Pulse or digital communications – Bandwidth reduction or expansion – Television or motion video signal
Reexamination Certificate
1999-10-22
2004-07-20
Vo, Tung (Department: 2613)
Pulse or digital communications
Bandwidth reduction or expansion
Television or motion video signal
Reexamination Certificate
active
06765965
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a motion vector detecting apparatus, and more particularly to a motion vector detecting apparatus used in a digital motion picture compression system.
2. Description of the Background Art
To transmit a huge amount of image data efficiently at high speed, the image data are compressed to reduce the data amount to be transferred. One of such image data compression method is an MPEG method that deals with motion pictures. In this MPEG method, a motion vector is detected on a picture block basis according to a block matching method. This motion vector as well as a difference value between pixels in a current picture block and in a predictive picture block is transmitted. Since the difference value is used as the data to be transmitted, if the predictive picture block and the current picture block are well-consistent (i.e., if the degree of motion is small), the difference value is small, which leads to a reduced amount of data to be transmitted. Conversely, if the predicted picture block and the current picture block are poorly consistent, the data amount to be transmitted will increase.
As the block size decreases, precision in detection will increase, but resultant transmission data amount will increase. Typically, in the block matching method in the MPEG method, a macroblock having a size of 16 pixels by 16 pixels is used as a unit block for motion vector detection. In the MPEG method, picture data are transmitted on a frame basis, each frame consisting of a plurality of macroblocks.
FIG. 42
schematically shows a construction of a divided display screen according to the block matching method. In
FIG. 42
, one picture IG is divided into a plurality of macroblocks MB. By way of example,
FIG. 42
shows the picture that is divided into macroblocks of 4 rows by 5 columns. Macroblock MB generally has a size of 16 pixels by 16 pixels, and the number of macroblocks MB included in one picture IG is determined appropriately dependent on the number of pixels constituting picture IG.
FIG. 43
illustrates a motion vector detecting operation according to the block matching method. In
FIG. 43
, a search area SE having a prescribed size is determined for a current picture block CB of a target for the motion vector detection. This search area SE has a size that is predetermined both horizontally and vertically with its center being a position of macroblock (or, a template block) MB included in picture IG shown in FIG.
42
. In this search area SE, from a position (a real rear point) corresponding to the position of current picture block CB, a positional vector (i, j) of a reference picture block RB included in search area SE is calculated and determined as a motion vector candidate. The correlation between current picture block CB and reference picture block RB is obtained. Typically, a sum of absolute difference values or squared difference values between pixels at corresponding positions in current picture block CB and in reference picture block RB is obtained as an estimation value. This estimation value is obtained for every positional vector in search area SE, and the displacement vector of the reference picture block having the smallest estimation value is determined as a motion vector for the current picture block CB.
When pixel data of current picture block CB is expressed as “aij” and pixel data of reference picture block RB as “bij”, the estimation value is calculated from the following equation:
E=&Sgr;|aij−bij|,
or
E
=(&Sgr;
aij
2
−bij
2
)
Thus, a huge number of calculations are necessary to obtain the estimation value. Moreover, after obtaining the estimation value for every reference block RB in search area SE, the motion vector should be determined, for which another huge number of calculations are required. To perform this motion vector detection at high speed, a variety of operation algorithms have been proposed. Operation algorithms in the motion picture compression system according to the MPEG method are described in the following articles.
P. Pirsch et al., “VLSI Architecture for Video Compression—A Survey”, Proc. IEEE Vol. 83, No. 2, pp.220-246, 1995.
M. Yoshimoto et al., “ULSI Realization of MPEG
2
Realtime Video Encoder and Decoder—An Overview”, IEICE Trans. Electron., Vol. E78-C, No. 12, pp.1668-1681, 1995.
Tanaka et al., “MPEG2 Encoding LSIs About To Change Audiovisual Equipment For Household Use”, Nikkei Electronics, No. 711, Mar. 9, 1998.
LSIs for motion detecting operations designed to detect motion vectors at high speed are described in the following articles.
K. Ishihara et al., “A-Half-Pel Precision MPEG2 Motion Estimation Processor with Concurrent Three-Vector Search”, ISSCC Digest of Technical Papers, pp.288-289, 1995.
A. Ohtani et al., “A Motion Estimation Processor for MPEG2 Video Real Time Encoding at Wide Search Range”, Proc. IEEE Custom Integrated Circuits Conference, pp.405-408, 1995.
A. Hanami et al., “A 165-GOPS Motion Estimation Processor with Adaptive Dual-Array Architecture for High Quality Video-Encoding Applications”, Proc. IEEE Custom Integrated Circuits Conference, pp.169-172, 1998.
An all-sample full-search method enables the most accurate motion vector detection, in which method the above-described estimation values are calculated for all motion vector candidates within a search area. Specifically, an estimation value is obtained by calculating differences for all the pixel data in a picture block under search (i.e., a template block) and in a reference picture block. A positional vector of the reference picture block having the minimal estimation value among the estimation values estimated for all the estimation points in the search area is determined as a motion vector.
This all-sample full-search method, however, requires a huge amount of calculations and takes a long time to determine the motion vector. Thus, for high-speed motion vector detection through reduction of the computational amount, it is required to narrow the search area to reduce the number of estimation points. This means that, when a motion vector detecting apparatus is implemented with a single LSI, it is necessary to narrow a motion vector search range of this motion vector detecting apparatus of one LSI. Consequently, in order to search a motion vector over a wide range according to the all-sample full-search method, a plurality of LSIs (motion vector detecting apparatuses) should be operated in parallel. This leads to an increased number of LSIs to be used, resulting in an increase in power consumption as well as in apparatus scale.
To reduce the amount of computation in the all-sample full-search method, a variety of approaches have been proposed as follow: a sub-sampling approach, in which the differential operation is performed for only data of a part of pixels at each searching position (for each motion vector candidate, or at each estimation point); an algorithmic searching approach, in which the differential operation between a template block and a reference picture block is performed at only a part of coordinate positions in a search area according to a specific algorithm; and a combination of the sub-sampling approach and the algorithmic searching approach. A motion picture compression apparatus with a motion vector detecting circuit using the combined approach of sub-sampling and algorithmic searching is described, for example, in the following article.
M. Mizuno et al., “A 1.5 W Single-Chip MPEG2 MP@ML Encoder with Low-Power Motion Estimation and Clocking”, ISSCC Digest of Technical Papers, pp. 256-257, 1997.
In these motion vector searching approaches, however, a size of search area (motion vector search range) is fixed, and the motion vector is searched according to an algorithm that is statically determined independent of the characteristics of input current picture data. Therefore, such a situation may arise that, for some current picture data, it is possible to detect a motion vector having an estimation value substant
Hanami Atsuo
Ishihara Kazuya
Kumaki Satoshi
Matsumura Tetsuya
McDermott Will & Emery LLP
Renesas Technology Corp.
Vo Tung
LandOfFree
Motion vector detecting 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 detecting apparatus, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Motion vector detecting apparatus will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3204622