Image analysis – Image compression or coding – Including details of decompression
Reissue Patent
1998-06-11
2002-09-24
Tran, Phuoc (Department: 2621)
Image analysis
Image compression or coding
Including details of decompression
C382S232000, C382S236000, C382S248000, C375S240160
Reissue Patent
active
RE037858
ABSTRACT:
BACKGROUND OF THE INVENTION
i) Field of the Invention
The present invention relates generally to image encoding systems and, more particularly, to motion compensation predicting coding systems.
ii) Description of the Prior Art
FIG. 1
shows a conventional motion compensation predicting coding apparatus, as disclosed in “A Study on HDTV Signal Coding with Motion Adaptive Noise Reduction” by S. Nogaki, M. Ohta and T. Omachi, The Third HDTV International Workshop Preliminary Report, Vol. 3, 1989. The conventional motion compensation predicting coding apparatus encodes a digital image signal
101
to produce a coded signal
106
that is sent over a transmission line
109
. The conventional motion compensation predicting coding apparatus includes a frame memory
1
, a motion detector
2
, a subtracter
3
, a coder
4
, a local decoder
5
, an adder
6
and a multiplexer (MUX)
7
. At a destination, the coded data is decoded to reproduce the original image signal
101
. The destination has an architecture similar to that shown in
FIG. 1
, except that the destination in configured to decode rather than encode.
Before discussing the operation of the conventional coding apparatus, it is helpful to review the format of the image data to be encoded. An image may be viewed as being made of a number of picture elements for pixels. Each of these pixels has a certain level of luminance that is encoded according to a grey scale or other scale. The luminance for each pixel may be encoded into a digital signal. The digital signals for an entire screen of the display form a frame. This frame may be broken down into rows of pixels. Since many video displays, such as conventional cathode ray tubes, scan odd numbered rows in a first pass and even numbered rows in a second pass when generating an image (i.e., interlaced scanning), it is helpful to view the frame as being composed of an even field having data for even numbered rows and an odd field having data for odd numbered rows. Each of the fields (i.e., the odd field and the even field) may be further divided into blocks of pixel data such as regions of 8 pixels by 8 pixels.
The conventional motion compensation predicting coding apparatus operates as follows. An input image signal
101
or, for example, a digital signal encoding the luminances of pixels in a frame is provided in accordance with an interlaced scanning format. In the interlaced scanning format, the odd rows of pixels are first scanned and then the even rows of pixels are scanned. Hence, the input signal provides a sequence of pixel data beginning with the data for all of the odd rows which is followed by the data for all of the even rows. For purposes of the present discussion, it is assumed that the input image signal
101
is already organized into blocks. Each block is made of pixels of the same field. The system operates on a block at a time until all the blocks of a given field are encoded. The system then proceeds to encode the blocks of the other field in a frame. Once a frame is fully encoded, the process of encoding is repeated with a next frame.
The input image signal
101
of the present frame is compared with the image signals for the same field in a preceding frame to carry out motion detection. For instance, suppose a first block in the input image signal
101
encodes luminance information for a block of pixels in the odd field of a present frame. The motion detection of this first block is performed in the motion detector
2
by searching for the most analogous block in the neighboring blocks
102
in the previous frame that are positioned around the corresponding position of the first block. The neighboring blocks
102
are read out of the frame memory
1
and provided to the motion detector
2
. The frame memory
1
, stores image data (of local decoded signals
108
) which has been locally decoded in the local decoder
5
and which has been summed with a motion compensation signal
104
in the adder
6
. The frame memory
1
may be realized as a conventional RAM.
The motion detector
2
determines the similarities between the present block and the respective neighboring blocks to select the most analogous neighboring blocks. As a yardstick of similarity, the motion detector
2
may calculate a different absolute value sum between blocks that is obtained by summing absolute values of each difference in luminance values for the corresponding pixels in the blocks, or the motion detector may calculate a differential square sum that is obtained by summing square values of differences in luminance values between the corresponding pixels in the blocks.
Once the most analogous neighboring block is found, the motion detector
2
calculates and outputs a motion vector
103
to the frame memory
1
and the MUX
7
. The motion vector indicates the vector displacement difference between the most analogous neighboring block and the first block. This motion vector
103
includes vector components in the horizontal and vertical directions. Also, once the most analogous neighboring block is found, a motion compensation prediction signal
104
that encodes the illuminance of the pixels of the most analogous neighboring block is read out of the frame memory
1
and sent to a subtracter
3
.
The subtracter
3
subtracts the motion compensation prediction signal
104
from the input image signal
101
to obtain a prediction error signal
105
. In particular, the illuminance of each pixel in the motion compensation prediction signal is subtracted from the illuminance of each corresponding pixel in the input image signal
101
. Chips for performing the subtraction are commercially available. The subtracter
3
then outputs the prediction error signal
105
to the coder
4
. The coder
4
executes the coding of the prediction error signal
105
to remove spatial redundancy in the signal. The coder
4
, thus, helps to compress the prediction error signal
105
. The signal
105
includes both low frequency components and high frequency components. Usually, in compressing the prediction error signal
105
, the coder
4
quantizes the low frequency components of the signal
105
using many bits and quantizes the high frequency components using few bits. More bits are used in the encoding of the low frequency components because the low frequency components typically have much grater power than do the high frequency components. Assigning more bits to the low frequency components enhances coding efficiency and reduces image degradation. The coder
4
, may perform, for example, an orthogonal transformation, such as a discrete cosine transformation (DCT), on an 8×8 pixel block to effect a frequency conversion that results in scalar quantization of a conversion factor. Scalar-quantized coded data
106
for the block is then sent from the coder
4
to the local decoder
5
and to the MUX
7
.
The MUX
7
not only multiplexes the coded data
106
and the motion vector
103
but also encodes the data in a format proper for sending the data down the transmission line
109
. The local decoder
5
performs a reciprocal operation to that performed by the coder
4
. In particular, an inverse scalar quantization or an inverse orthogonal transformation is carried out to obtain a decoded error signal
107
. The adder
6
adds the motion compensation predicting signal
104
to the decoded error signal
107
to obtain the local decoded signal
108
, which corresponds to the input image signal
101
. Chips for implementing the adder
6
are well known in the art. The local decoded signal
108
is stored in the frame memory
1
. The local decoded signal is subsequently read out of the frame memory
1
for use in carrying out the motion detection of the odd field in the next frame.
For the even field of the input image signal
101
, the motion detection in the motion detector
2
, the coding in the coder
4
, and so forth are carried out in the same manner as described above with respect to the odd field.
In the conventional motion compensation predicting coding apparatus, as described above, the removal
Asai Kohtaro
Kato Yoshiaki
Murakami Tokumichi
Yamada Yoshihisa
Mitsubishi Denki & Kabushiki Kaisha
Tran Phuoc
LandOfFree
Motion compensation predicting encoding method and 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 compensation predicting encoding method and apparatus, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Motion compensation predicting encoding method and apparatus will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2838194