Motion Compensation - Revision history

DonDiego: categorize

2007-05-23T10:48:43Z

categorize

← Older revision		Revision as of 03:48, 23 May 2007
Line 1:		Line 1:
	'''Motion compensation''' encompasses a number of methods for using predicted motion vectors to recreate pixels within a frame from past, future video frames or even the current video frame. Wikipedia has a [http://en.wikipedia.org/wiki/Motion_compensation fairly extensive article] discussing and detailing some of the defined techniques.		'''Motion compensation''' encompasses a number of methods for using predicted motion vectors to recreate pixels within a frame from past, future video frames or even the current video frame. Wikipedia has a [http://en.wikipedia.org/wiki/Motion_compensation fairly extensive article] discussing and detailing some of the defined techniques.

			[[Category:Compression Theory]]

Superdump: Moving some material to a more appropriate home in Motion Prediction

2007-05-21T15:50:38Z

Moving some material to a more appropriate home in Motion Prediction

← Older revision		Revision as of 08:50, 21 May 2007
Line 1:		Line 1:
	'''Motion compensation''' encompasses a number of methods for using predicted motion vectors to recreate pixels within a frame from past, future video frames or even the current video frame. Wikipedia has a [http://en.wikipedia.org/wiki/Motion_compensation fairly extensive article] discussing and detailing some of the defined techniques.		'''Motion compensation''' encompasses a number of methods for using predicted motion vectors to recreate pixels within a frame from past, future video frames or even the current video frame. Wikipedia has a [http://en.wikipedia.org/wiki/Motion_compensation fairly extensive article] discussing and detailing some of the defined techniques.


	~~== Direct Prediction ==~~

	'''Direct prediction''' offers methods for calculating motion vectors for already coded vectors to reduce the bit cost of coding B-frames and take advantage of motion coherences in different domains. There are two methods of direct prediction - spatial and temporal.

	The '''spatial''' mode calculates motion vectors from the vectors (left, top left, top? if available?) of neighbouring blocks. Depending on the prediction method used in said neighbouring blocks, the resultant B-frame prediction mode can be one of forward, backward or interpolation.

	'''Temporal''' direct prediction necessarily uses the interpolation mode. The temporal mode considers the forward motion vector of the colocated block in the first list 1 reference frame. If there are no reference B-frames or B-frame reordering, the first list 1 reference frame is usually the next P-frame.

	~~[[Image:H264_direct_temporal_3.png]]~~

	~~[[Image:H264_direct_temporal_4.png]]~~

	~~''Note: As poc(L1) > poc(cur), the signs of the components of MV_L1 are swapped.''~~

	Where MV_L1 colocated is the motion vector of the colocated block in the first L1 reference frame and poc() gives the position of frame in the sequence of frames (c.f. picture order count §8.2.1 of the H.264 spec).

	~~Consider the following setup where we have a sequence of frames PbbbP with no reference B-frames / frame reordering:~~

	~~[[Image:H264_direct_temporal_1.png]]~~

	To visualise the process, the end points of the forward motion vector remain in the same frames (starting at P2, ending at P1) but the vector is translated such that it appropriately intersects the current macroblock (in red) in the current frame (B1).

	~~[[Image:H264_direct_temporal_2.png]]~~

	~~The translated vector is essentially 'split' setting the start point at the appropriate position of the current macroblock and pointing along the translated vector forward to P1 and backward to P2.~~

Superdump at 01:52, 21 May 2007

2007-05-21T01:52:17Z

← Older revision		Revision as of 18:52, 20 May 2007
Line 16:		Line 16:
	''Note: As poc(L1) > poc(cur), the signs of the components of MV_L1 are swapped.''		''Note: As poc(L1) > poc(cur), the signs of the components of MV_L1 are swapped.''

	Where MV_L1 colocated is the motion vector of the colocated block in the first L1 reference frame and poc() gives the position of frame in the sequence of frames.		Where MV_L1 colocated is the motion vector of the colocated block in the first L1 reference frame and poc() gives the position of frame in the sequence of frames (c.f. picture order count §8.2.1 of the H.264 spec).

	Consider the following setup where we have a sequence of frames PbbbP with no reference B-frames / frame reordering:		Consider the following setup where we have a sequence of frames PbbbP with no reference B-frames / frame reordering:

Superdump at 02:01, 20 May 2007

2007-05-20T02:01:29Z

← Older revision		Revision as of 19:01, 19 May 2007
Line 14:		Line 14:
	[[Image:H264_direct_temporal_4.png]]		[[Image:H264_direct_temporal_4.png]]

	''Note: As poc(L1) > poc(cur), the signs of the ~~directions~~ of MV_L1 are swapped.''		''Note: As poc(L1) > poc(cur), the signs of the components of MV_L1 are swapped.''

	Where MV_L1 colocated is the motion vector of the colocated block in the first L1 reference frame and poc() gives the position of frame in the sequence of frames.		Where MV_L1 colocated is the motion vector of the colocated block in the first L1 reference frame and poc() gives the position of frame in the sequence of frames.

Superdump at 02:00, 20 May 2007

2007-05-20T02:00:17Z

← Older revision		Revision as of 19:00, 19 May 2007
Line 8:		Line 8:
	The '''spatial''' mode calculates motion vectors from the vectors (left, top left, top? if available?) of neighbouring blocks. Depending on the prediction method used in said neighbouring blocks, the resultant B-frame prediction mode can be one of forward, backward or interpolation.		The '''spatial''' mode calculates motion vectors from the vectors (left, top left, top? if available?) of neighbouring blocks. Depending on the prediction method used in said neighbouring blocks, the resultant B-frame prediction mode can be one of forward, backward or interpolation.

	'''Temporal''' direct prediction necessarily uses the interpolation mode. The temporal mode considers the forward motion vector of the ~~collocated~~ block in the first list 1 reference frame. If there are no reference B-frames or B-frame reordering, the first list 1 reference frame is usually the next P-frame.		'''Temporal''' direct prediction necessarily uses the interpolation mode. The temporal mode considers the forward motion vector of the colocated block in the first list 1 reference frame. If there are no reference B-frames or B-frame reordering, the first list 1 reference frame is usually the next P-frame.

	[[Image:H264_direct_temporal_3.png]]		[[Image:H264_direct_temporal_3.png]]

	LaTeX notation: {MV}_{forward} = \frac { {MV}_{L1}^{collocated} * \left ( poc \left ( cur \right ) - poc \left ( L0 \right ) \right ) } { \left \| poc \left ( L1 \right ) - poc \left ( L0 \right ) \right \| }

	[[Image:H264_direct_temporal_4.png]]		[[Image:H264_direct_temporal_4.png]]
Line 18:		Line 16:
	''Note: As poc(L1) > poc(cur), the signs of the directions of MV_L1 are swapped.''		''Note: As poc(L1) > poc(cur), the signs of the directions of MV_L1 are swapped.''

	LaTeX notation: {MV}_{backward} = \frac { {MV}_{L1}^{collocated} * \left ( poc \left ( cur \right ) - poc \left ( L1 \right ) \right ) } { \left \| poc \left ( L1 \right ) - poc \left ( L0 \right ) \right \| }		Where MV_L1 colocated is the motion vector of the colocated block in the first L1 reference frame and poc() gives the position of frame in the sequence of frames.

	Where MV_L1 ~~collocated~~ is the motion vector of the ~~collocated~~ block in the first L1 reference frame and poc() gives the position of frame in the sequence of frames.

	Consider the following setup where we have a sequence of frames PbbbP with no reference B-frames / frame reordering:		Consider the following setup where we have a sequence of frames PbbbP with no reference B-frames / frame reordering:

Superdump: Add images of equations as LaTeX interpreter plugin isn't in place.

2007-05-20T01:43:01Z

Add images of equations as LaTeX interpreter plugin isn't in place.

← Older revision		Revision as of 18:43, 19 May 2007
Line 10:		Line 10:
	'''Temporal''' direct prediction necessarily uses the interpolation mode. The temporal mode considers the forward motion vector of the collocated block in the first list 1 reference frame. If there are no reference B-frames or B-frame reordering, the first list 1 reference frame is usually the next P-frame.		'''Temporal''' direct prediction necessarily uses the interpolation mode. The temporal mode considers the forward motion vector of the collocated block in the first list 1 reference frame. If there are no reference B-frames or B-frame reordering, the first list 1 reference frame is usually the next P-frame.

	<math>{MV}_{forward} = \frac { {MV}_{L1}^{collocated} * \left \| poc \left ( cur \right ) - poc \left ( L1 \right ) \right \| } { \left \| poc \left ( L1 \right ) - poc \left ( L0 \right ) \right \| }</math>		[[Image:H264_direct_temporal_3.png]]

	~~<math>~~{MV}_{~~backward~~} = \frac { {MV}_{L1}^{collocated} * \left \| poc \left ( cur \right ) - poc \left ( L0 \right ) \right \| } { \left \| poc \left ( L1 \right ) - poc \left ( L0 \right ) \right \| }~~</math>~~		LaTeX notation: {MV}_{forward} = \frac { {MV}_{L1}^{collocated} * \left ( poc \left ( cur \right ) - poc \left ( L0 \right ) \right ) } { \left \| poc \left ( L1 \right ) - poc \left ( L0 \right ) \right \| }

	~~Where <math~~>{MV}_{L1}^{collocated}~~</math>~~ is the motion vector of the collocated block in the first L1 reference frame and ~~<math>~~poc ~~\left~~ ( ~~L1 \right~~ )~~</math> is~~ the position of ~~the first L1 reference~~ in the sequence of frames.		[[Image:H264_direct_temporal_4.png]]

			''Note: As poc(L1) > poc(cur), the signs of the directions of MV_L1 are swapped.''

			LaTeX notation: {MV}_{backward} = \frac { {MV}_{L1}^{collocated} * \left ( poc \left ( cur \right ) - poc \left ( L1 \right ) \right ) } { \left \| poc \left ( L1 \right ) - poc \left ( L0 \right ) \right \| }

			Where MV_L1 collocated is the motion vector of the collocated block in the first L1 reference frame and poc() gives the position of frame in the sequence of frames.

	Consider the following setup where we have a sequence of frames PbbbP with no reference B-frames / frame reordering:		Consider the following setup where we have a sequence of frames PbbbP with no reference B-frames / frame reordering:

Superdump: LaTeX equations - though they don't work at this precise moment.

2007-05-20T01:01:27Z

LaTeX equations - though they don't work at this precise moment.

← Older revision		Revision as of 18:01, 19 May 2007
Line 8:		Line 8:
	The '''spatial''' mode calculates motion vectors from the vectors (left, top left, top? if available?) of neighbouring blocks. Depending on the prediction method used in said neighbouring blocks, the resultant B-frame prediction mode can be one of forward, backward or interpolation.		The '''spatial''' mode calculates motion vectors from the vectors (left, top left, top? if available?) of neighbouring blocks. Depending on the prediction method used in said neighbouring blocks, the resultant B-frame prediction mode can be one of forward, backward or interpolation.

	'''Temporal''' direct prediction necessarily uses the interpolation mode. The temporal mode considers the forward motion vector of the collocated block in the first list 1 reference frame.		'''Temporal''' direct prediction necessarily uses the interpolation mode. The temporal mode considers the forward motion vector of the collocated block in the first list 1 reference frame. If there are no reference B-frames or B-frame reordering, the first list 1 reference frame is usually the next P-frame.

	~~Insert mathematical theory here~~. :)		<math>{MV}_{forward} = \frac { {MV}_{L1}^{collocated} * \left \| poc \left ( cur \right ) - poc \left ( L1 \right ) \right \| } { \left \| poc \left ( L1 \right ) - poc \left ( L0 \right ) \right \| }</math>

			<math>{MV}_{backward} = \frac { {MV}_{L1}^{collocated} * \left \| poc \left ( cur \right ) - poc \left ( L0 \right ) \right \| } { \left \| poc \left ( L1 \right ) - poc \left ( L0 \right ) \right \| }</math>

			Where <math>{MV}_{L1}^{collocated}</math> is the motion vector of the collocated block in the first L1 reference frame and <math>poc \left ( L1 \right )</math> is the position of the first L1 reference in the sequence of frames.

	Consider the following setup where we have a sequence of frames PbbbP with no reference B-frames / frame reordering:		Consider the following setup where we have a sequence of frames PbbbP with no reference B-frames / frame reordering:

Superdump at 23:18, 19 May 2007

2007-05-19T23:18:52Z

← Older revision		Revision as of 16:18, 19 May 2007
Line 4:		Line 4:
	== Direct Prediction ==		== Direct Prediction ==

	Direct prediction offers methods for calculating motion vectors for already coded vectors to reduce the bit cost of coding B-frames and take advantage of motion coherences in different domains. There are two methods of direct prediction - spatial and temporal.		'''Direct prediction''' offers methods for calculating motion vectors for already coded vectors to reduce the bit cost of coding B-frames and take advantage of motion coherences in different domains. There are two methods of direct prediction - spatial and temporal.

	The '''spatial''' mode calculates motion vectors from the vectors (left, top left, top? if available?) of neighbouring blocks. Depending on the prediction method used in said neighbouring blocks, the resultant B-frame prediction mode can be one of forward, backward or interpolation.		The '''spatial''' mode calculates motion vectors from the vectors (left, top left, top? if available?) of neighbouring blocks. Depending on the prediction method used in said neighbouring blocks, the resultant B-frame prediction mode can be one of forward, backward or interpolation.

Superdump: First edition.

2007-05-19T23:18:13Z

First edition.

New page

'''Motion compensation''' encompasses a number of methods for using predicted motion vectors to recreate pixels within a frame from past, future video frames or even the current video frame. Wikipedia has a [http://en.wikipedia.org/wiki/Motion_compensation fairly extensive article] discussing and detailing some of the defined techniques.

== Direct Prediction ==

Direct prediction offers methods for calculating motion vectors for already coded vectors to reduce the bit cost of coding B-frames and take advantage of motion coherences in different domains. There are two methods of direct prediction - spatial and temporal.

The '''spatial''' mode calculates motion vectors from the vectors (left, top left, top? if available?) of neighbouring blocks. Depending on the prediction method used in said neighbouring blocks, the resultant B-frame prediction mode can be one of forward, backward or interpolation.

'''Temporal''' direct prediction necessarily uses the interpolation mode. The temporal mode considers the forward motion vector of the collocated block in the first list 1 reference frame.

Insert mathematical theory here. :)

Consider the following setup where we have a sequence of frames PbbbP with no reference B-frames / frame reordering:

[[Image:H264_direct_temporal_1.png]]

To visualise the process, the end points of the forward motion vector remain in the same frames (starting at P2, ending at P1) but the vector is translated such that it appropriately intersects the current macroblock (in red) in the current frame (B1).

[[Image:H264_direct_temporal_2.png]]

The translated vector is essentially 'split' setting the start point at the appropriate position of the current macroblock and pointing along the translated vector forward to P1 and backward to P2.