Motion vector detection device

Details Motion vector detection device Motion vector detection device

1998-12-21

2002-04-16

Kelley, Chris (Department: 2613)

Pulse or digital communications

Bandwidth reduction or expansion

Television or motion video signal

C375S240240, C382S236000

Reexamination Certificate

active

06373893

ABSTRACT:

BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a motion vector detecting device, especially to a motion vector detecting device suitable for use in the inter-frame coding of the motion compensation being the compression coding system of dynamic image information.
2. Description of the Related Art
Dynamic images include many similar images in two continuous screens. To reuse the previous images of these similar images in two continuous screens will make it possible to reduce the transmission of information to form the images. That is, the reuse of the previous images will achieve the compression coding of the dynamic images. To achieve this requires information indicating the location in the present screen which a part of the previous screen moves to (namely, a motion vector).
The detection of the motion vector generally employs the following method. First, the method gazes a. block consisting of M×N pixels (including M=N) of the present screen. Further, it gazes each of the blocks of M×N pixels contained in a (M+&agr;)×(N+&bgr;) pixel region (hereunder, referred to as a “reference region”, also including &agr;=&bgr;) which is wider than the block consisting of the M×N pixels of the previous screen (hereunder, referred to as the “reference screen”). And, it searches the most similar block to the gazed block of the present screen among the blocks of the M×N pixels contained in the reference region. Concretely, the sums of the differential absolute values are calculated between corresponding pixel data of the blocks of the present screen and the reference region. The block is searched in the reference region, which reduces the sum of the differential absolute values to the minimum. The motion vector is expressed by the difference (&Dgr;x, &Dgr;y) of the coordinates between the searched block and the gazed block of the present screen.
In practice, the detection of the motion vector is executed by a plurality of steps in order to enhance the processing efficiency or to increase the processing speed in many cases. The actual steps in an example are as follows.
First, the gazed block of the present screen and the reference region of the reference screen each are reduced, when a block of the present screen is compared with each of the blocks of the reference region.
The first method to reduce the foregoing block thins out every other pixel, for example, from the M×N pixels into a block of M/2×N/2 pixels as to the block of the present screen, and thins out every other pixel from the (M+&agr;)×(N+&bgr;) pixels into a block of (M+&agr;)/2×(N+&bgr;)/2 pixels as to the reference region. And, the second method obtains the averages of four pixels each, for example, from the M×N pixels to make a block of M/2×N/2 pixels as to the block of the present screen, and obtains the averages of the four pixels each from the (M+&agr;)×(N+&bgr;) pixels to make a block of (M+&agr;)/2×(N+&bgr;)/2 pixels as to the reference region.
Next, the second step gazes a part including the block that was searched in the first step, of the reference region not reduced; and it searches the most similar block to the block of the present screen from the part region gazed. Further, some cases need a third step and the succeeding steps (for example, the detection of the motion vector employing the reference region of the half pixel accuracy explained later).
However, the conventional motion vector detecting device involves the following problems in the improvement of the processing speed and/or the simplification of the device construction.
(1): The detection of the motion vector through a plurality of the foregoing steps has generally adopted the following procedure in the conventional technique.
FIG. 36
is a conceptual chart to explain the procedure.
A motion vector detecting device
11
contains an internal memory (also called as a local memory)
11
a
and a motion vector detecting unit
11
b.
The internal memory
11
a
is generally made up with an SRAM.
The image data of the present screen and the reference screen are stored in an external memory
13
provided outside the motion vector detecting device
11
. The external memory
13
is generally made up with a DRAM.
In the conventional method of the detection of the motion vector, the data of the M×N pixels (for example, 16×16 pixels) are inputted as they are from the external memory
13
into the internal memory
11
a,
as an information of the present screen. Further, the data of the (M+&agr;)×(N+&bgr;) pixels (for example, 48×48 pixels) are inputted as they are, as an information of the reference region. And, at the first step to execute the comparison between the reduced blocks, the motion vector detecting unit
11
b
reduces each of the blocks of the present screen and the reference region of the reference screen as mentioned above.
The reason that the M×N pixels as the information of the present screen and the (M+&agr;)×(N+&bgr;) pixels as the information of the reference region each have been inputted in the internal memory as they are mainly lies in the facility in the process of the second step and after, namely, the process of the comparison between the blocks not reduced.
However, this conventional method requires the storage capacity of at least M×N+(M+&agr;)×(N+&bgr;)+M/2×N/2+(M+&agr;)/2×(N+&bgr;)/2 for the internal memory
11
a,
and therefore, a memory of a large storage capacity has been necessary for the internal memory.
(2): Further, in the conventional method, the data of the blocks of the present screen and the data of the reference region of the reference screen each are stored in the internal memory
11
a
from the external memory
13
in the state that they are not reduced, and thereafter, they are reduced. Therefore, the conventional method requires a time to store the data and a time to reduce the data separately, and needs a longer processing time.
(3): Further, in the conventional method, every time when the motion vector of one block of the present screen is being detected, the data of the reference region corresponding to the one block is transferred to the internal memory
11
a.
However, the reference regions corresponding to the adjoining blocks of the present screen have an overlapping region in many cases. As shown in
FIG. 37
, suppose that the present screen
21
contains first and second blocks
21
a,
21
b,
each adjoining. The reference screen
23
contains first and second reference regions
23
a,
23
b
corresponding to the first and second blocks
21
a,
21
b.
These reference regions
23
a,
23
b
often contain overlapping image data
25
. In such a situation, the conventional method transfers this overlapping image data
25
to the internal memory
11
a
during detection of the motion vector of the first block
21
a
and during detection of the motion vector of the second block
21
b.
Accordingly, the conventional method requires a longer time to transfer the data.
(4): Further, the compression coding of dynamic images frequently employs inter-frame predictive coding. And generally, the judgment whether or not to execute the inter-frame predictive coding with regard to a block (also referred to as a gazed block) of the present screen is performed as follows. First, the sum of the differential absolute values between the average of all the pixel values of the gazed block and each of the pixel values of the gazed block is calculated (this is also referred to as a second sum of the differential absolute values). And, this second sum of the differential absolute values is compared with the sum of the differential absolute values (also referred to as a first sum of the differential absolute values) that was obtained for detecting the foregoing motion vector. From this comparison result, whether or not to execute the inter-frame predic

Affiliated with

Also associated with

Rating

Motion vector detection device 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 device, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Motion vector detection device will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-2820145