B picture than the 1 of the frame, the motion information of the block at the same position in the 0 is used to obtain the motion vector of the direct mode.

If the list 0 and list 1 reference pictures in the direct mode exist in front of and behind the B picture temporally (Fig. 9 (a)), or the list 0 and list 1 reference pictures in the direct mode in time than the B picture. If the list 0 reference picture exists later than the list 1 picture (FIG. 9B), the list 0, list 1 motion vectors MV _F and MV _B in the direct mode of the B frame are calculated as follows.

MV _F = TD _B x MV ₀ / TD _{D, 0}

MV _B = (TD _B -TD _{D, 0} ) x MV ₀ / TD _{D, 0}

* The above expression can be expressed as

Z = TD _B x 256 / TD _{D, 0} MV _F = (Z x MV ₀ + 128) >> 8

W = Z-256 MV _B = (W x MV ₀ + 128) >> 8

If a block at the same position in feed 0 of the list 1 reference frame has a motion vector that points to the feed of a frame that is later in time than the B picture, then both list 0 and list 1 reference pictures in direct mode are less than the B picture. When the list 0 reference picture is present before the list 1 reference picture while being present later in time (FIG. 9C), the list 0, list 1 motion vectors MV _F and MV _B in the direct mode are calculated as follows.

MV _F =-TD _B x MV ₀ / TD _{D, 0}

MV _B =-(TD _B + TD _{D, 0} ) x MV ₀ / TD _{D, 0}

Or it may be represented by the following equation.

Z =-TD _B x 256 / TD _{D, 0} MV _F = (Z x MV ₀ + 128) >> 8

W = Z-256 MV _B = (W x MV ₀ + 128) >> 8

Here, TD _B is the temporal distance between the current B frame and the list 0 reference frame, TD _{D, 1} is the temporal distance between the 0 and list 0 reference frames of the list 1 reference frame, and MV ₀ is the direct mode. This is the motion vector of the block at the same position in list 1 reference feed at.

[3] The list 1 reference picture in direct mode is temporally ahead of the B picture

In this case, both the list 0 and list 1 reference pictures are always in front of the B picture in time.

Case 1: When the macroblock of the B picture and the macroblock at the same position of the list 1 reference picture are both in frame mode

As shown in FIG. 4, a block at the same position of the list 1 reference picture may have one motion vector or two motion vectors. In the present invention, when a block having the same position has two motion vectors, one motion vector (L0 MV) Alternatively, L1 MV) is selected, and a motion vector of direct mode is derived from the selected motion vector (hereinafter, description will be made based on the case where L0 MV (list 0 motion vector) is selected).

Therefore, FIG. 4 (a) (c) (e) (g) (h) becomes the same as FIG. 10 (a), and FIG. 4 (b) (d) is simplified as shown in FIG. 10 (b). It can be displayed.

If the list 0 reference picture in the direct mode exists before the list 1 reference picture in time, the motion vectors MV _F and MV _B in the direct mode are calculated as follows (Fig. 10 (a)).

MV _F = TD _B x MV / TD _D

MV _B = (TD _B -TD _D ) x MV / TD _D

Where TD _B is the temporal distance between the current B frame and the list 0 reference frame, TD _D is the temporal distance between the list 1 reference list and the list 0 reference frame, and MV is the same as in the list 1 reference picture in the direct mode. The motion vector of the block of position.

And the above expression can be expressed as

Z = TD _B x 256 / TD _D MV _F = (Z x MV + 128) >> 8

W = Z-256 MV _B = (W x MV + 128) >> 8

If the list 0 reference picture in the direct mode exists later in time than the list 1 reference picture, the motion vectors MV _F and MV _B in the direct mode are calculated as follows (Fig. 10 (b)).

MV _F =-TD _B x MV / TD _D

MV _B =-(TD _B + TD _D ) x MV / TD _D

Or it may be represented by the following equation.

Z =-TD _B x 256 / TD _D MV _F = (Z x MV + 128) >> 8

W = Z-256 MV _B = (W x MV + 128) >> 8

Where TD _B is the temporal distance between the current B frame and the list 0 reference frame, TD _D is the temporal distance between the list 1 reference list and the list 0 reference frame, and MV is the same as in the list 1 reference picture in the direct mode. The motion vector of the block of position.

Case 2: When both the macroblock of the B picture and the macroblock at the same position of the list 1 reference picture are both in feed mode

If the list 0 reference picture in direct mode exists temporally before the list 1 reference picture, then the list 0, list 1 motion vectors MV _{F, i} and MV _{B, i} It is calculated as follows (FIGS. 11A and 11B).

MV _{F, i} = TD _{B, i} x MV _i / TD _{D, i}

MV _{B, i} = (TD _{B, i} -TD _{D, i} ) x MV _i / TD _{D, i}

The above expression can be expressed as

Z = TD _{B, i} x 256 / TD _{D, i} MV _{F, i} = (Z x MV _i + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV _i + 128) >> 8

Here, TD _{B, i} is the temporal distance between the current B feed and the list 0 reference feed, TD _{D, i} is the temporal distance between the list 1 reference list and the list 0 reference feed, and MV _i is the direct The motion vector of the same block in the mode's list 1 reference field.

If a block at the same position in the frame i of the list 1 reference frame has a motion vector that points to the feed of the frame that is later in time, the list 0 reference picture in direct mode exists later in time than the list 1 reference picture. In the case of direct mode, list 0 and list 1 motion vectors MV _{F, i} and MV _{B, i} are calculated as follows ((c) and (d) of FIG. 11).

MV _{F, i} = -TD _{B, i} x MV _i / TD _{D, i}

MV _{B, i} =-(TD _{B, i} + TD _{D, i} ) x MV _i / TD _{D, i}

Or it may be represented by the following equation.

Z =-TD _{B, i} x 256 / TD _{D, i} MV _{F, i} = (Z x MV _i + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV _i + 128) >> 8

Here, TD _{B, i} is the temporal distance between the current B feed and the list 0 reference feed, TD _{D, i} is the temporal distance between the list 1 reference list and the list 0 reference feed, and MV _i is the direct The motion vector of the same block in the mode's list 1 reference field.

Case 3: When the macroblock of the B picture is in shield mode and the macroblock at the same position of the list 1 reference picture is in frame mode

If the list 0 reference picture in direct mode exists temporally before the list 1 reference picture, then the list 0, list 1 motion vectors MV _{F, i} and MV _{B, i} It is calculated as follows (Fig. 12 (a)).

MV _{F, i} = TD _{B, i} x MV / TD _D

MV _{B, i} = (TD _{B, i} -TD _D ) x MV / TD _D

The above expression can be expressed as

Z = TD _{B, i} x 256 / TD _D MV _{F, i} = (Z x MV + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV + 128) >> 8

Here, TD _{B, i} is the temporal distance between the current B feed and the list 0 reference feed, TD _D is the temporal distance between the list 1 reference list and the list 0 reference frame, and MV is the list 1 reference of the direct mode. This is the motion vector of the same block in the frame.

If a block in the same position in the list 1 reference frame has a motion vector pointing to a frame that is later in time, if each list 0 reference picture in direct mode is later in time than the list 1 reference picture, each picture in frame B List 0 and list 1 motion vectors MV _{F, i} and MV _B, i of the direct mode for the yield i are calculated as follows (Fig. 12 (b)).

MV _{F, i} =-TD _{B, i} x MV / TD _D

MV _{B, i} =-(TD _{B, i} + TD _D ) x MV / TD _D

Or it may be represented by the following equation.

Z =-TD _{B, i} x 256 / TD _D MV _{F, i} = (Z x MV + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV + 128) >> 8

Here, TD _{B, i} is the temporal distance between the current B feed and the list 0 reference feed, TD _D is the temporal distance between the list 1 reference list and the list 0 reference frame, and MV is the list 1 reference of the direct mode. This is the motion vector of the same block in the frame.

Case 4: When the macroblock of the B picture is in frame mode and the macroblock at the same position of the list 1 reference picture is in the feed mode

Since the distance 1 is closer to the B picture than the field 1 (f1) of the list 1 reference frame, the motion information of the block at the same position in the field 1 (f1) determines the motion vector of the direct mode. It is used to

If the list 0 reference picture in direct mode exists temporally before the list 1 reference picture, the list 0, list 1 motion vectors MV _F and MV _B in direct mode for each frame i of frame _B are calculated as follows. (FIG. 13A).

MV _F = TD _B x MV ₁ / TD _{D, 1}

MV _B = (TD _B -TD _{D, 1} ) x MV ₁ / TD _{D, 1}

The above expression can be expressed as

Z = TD _B x 256 / TD _{D, 1} MV _F = (Z x MV ₁ + 128) >> 8

W = Z-256 MV _B = (W x MV ₁ + 128) >> 8

Here, TD _B is a temporal distance between a current B frame and a list 0 reference frame, TD _{D, 1} is a temporal distance between a feed 1 and a list 0 reference feed of a list 1 reference frame, and MV ₁ is the direct mode. This is the motion vector of the block at the same position in feed 1 of the list 1 reference frame.

If a block at the same position in feed 1 (f1) of the list 1 reference frame has a motion vector that points to the feed of a frame that is later in time, the list 0 reference picture in direct mode is temporal than the list 1 reference picture. If it exists later, list 0, list 1 motion vectors MV _F and MV _B in direct mode are calculated as follows (Fig. 13 (b)).

MV _F =-TD _B x MV ₁ / TD _{D, 1}

MV _B =-(TD _B + TD _{D, 1} ) x MV ₁ / TD _{D, 1}

Or it may be represented by the following equation.

Z = -TD _B x 256 / TD _{D, 1} MV _F = (Z x MV ₁ + 128) >> 8

W = Z-256 MV _B = (W x MV ₁ + 128) >> 8

Here, TD _B is a temporal distance between a current B frame and a list 0 reference frame, TD _{D, 1} is a temporal distance between a feed 1 and a list 0 reference feed of a list 1 reference frame, and MV ₁ is the direct mode. This is the motion vector of the block at the same position in feed 1 of the list 1 reference frame.

[4] calculating a direct mode motion vector by expressing a temporal distance between pictures as a signed value

When the list 1 reference picture of the direct mode is located ahead of or behind the B picture in time, two kinds of equations exist for each case. If the value is expressed as, it can be simplified as follows.

Case 1: When the macroblock of the B picture and the macroblock at the same position of the list 1 reference picture are both in frame mode

When the macroblocks of the B picture and the macroblocks at the same position of the list 1 reference picture are both in the frame mode, the motion vectors MV _F and MV _B in the direct mode of the B picture can be obtained from the following equation.

MV _F = TD _B x MV / TD _D

MV _B = (TD _B -TD _D ) x MV / TD _D

or

Z = TD _B x 256 / TD _D MV _F = (Z x MV + 128) >> 8

W = Z-256 MV _B = (W x MV + 128) >> 8

Here, TD _B is the temporal distance between the current B frame and the list 0 reference frame. The temporal distance calculated from the B frame is represented by a positive sign, and the temporal distance calculated from the list 0 reference frame is represented by a negative sign. TD _D is the temporal distance between the list 1 reference frame and the list 0 reference frame. The temporal distance calculated from the list 1 reference frame is indicated by a positive sign and the temporal distance calculated from the list 0 reference frame is negative. The MV is a motion vector of a block of the same position in the list 1 reference picture of the direct mode.

Case 2: When both the macroblock of the B picture and the macroblock at the same position of the list 1 reference picture are both in feed mode

When the macroblocks of the B picture and the macroblocks at the same position of the list 1 reference picture are both in the field mode, the motion vectors MV _{F, i} and MV _B in the direct mode for each field i of the B frame. _{, i} can be obtained from the following equation.

MV _{F, i} = TD _{B, i} x MV _i / TD _{D, i}

MV _{B, i} = (TD _{B, i} -TD _{D, i} ) x MV _i / TD _{D, i}

or

Z = TD _{B, i} x 256 / TD _{D, i} MV _{F, i} = (Z x MV _i + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV _i + 128) >> 8

Where TD _{B, i} is the temporal distance between the current B and list 0 reference feeds. The time distance calculated from the B feeds is represented by a positive sign and the time distance calculated from the list 0 reference feed is negative. TD _{D, i} is the temporal distance between list 1 reference list and list 0 reference feed, and the time distance calculated from list 1 reference feed is represented by a positive sign and list The temporal distance calculated from the zero reference period is represented by a negative sign, and MV _i is a motion vector of the same position block in the list 1 reference shield of the direct mode.

Case 3: When the macroblock of the B picture is in shield mode and the macroblock at the same position of the list 1 reference picture is in frame mode

When the macroblock of the B picture is in the shield mode and the macroblock at the same position of the list 1 reference picture is in the frame mode, the motion vectors MV _{F, i} , MV _{B, i} can be obtained from the following equation.

MV _{F, i} = TD _{B, i} x MV / TD _D

MV _{B, i} = (TD _{B, i} -TD _D ) x MV / TD _D

or

Z = TD _{B, i} x 256 / TD _D MV _{F, i} = (Z x MV + 128) >> 8

W = Z-256 MV _{B, i} = (W x MV + 128) >> 8

Where TD _{B, i} is the temporal distance between the current B and list 0 reference feeds. The time distance calculated from the B feeds is represented by a positive sign and the time distance calculated from the list 0 reference feed is negative. TD _D is the temporal distance between the list 1 reference frame and the list 0 reference frame. The temporal distance calculated from the list 1 reference frame is represented by a positive sign and is calculated from the list 0 reference frame. The temporal distance is indicated by a negative sign, and MV is a motion vector of a block in the same position in the list 1 reference frame of the direct mode.

Case 4: When the macroblock of the B picture is in frame mode and the macroblock at the same position of the list 1 reference picture is in the feed mode

If the macroblock of the B picture is in frame mode and the macroblock at the same position of the list 1 reference picture is in the shield mode, and the list 1 reference picture is later in time than the B picture, the 0 of the list 1 reference frame Since the temporal distance is closer to the B picture than the FID 1, the motion information of the block of the same position in FID 0 is used to obtain the motion vector of the direct mode. Therefore, the motion vectors MV _F and MV _B in the direct mode of the B frame can be obtained from the following equation in which the motion information of the block at the same position in the feed 0 of the list 1 reference frame is used to obtain the motion vector in the direct mode. .

MV _F = TD _B x MV ₀ / TD _{D, 0}

MV _B = (TD _B -TD _{D, 0} ) x MV ₀ / TD _{D, 0}

or

Z = TD _B x 256 / TD _{D, 0} MV _F = (Z x MV ₀ + 128) >> 8

W = Z-256 MV _B = (W x MV ₀ + 128) >> 8

Here, TD _B is the temporal distance between the current B frame and the list 0 reference frame. The temporal distance calculated from the B frame is represented by a positive sign, and the temporal distance calculated from the list 0 reference frame is represented by a negative sign. TD _{D, 0} is the temporal distance between feed ₀ of the list 1 reference frame and the list 0 reference feed, and the calculated time distance from the feed 0 of the list 1 reference frame is represented by a positive sign. The temporal distance calculated from the list 0 reference period is represented by a negative sign, and MV ₀ is a motion vector of a block at the same position in the field 0 of the list 1 reference frame in the direct mode.

If the list 1 reference picture is present in front of the B picture in time, the motion of the block of the same position in the first location is because the distance 1 of the list 1 reference frame is closer to the B picture than the time 0. The information is used to find the motion vector of the direct mode. Accordingly, the motion vectors MV _F and MV _B in the direct mode of the B frame can be obtained from the following equation in which the motion information of the block at the same position in the feed 1 of the list 1 reference frame is used to obtain the motion vector in the direct mode. .

MV _F = TD _B x MV ₁ / TD _{D, 1}

MV _B = (TD _B -TD _{D, 1} ) x MV ₁ / TD _{D, 1}

or

Z = TD _B x 256 / TD _{D, 1} MV _F = (Z x MV ₁ + 128) >> 8

W = Z-256 MV _B = (W x MV ₁ + 128) >> 8

Here, TD _B is the temporal distance between the current B frame and the list 0 reference frame. The temporal distance calculated from the B frame is represented by a positive sign, and the temporal distance calculated from the list 0 reference frame is represented by a negative sign. TD _{D, 1} is the temporal distance between feed ₁ of the list 1 reference frame and the list 0 reference feed, and the calculated time distance from feed 1 of the list 1 reference frame is represented by a positive sign and The temporal distance calculated from the zero reference period is represented by a negative sign, and MV ₁ is a motion vector of the same position block in the first position of the list 1 reference frame of the direct mode.

As described above, according to the method of calculating a direct mode motion vector of a B picture according to the present invention, a direct mode motion vector extraction technique is proposed in a B (Bi-predictive) picture defined in a next-generation video compression technique, and the direct mode is a macro. There is an advantage in that the B picture coding efficiency can be improved by increasing the probability of being selected as the prediction mode of the block.

1 is a view for explaining a list 1 reference picture (list 1 reference picture for direct mode) of the direct mode in a general IBBBP pattern.

2 is a view for explaining a list 1 reference picture in direct mode in a general IBBB pattern.

3 is a diagram illustrating a case where a list 1 reference picture in a general direct mode is located later in time than a B picture (L0 MV: list 0 motion vector and L1 MV: list 1 motion vector).

FIG. 4 is a diagram illustrating a case where a list 1 reference picture in a general direct mode is located in front of a B picture in time (L0 MV: list 0 motion vector and L1 MV: list 1 motion vector). FIG.

FIG. 5 is a diagram for explaining calculation of a predictive motion vector of block E using motion vectors of neighboring blocks A, B, and C in consideration of general spatial redundancy. FIG.

FIG. 6 is a diagram illustrating a case in which both macroblocks of a B picture and macroblocks at the same position of the list 1 reference picture are in frame mode when the list 1 reference picture of the general direct mode is located later in time than the B picture. FIG.

FIG. 7 is a diagram illustrating a case in which both macroblocks of a B picture and macroblocks at the same position of the list 1 reference picture are in a closed mode when the list 1 reference picture in the general direct mode is located later in time than the B picture. FIG.

FIG. 8 illustrates a case in which a macroblock of a B picture is a shield mode and a macroblock in the same position of the list 1 reference picture is a frame mode when the list 1 reference picture of the general direct mode is located later in time than the B picture. drawing.

FIG. 9 illustrates a case in which a macroblock of a B picture is a frame mode and a macroblock at the same position of the list 1 reference picture is a feed mode when the list 1 reference picture in the general direct mode is located later in time than the B picture. drawing.

FIG. 10 is a diagram illustrating a case in which both macroblocks of a B picture and macroblocks at the same position of the list 1 reference picture are both in frame mode when the list 1 reference picture in the general direct mode is temporally positioned before the B picture. FIG.

FIG. 11 is a diagram illustrating a case in which both macroblocks of a B picture and macroblocks at the same position of the list 1 reference picture are in a fail mode when a list 1 reference picture in a general direct mode is located ahead of time in a B picture. FIG.

12 illustrates a case in which a macroblock of a B picture is a shield mode and a macroblock at the same position of the list 1 reference picture is a frame mode when the list 1 reference picture of the general direct mode is located in front of the B picture temporally. drawing.

FIG. 13 illustrates a case in which a macroblock of a B picture is a frame mode and a macroblock at the same position of the list 1 reference picture is a feed mode when a list 1 reference picture in a general direct mode is located in front of a B picture temporally. drawing.

Claims (12)

A method of determining a motion vector for deriving motion vectors of a bi-predictive image block, Regardless of whether a co-located block co-located with respect to a bi-predicted image block includes a list 1 motion vector, the list 0 motion vector of the co-located image block Selecting as a motion vector for deriving the motion vectors. The method of claim 1, And said at least one image block is associated with a list 1 reference picture. The method of claim 2, And determining the reference picture indicated by the selected list 0 motion vector as a list 0 reference picture. The method of claim 3, wherein Deriving the motion vectors for the bi-predictive image block from the selected list 0 motion vector. The method of claim 1, And determining the reference picture indicated by the selected list 0 motion vector as a list 0 reference picture. The method of claim 1, And determining a list 0 reference picture from reference information of the image block at the same location including the selected list 0 motion vector. The method of claim 1, And determining, as a list 0 reference picture, a reference picture referenced by the image block at the same location including the selected list 0 motion vector. The method of claim 1, And deriving the motion vectors for the bipredicted image block from the selected list 0 motion vector. The method of claim 1, And wherein said selecting step is performed by a decoder operation in direct mode. The method of claim 1, And the image block in the same position includes list 1 and list 0 motion vectors. The method of claim 1, And said at least one image block has two or more prediction modes. A method of determining a motion vector for deriving motion vectors of a bi-predictive image block, If a co-located block in the same position with respect to the bi-predictive image block contains a list 0 motion vector, the list 0 motion vector of the image block in the same position is ignored and the list 0 motion vector is ignored. Selecting as a motion vector for deriving the motion vectors of the bi-prediction image block.