Pulse or digital communications – Bandwidth reduction or expansion – Television or motion video signal
Reexamination Certificate
2000-11-09
2003-09-09
Le, Vu (Department: 2613)
Pulse or digital communications
Bandwidth reduction or expansion
Television or motion video signal
C348S699000
Reexamination Certificate
active
06618441
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Technical Field of the Invention
The present invention relates to a motion vector detection apparatus to be used in an apparatus for encoding dynamic images. In particular, the present invention relates to a motion vector detection apparatus for detecting a motion vector representing the movement direction and movement quantity of a dynamic image signal between dynamic image frames.
2. Description of the Prior Art
With development of computers and communication networks, digital processing of video signals is rapidly spread. For example, dynamic images are distributed to homes by using a communication satellite, and utilization of the digital image technique using the digital movie, DVD (digital versatile disc), or the like is spread rapidly.
Implementation of faster communication networks for transmitting image data greatly contributes to development of such a technique. Together therewith, advancement of an image compression technique such as the MPEG (moving picture experts group: ISO) cannot be overlooked. For example, in the MPEG, there is detected a motion vector indicating a position in the next frame to which an image included in a certain frame has moved, in order to compress a dynamic image between frames. As for an image part which does not move in successive frames, a previously detected image is used as it is. As for an image part having a motion between frames, such signal processing as to shift the image part by using a detected motion vector is conducted to increase the compression factor of the image. Under such circumstances, a technique for detecting motion vectors is attracting attention.
FIG. 12
shows an example of a relation between a frame and a macro block used in motion vector detection. In
FIG. 12
, a rectangle surrounded by an outer frame represents a frame
11
displayed on the screen at a certain time. In this example, the frame
11
is formed of 352 pixels by 240 lines. The frame
11
is divided into macro blocks
12
each formed of a rectangular region having 16 pixels by 16 pixels. In this case, the frame
11
is formed of 22×15 macro blocks
12
. Each macro block
12
becomes the unit of motion vector detection.
The reason why a region of some pixels is thus defined as the macro block
12
is that a region of such a degree that it can be recognized as a pattern of a certain degree is needed in order to judge the motion of the image within the framework. If the size of the macro block
12
is expanded more than needed, then there occur problems such as a problem that a region which is stationary between frames cannot be pick out efficiently. Furthermore, as the size of the macro block is made smaller, the quantity of information for comparison decreases and the data quantity of whole processing for detecting a motion vector increases. Under such circumstances, the macro block size becomes in many cases equal to approximately the above described pixel size.
FIG. 13
is a diagram showing the concept of the motion vector. It is now assumed that there are a first frame
11
1
and a second frame
11
2
which is temporally later than the first frame
11
1
, on a time axis (t). It is assumed that a dynamic image corresponding to an arbitrary macro block
12
1
shaded in the first frame
11
1
is judged to have moved from the same position
14
in the frame to a different position
15
in the second frame
11
2
obtained after the elapse of a predetermined time. In this case, a motion vector
16
can be represented as a connection of the position
14
before the movement to the position
15
after the movement in the same frame
11
2
.
FIG. 14
shows the principle of detecting such a motion vector. In the same way as the description given by referring to
FIG. 13
, it is now assumed that there are a first frame
11
1
and a second frame
11
2
which is temporally later than the first frame
11
1
. The second frame
11
2
is a picture which is now being subjected to processing for detecting a motion vector. The first frame
11
1
is a picture for which processing has already been finished. As for a macro block
12
2
in the position
15
of the second frame
11
2
for which a motion vector is to be detected, retrieval is effected to find a position of the first frame
11
1
having a macro block which resembles the macro block
12
2
most closely and thereby the motion vector is detected.
At this time, a concept referred to as a retrieval range
21
is introduced in order to reduce the burden of the search processing. Only in the retrieval range
21
, a macro block of a dynamic image corresponding to the macro block
12
2
is detected. This detection operation is conducted by scanning the image pattern in the same range as that of the macro block
12
2
as in TV raster scan, beginning from the left top corner and successively in the retrieval range
21
. A scan position which has caused the highest coincidence becomes a start point of the motion vector. For the purpose of calculation processing for judging the start point of the motion vector, a concept called vector evaluation value is introduced in some cases.
FIG. 15
is a diagram showing the concept of the vector evaluation value so as to correspond to FIG.
14
. In the retrieval range
21
shown in
FIG. 14
, an inspection framework
31
1
having the same size as that of the macro block
12
2
is disposed. And absolute values of differences respectively between 16 by 16 pixels included in the inspection framework
31
1
(see a right bottom part of
FIG. 12
) and 16 by 16 pixels forming the macro block
12
2
are derived, and they are added together. A resultant sum &Sgr;E
1
is adopted as the vector evaluation value for the inspection framework
31
1
. The reason why absolute values are derived is that the difference of signal levels of pixels is to be found. If the inspection framework
31
1
completely corresponds to the macro block
12
2
, then ideally the vector evaluation value &Sgr;E
1
becomes “0.”
If the vector evaluation value &Sgr;E
1
between the inspection framework
31
1
and the macro block
12
2
is thus derived, then an inspection framework
31
2
obtained by moving the inspection framework
31
1
in a direction of an arrow
32
shown in
FIG. 15
by one pixel is disposed in the same way. Between the inspection framework
31
2
and the macro block
12
2
, absolute values of differences respectively of 16 by 16 pixels are derived. The absolute values are added. A resultant vector evaluation value is &Sgr;E
2
. Then an inspection framework
31
3
obtained by further moving the inspection framework
31
2
in a direction of the arrow
32
by one pixel is disposed in the same way. Between the inspection framework
31
3
and the macro block
12
2
, absolute values of differences respectively of 16 by 16 pixels are derived. The absolute values are added. A resultant vector evaluation value is &Sgr;E
3
. Hereafter, all regions in the retrieval range
21
as described with reference to
FIG. 14
are thus scanned in the same way. Respective vector evaluation values &Sgr;E are thus derived. It is understood that such a position of the framework
31
in the retrieval range
21
that the sum of absolute values becomes minimum is a position which has become the start point of movement of the macro block
12
2
. By thus calculating the vector evaluation value in every position in the retrieval range
21
, comparing them with each other, and finding an inspection framework
31
having a minimum value, a motion vector in the position of the macro block
12
2
can be derived.
With respect to each position of macro blocks
12
of the current frame (the second frame
11
2
), a vector evaluation value having a minimum value between the macro block
12
and a past frame to be referred to (the first frame
11
1
), in the retrieval range
21
is thus derived. By doing so, a motion vector in each position of the macro block
12
of the current frame ((the second frame
11
2
) can be derived. For that purpose, however, massive calculation needs to be conducted between fr
Le Vu
NEC Corporation
Whitham Curtis & Christofferson, P.C.
LandOfFree
Motion vector detection apparatus does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Motion vector detection apparatus, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Motion vector detection apparatus will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3033543