Prediction method for discrete cosine transform coefficient

Details Prediction method for discrete cosine transform coefficient Prediction method for discrete cosine transform coefficient

1997-09-19

2001-01-09

Mehta, Bhavesh (Department: 2721)

Image analysis

Image compression or coding

Transform coding

C382S233000, C382S238000, C382S251000

Reexamination Certificate

active

06173080

ABSTRACT:

BACKGROUND OF THE INVENTION
The present invention relates to a video signal encoding/decoding method, and more particularly, to a prediction method for a discrete cosine transform coefficient, for predicting DC/AC coefficients of a current block using location information of nonzero AC coefficients in a discrete cosine transform (DCT) domain.
A discrete cosine encoder encodes without loss of DC/AC coefficients obtained after quantization. Here, the DC/AC coefficients of the current block are predicted using the fact that adjacent blocks are similar to each other.
FIG. 1
is a diagram illustrating a general DC/AC coefficient prediction method. In order to predict the coefficients of the current block
101
which has been quantized after the DCT, coefficients of adjacent blocks
102
and
103
which are obtained through the same process are used. DC coefficient
104
of the current block
101
is predicted using the DC coefficient
107
of an upper block
102
or the DC coefficient
109
of a left block
103
. This is applied only for the prediction of the DC coefficient. An AC coefficient area
105
having the vertical characteristics of the current block
101
to be coded is predicted from the AC coefficients of an area
108
of the upper block
102
. This prediction method is profitable when the vertical components between blocks of an image are similar. An AC coefficient area
106
having the horizontal characteristics is predicted from the AC coefficients of an area
110
of the left block
103
. This prediction method is profitable when the horizontal components between blocks of the image are similar. The prediction of the DC/AC coefficients of the current block
101
is performed in the areas
104
,
105
and
106
.
FIG. 2
illustrates a general lossless encoding method of quantized coeficients in a DCT coding. In the block, a run-length coding is performed by zigzag scanning
201
.
FIG. 3
shows the structure of an encoder/decoder of DCT coefficients using a general DC/AC coefficient prediction method. Data
301
is converted into a DCT coefficient
303
by a discrete cosine transformer (DCT)
302
. The DCT coefficient
303
is quantized by a quantizer
304
. A quantized DCT coefficient
305
is input to a DC/AC coefficient predictor
306
for the prediction. Predicted coefficient
307
is encoded by a variable length code (VLC) encoder
308
and output as a bit stream
309
. The bit stream
309
is decoded by a VLC decoder
310
. Decoded coefficient
311
is compensated by a DC/AC coefficient compensator
312
, resulting in a compensated quantized coefficient
313
. The compensated quantized coefficient
313
is inversely quantized by an inverse quantizer
314
, and the obtained DCT coefficient
315
is inversely transformed by an inverse DCT (IDCT)
316
, resulting in final restored data
317
.
As described above, according to the general DC/AC coefficient prediction method, DC and AC coefficients in a horizontal or vertical area are predicted from the corresponding area of an adjacent block as shown in FIG.
1
. When performing a DCT on a block, a DC coefficient corresponds to the brightness of the block, which is similar among adjacent blocks. Also, in the case of the AC coefficient, the vertical and horizontal components of the block are reflected in AC coefficients in a DCT domain, so that the prediction can be achieved if the vertical and horizontal components of each adjacent block have similarity.
However, the general DC/AC coefficient prediction method uses the results of the DCT of the adjacent blocks regardless of their value, so that the improvement in the prediction performance is less than optimum. Also, when predicting an AC coefficient having a value of zero, there is a high probability that the predicted AC coefficient is not equal to zero, thus the prediction performance can not improve. A new algorithm is required for solving these problems.
SUMMARY OF THE INVENTION
To solve the above problem, it is an object of the present invention to provide a discrete cosine transform (DCT) coefficient prediction method for predicting DC/AC coefficients using location information of nonzero AC coefficients in a DCT domain.
It is another object of the present invention to provide a method for encoding DCT coefficient using the DCT coefficient prediction method.
It is still another object of the present invention to provide a DCT coefficient prediction method for predicting AC coefficients using projected values of adjacent blocks or weighted values.
To achieve the first object of the present invention, there is provided a discrete cosine transform (DCT) coefficient prediction method comprising the steps of:
scanning a vertical or horizontal component area of quantized AC coefficients in a DCT domain to obtain location information of all nonzero AC coefficients of a current block; and
predicting a DC coefficient and all nonzero AC coefficients in the vertical or horizontal component area, of the current block, using the location information and quantized DC and AC coefficients of the adjacent blocks.
To achieve the second object of the present invention, there is provided a discrete cosine transform (DCT) coefficient encoding method comprising the steps of:
scanning a vertical or horizontal component area of quantized AC coefficients in a DCT domain to obtain location information of all nonzero AC coefficients of a current block;
predicting a DC coefficient, and all nonzero AC coefficients in the vertical or horizontal component area, of the current block, using the location information and quantized DC and AC coefficients of the adjacent blocks; and
encoding the absolute difference between the DC coefficient of the current block and those of the adjacent blocks, and the absolute differences between the nonzero AC coefficients in the vertical or horizontal area of the current block, and those of the adjacent blocks, using the location information and the prediction coefficients, and applying a modified zigzag scanning to the remaining non-predicted AC coefficients for encoding.
According to an aspect of the third object, there is provided a discrete cosine transform (DCT) coefficient prediction method comprising the steps of:
(a) projecting pixel data, which are the closest to a current block in a spatial domain, in the horizontal or vertical direction; and
(b) performing DCT and quantization on the pixel data projected in the horizontal or vertical direction, and predicting DC and AC coefficients of the current block using the quantized DC and AC coefficients of the adjacent blocks, in a DCT domain.
According to another aspect of the third object, there is provided a discrete cosine transform (DCT) coefficient prediction method comprising the steps of:
(a) applying weighted values to pixel data of blocks adjacent to a current block in a spatial domain, wherein a smaller weight is applied to pixel values which are further from the current block; and
(b) performing DCT and quantization on the pixel data of the adjacent blocks to which the weighted values are applied, and predicting DC and AC coefficients of the current block using the quantized DC and AC coefficients of the adjacent blocks, in a DCT domain.
According to the present invention, the location information of the vertical and horizontal components of the AC coefficients is used for predicting AC coefficients. Here, the vertical and horizontal component areas of the AC coefficients, and the remaining coefficient area, are separately coded by using variable modified zigzag scanning. The performance of the AC coefficient prediction encoding is improved as the absolute values of the predicted coefficients are small, and as the number of coefficients of value zero increases. That is, because the AC coefficients in the vertical and horizontal component areas which are equal to
0
may be predicted to be nonzero, those AC coefficients of zero are excluded from the prediction.
In order to strengthen DC/AC coefficient prediction characteristics in the DCT domain, a large weighted value is applied to the pixel val

Affiliated with

Also associated with

Rating

Prediction method for discrete cosine transform coefficient does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Prediction method for discrete cosine transform coefficient, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Prediction method for discrete cosine transform coefficient will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-2492981