Methods and apparatus for video size conversion
View Patent ↗Methods and apparatus for performing 2:1 downscaling on video data are provided. At least one input matrix of N×N (e.g., N=16) Discrete Cosine Transform (DCT) coefficients is formed from the video data by combining four N/2×N/2 field-mode DCT blocks. Vertical downsampling and de-interlacing are performed to the input matrix to obtain two N/2×N/2 frame-mode DCT blocks. An N×N/2 input matrix is formed from the two frame-mode DCT blocks. Horizontal downsampling is performed to the N×N/2 matrix to obtain one N/2×N/2 frame-mode DCT block.
1. A method for performing 2:1 downscaling on video data, comprising the steps of:
forming at least one input matrix of N×N Discrete Cosine Transform (DCT) coefficients from the video data by combining four N/2×N/2 field-mode DCT blocks;
performing vertical downsampling and de-interlacing to the input matrix to obtain two N/2×N/2 frame-mode DCT blocks;
forming an N×N/2 input matrix from the two frame-mode DCT blocks; and
performing horizontal downsampling to the N×N/2 matrix to obtain one N/2×N/2 frame-mode DCT block.
2. The method of claim 1 , wherein N=16.
3. The method of claim 1 , wherein:
the vertical downsampling also achieves low pass filtering of the N×N input matrix.
4. The method of claim 1 , wherein:
the vertical downsampling uses a sparse downsampling matrix.
5. The method of claim 4 , wherein:
the sparse downsampling matrix=0.5[I 8 I 8 ], where I 8 is an 8×8 identity matrix.
6. The method of claim 1 , wherein:
the horizontal downsampling uses a sparse downsampling matrix composed of odd “O” and even “E” matrices.
7. The method of claim 6 , wherein:
the even matrix has the following form:
E
=
[
e
(
0
)
0
0
0
,
0
e
(
1
)
0
e
(
2
)
,
0
0
0
0
,
0
e
(
3
)
0
e
(
4
)
,
0
0
e
(
5
)
0
,
0
e
(
6
)
0
e
(
7
)
,
0
0
0
0
,
0
e
(
8
)
0
e
(
9
)
]
where e(1) through e(9) are non-zero coefficients; and
the odd matrix has the following form:
O
=
[
0
0
0
0
,
o
(
0
)
0
o
(
1
)
0
,
0
o
(
2
)
0
0
,
o
(
3
)
0
o
(
4
)
0
,
0
0
0
0
,
o
(
5
)
0
o
(
6
)
0
,
0
0
0
o
(
7
)
,
o
(
8
)
0
o
(
9
)
0
]
where o(1) through o(9) are non-zero coefficients.
8. An apparatus for performing 2:1 downscaling on video data, comprising:
means for forming at least one input matrix of N×N Discrete Cosine Transform (DCT) coefficients from the video data by combining four N/2×N/2 field-mode DCT blocks;
means for performing vertical downsampling and de-interlacing to the input matrix to obtain two N/2×N/2 frame-mode DCT blocks;
means for forming an N×N/2 input matrix from the two frame-mode DCT blocks; and
means for performing horizontal downsampling to the N×N/2 matrix to obtain one N/2×N/2 frame-mode DCT block.
9. The apparatus of claim 8 , wherein N=16.
10. The apparatus of claim 8 , wherein:
the means for performing vertical downsampling also achieves low pass. filtering of the N×N input matrix.
11. The apparatus of claim 8 , wherein:
the means for performing vertical downsampling uses a sparse downsampling matrix.
12. The apparatus of claim 11 , wherein:
the sparse downsampling matrix=0.5[I 8 I 8 ], where I 8 is an 8×8 identity matrix.
13. The apparatus of claim 8 , wherein:
the means for performing horizontal downsampling uses a sparse downsampling matrix composed of odd “O” and even “E” matrices.
14. The apparatus of claim 13 , wherein:
the even matrix has the following form:
E
=
[
e
(
0
)
0
0
0
,
0
e
(
1
)
0
e
(
2
)
,
0
0
0
0
,
0
e
(
3
)
0
e
(
4
)
,
0
0
e
(
5
)
0
,
0
e
(
6
)
0
e
(
7
)
,
0
0
0
0
,
0
e
(
8
)
0
e
(
9
)
]
where e(1) through e(9) are non-zero coefficients; and
the odd matrix has the following form:
O
=
[
0
0
0
0
,
o
(
0
)
0
o
(
1
)
0
,
0
o
(
2
)
0
0
,
o
(
3
)
0
o
(
4
)
0
,
0
0
0
0
,
o
(
5
)
0
o
(
6
)
0
,
0
0
0
o
(
7
)
,
o
(
8
)
0
o
(
9
)
0
]
where o(1) through o(9) are non-zero coefficients.