Scaling filter for video sharpening
A device has a single scaling filter to filter a video signal once to perform both sharpening and scaling. A memory stores original scaling filter coefficients for the scaling filter. An integrated circuit calculates new sharpening-scaling filter coefficients derived from the original scaling filter coefficients and one of sharpening filter coefficients for a sharpening filter and a sharpening strength and applies the new sharpening-scaling filter coefficients to the single scaling filter.
1. A device comprising:
a single scaling filter to filter a video signal once to perform both sharpening and scaling;
a memory to store original scaling filter coefficients for the scaling filter; and
an integrated circuit to calculate new sharpening-scaling filter coefficients derived from the original scaling filter coefficients and one of sharpening filter coefficients for a sharpening filter and a sharpening strength and to apply the new sharpening-scaling filter coefficients to the single scaling filter;
wherein the single scaling filter filters the video signal according to
z
q
=
∑
i
=
-
m
2
+
1
m
2
d
i
x
⌊
q
s
⌋
+
i
where x is the video signal; d i are the new sharpening-scaling filter coefficients; m denotes an even number of taps; s denotes the scaling ratio; q denotes a coordination index after scaling; and i is an index for a tap of the single scaling filter.
2. The device of claim 1 , wherein the single scaling filter comprises a polyphase m-tap finite impulse response (FIR) scaling filter.
3. The device of claim 2 , wherein the new sharpening-scaling filter coefficients are derived according to
[
d
-
1
d
0
d
1
d
2
]
=
[
h
0
+
h
-
1
b
1
h
-
1
0
h
1
a
4
h
1
+
h
-
1
b
2
h
0
h
-
1
h
1
a
3
h
-
1
b
3
h
1
h
0
h
-
1
+
h
1
a
2
h
-
1
b
4
0
h
1
h
0
+
h
1
a
1
]
[
c
-
1
c
0
c
1
c
2
]
where d i , −1≦i≦2, are the new sharpening-scaling filter coefficients; c i , −1≦i≦2, are the original scaling filter coefficients; h i , −1≦i≦1, are the sharpening filter coefficients; a i , 1≦i≦4, are forward prediction coefficients; and b i , 1≦i≦4, are backward prediction coefficients.
4. The device of claim 3 , wherein the polyphase m-tap finite impulse response (FIR) scaling filter has a plurality of phases.
5. The device of claim 4 , wherein the new sharpening-scaling filter coefficients are calculated according to a first order approximation of
[
d
p
,
-
1
d
p
,
0
d
p
,
1
d
p
,
2
]
=
(
[
α
+
256
-
α
0
0
-
α
2
α
+
256
-
α
0
0
-
α
2
α
+
256
-
α
0
0
-
α
α
+
256
]
[
c
p
,
-
1
c
p
,
0
c
p
,
1
c
p
,
2
]
+
[
128
128
128
128
]
)
>>
8
where
h
-
1
=
h
1
=
-
α
256
*
(
256
)
and
h
0
+
h
-
1
+
h
1
=
1.0
(
h
0
=
1
+
2
α
256
)
*
256
where α is the sharpening strength in a range of −127-+127; a 2 , a 3 and a 4 are set to zero; b 2 , b 3 and b 4 are set to zero; b 1 and a 1 are set to 1; c p,i , −1≦i≦2, represent the original scaling filter coefficients for a phase p; and d p,i , −1≦i≦2, represent the new sharpening-scaling filter coefficients for the phase p.
6. The device of claim 5 , wherein the integrated circuit expands the new sharpening-scaling filter coefficients from 10 bits to 11 bits.
7. The device of claim 5 , further comprising:
a sharpening strength adjuster to adaptively adjust the sharpening strength to prevent unwanted enhancement on artifacts.
8. The device of claim 7 , wherein the sharpening strength adjuster adaptively adjusts the sharpening strength to be reduced by a certain value if Qp is greater than a threshold according to
α=max(α 0 −k (max(0 ,Qp −τ)),α min )
where α min is a minimum sharpening strength; τ is a threshold determined by a distance to a last I frame and a codec type; k is a tunable constant; Qp is a quantization step; and α 0 is a default sharpening strength.
9. The device of claim 1 , wherein the new sharpening-scaling filter coefficients are calculated according to
[
d
-
m
2
+
1
d
-
m
2
+
2
d
-
m
2
+
3
⋮
d
0
d
1
⋮
d
m
2
]
=
[
h
0
+
h
-
1
h
-
1
h
-
2
⋯
h
-
m
2
+
1
0
⋯
0
h
1
h
0
h
-
1
⋯
h
-
m
2
+
2
h
-
m
2
+
1
⋯
0
h
2
h
1
h
0
⋯
h
-
m
2
+
3
h
-
m
2
+
2
⋯
0
⋮
⋮
⋮
⋮
⋮
⋮
⋮
⋮
h
m
2
-
1
h
m
2
-
2
h
m
2
-
3
⋯
h
0
h
-
1
⋯
0
0
h
m
2
-
1
h
m
2
-
2
⋯
h
1
h
0
⋯
h
-
m
2
+
1
⋮
⋮
⋮
⋮
⋮
⋮
⋮
⋮
0
0
0
⋯
0
h
m
2
-
1
⋯
h
0
+
h
1
]
[
c
-
m
2
+
1
c
-
m
2
+
2
c
-
m
2
+
3
⋮
c
0
c
1
⋮
c
m
2
]
where
d
i
,
-
m
2
+
1
≤
i
≤
m
2
,
are the new sharpening-scaling filter coefficients; h i are the sharpening filter coefficients; and
c
i
,
-
m
2
+
1
≤
i
≤
m
2
,
denotes the original scaling filter coefficients.
10. An integrated circuit comprising:
a single scaling filter to filter a video signal once to perform both sharpening and scaling;
a memory to store original scaling filter coefficients for the single scaling filter; and
a circuit to calculate new sharpening-scaling filter coefficients derived from the original scaling filter coefficients and one of sharpening filter coefficients for a sharpening filter and a sharpening strength and to apply the new sharpening-scaling filter coefficients to the single scaling filter;
wherein the single scaling filter filters the video signal according to
z
q
=
∑
i
=
-
m
2
+
1
m
2
d
i
x
⌊
q
s
⌋
+
i
where x is the video signal; d i are the new sharpening-scaling filter coefficients; m denotes an even number of taps; s denotes the scaling ratio; q denotes a coordination index after scaling; and i is an index for a tap of the single scaling filter.
11. The integrated circuit of claim 10 , wherein the single scaling filter comprises a polyphase m-tap finite impulse response (FIR) scaling filter.
12. The integrated circuit of claim 11 , wherein the new sharpening-scaling filter coefficients are derived according to
[
d
-
1
d
0
d
1
d
2
]
=
[
h
0
+
h
-
1
b
1
h
-
1
0
h
1
a
4
h
1
+
h
-
1
b
2
h
0
h
-
1
h
1
a
3
h
-
1
b
3
h
1
h
0
h
-
1
+
h
1
a
2
h
-
1
b
4
0
h
1
h
0
+
h
1
a
1
]
[
c
-
1
c
0
c
1
c
2
]
where d i , −1≦i≦2, are the new sharpening-scaling filter coefficients; c i , −1≦i≦2, are the original scaling filter coefficients; h i , −1≦i≦1, are the sharpening filter coefficients; a i , 1≦i≦4, are forward prediction coefficients; and b i , 1≦i≦4, are backward prediction coefficients.
13. The integrated circuit of claim 12 , wherein the polyphase m-tap finite impulse response (FIR) scaling filter has a plurality of phases.
14. The integrated circuit of claim 13 , wherein the new sharpening-scaling filter coefficients are calculated according to a first order approximation of
[
d
p
,
-
1
d
p
,
0
d
p
,
1
d
p
,
2
]
=
(
[
α
+
256
-
α
0
0
-
α
2
α
+
256
-
α
0
0
-
α
2
α
+
256
-
α
0
0
-
α
α
+
256
]
[
c
p
,
-
1
c
p
,
0
c
p
,
1
c
p
,
2
]
+
[
128
128
128
128
]
)
>>
8
where
h
-
1
=
h
1
=
-
α
256
*
(
256
)
and
h
0
+
h
-
1
+
h
1
=
1.0
(
h
0
=
1
+
2
α
256
)
*
256
where α is the sharpening strength in a range of −127-+127; a 2 , a 3 and a 4 are set to zero; b 2 , b 3 and b 4 are set to zero; b 1 and a 1 are set to 1; c p,i , −1≦i≦2, represent the original scaling filter coefficients for a phase p; and d p,i , −1≦i≦2, represent the new sharpening-scaling filter coefficients for the phase p.
15. The integrated circuit of claim 14 , wherein the circuit expands the new sharpening-scaling filter coefficients from 10 bits to 11 bits.
16. The integrated circuit of claim 14 , further comprising:
a sharpening strength adjuster to adaptively adjust the sharpening strength to prevent unwanted enhancement on artifacts.
17. The integrated circuit of claim 16 , wherein the sharpening strength adjuster adaptively adjusts the sharpening strength to be reduced by a certain value if Qp is greater than a threshold according to
α=max(α 0 −k (max(0 ,Qp −τ)),α mi )
where α min is a minimum sharpening strength; τ is a threshold determined by a distance to a last I frame and a codec type; k is a tunable constant; Qp is a quantization step; and a o is a default sharpening strength.
18. The integrated circuit of claim 10 , wherein the new sharpening-scaling filter coefficients are calculated according to
[
d
-
m
2
+
1
d
-
m
2
+
2
d
-
m
2
+
3
⋮
d
0
d
1
⋮
d
m
2
]
=
[
h
0
+
h
-
1
h
-
1
h
-
2
⋯
h
-
m
2
+
1
0
⋯
0
h
1
h
0
h
-
1
⋯
h
-
m
2
+
2
h
-
m
2
+
1
⋯
0
h
2
h
1
h
0
⋯
h
-
m
2
+
3
h
-
m
2
+
2
⋯
0
⋮
⋮
⋮
⋮
⋮
⋮
⋮
⋮
h
m
2
-
1
h
m
2
-
2
h
m
2
-
3
⋯
h
0
h
-
1
⋯
0
0
h
m
2
-
1
h
m
2
-
2
⋯
h
1
h
0
⋯
h
-
m
2
+
1
⋮
⋮
⋮
⋮
⋮
⋮
⋮
⋮
0
0
0
⋯
0
h
m
2
-
1
⋯
h
0
+
h
1
]
[
c
-
m
2
+
1
c
-
m
2
+
2
c
-
m
2
+
3
⋮
c
0
c
1
⋮
c
m
2
]
where
d
i
,
-
m
2
+
1
≤
i
≤
m
2
,
are the new sharpening-scaling filter coefficients; h i are the sharpening filter coefficients; and
c
i
,
-
m
2
+
1
≤
i
≤
m
2
,
denotes the original scaling filter coefficients.
19. A processor comprising:
a single scaling filter to filter a video signal once to perform both sharpening and scaling;
a memory to store original scaling filter coefficients for the scaling filter; and
an integrated circuit to calculate new sharpening-scaling filter coefficients derived from the original scaling filter coefficients and one of sharpening filter coefficients for a sharpening filter and a sharpening strength and to apply the new sharpening-scaling filter coefficients to the single scaling filter;
wherein the single scaling filter filters the video signal according to
z
q
=
∑
i
=
-
m
2
+
1
m
2
d
i
x
⌊
q
s
⌋
+
i
where x is the video signal; d i are the new sharpening-scaling filter coefficients; m denotes an even number of taps; s denotes the scaling ratio; q denotes a coordination index after scaling; and i is an index for a tap of the single scaling filter.
20. The processor of claim 19 , wherein the single scaling filter comprises a polyphase m-tap finite impulse response (FIR) scaling filter.
21. The processor of claim 20 , wherein the new sharpening-scaling filter coefficients are derived according to
[
d
-
1
d
0
d
1
d
2
]
=
[
h
0
+
h
-
1
b
1
h
-
1
0
h
1
a
4
h
1
+
h
-
1
b
2
h
0
h
-
1
h
1
a
3
h
-
1
b
3
h
1
h
0
h
-
1
+
h
1
a
2
h
-
1
b
4
0
h
1
h
0
+
h
1
a
1
]
[
c
-
1
c
0
c
1
c
2
]
where d i , −1≦i≦2, are the new sharpening-scaling filter coefficients; c i , −1≦i≦2, are the original scaling filter coefficients; h i , −1≦i≦1, are the sharpening filter coefficients; 1≦i≦4, are forward prediction coefficients; and b i , 1≦i≦4, are backward prediction coefficients.
22. The processor of claim 21 , wherein the polyphase m-tap finite impulse response (FIR) scaling filter has a plurality of phases.
23. The processor of claim 22 , wherein the new sharpening-scaling filter coefficients are calculated according to a first order approximation of
[
d
p
,
-
1
d
p
,
0
d
p
,
1
d
p
,
2
]
=
(
[
α
+
256
-
α
0
0
-
α
2
α
+
256
-
α
0
0
-
α
2
α
+
256
-
α
0
0
-
α
α
+
256
]
[
c
p
,
-
1
c
p
,
0
c
p
,
1
c
p
,
2
]
+
[
128
128
128
128
]
)
>>
8
where
h
-
1
=
h
1
=
-
α
256
*
(
256
)
and
h
0
+
h
-
1
+
h
1
=
1.0
(
h
0
=
1
+
2
α
256
)
*
256
where α is the sharpening strength in a range of −127-+127; a 2 , a 3 and a 4 are set to zero; b 2 , b 3 and b 4 are set to zero; b 1 and a 1 are set to 1; c p,i , −1≦i≦2, represent the original scaling filter coefficients for a phase p; and d p,i , −1≦i≦2, represent the new sharpening-scaling filter coefficients for the phase p.
24. The processor of claim 19 , wherein the new sharpening-scaling filter coefficients are calculated according to
[
d
-
m
2
+
1
d
-
m
2
+
2
d
-
m
2
+
3
⋮
d
0
d
1
⋮
d
m
2
]
=
[
h
0
+
h
-
1
h
-
1
h
-
2
⋯
h
-
m
2
+
1
0
⋯
0
h
1
h
0
h
-
1
⋯
h
-
m
2
+
2
h
-
m
2
+
1
⋯
0
h
2
h
1
h
0
⋯
h
-
m
2
+
3
h
-
m
2
+
2
⋯
0
⋮
⋮
⋮
⋮
⋮
⋮
⋮
⋮
h
m
2
-
1
h
m
2
-
2
h
m
2
-
3
⋯
h
0
h
-
1
⋯
0
0
h
m
2
-
1
h
m
2
-
2
⋯
h
1
h
0
⋯
h
-
m
2
+
1
⋮
⋮
⋮
⋮
⋮
⋮
⋮
⋮
0
0
0
⋯
0
h
m
2
-
1
⋯
h
0
+
h
1
]
[
c
-
m
2
+
1
c
-
m
2
+
2
c
-
m
2
+
3
⋮
c
0
c
1
⋮
c
m
2
]
where
d
i
,
-
m
2
+
1
≤
i
≤
m
2
,
are the new sharpening-scaling filter coefficients; h i are the sharpening filter coefficients; and
c
i
,
-
m
2
+
1
≤
i
≤
m
2
,
denotes the original scaling filter coefficients.
25. A wireless device comprising:
scaling filtering means for scaling filtering a video signal once to perform both sharpening and scaling;
storing means for storing original scaling filter coefficients for the scaling filtering means; and
calculating means for calculating new sharpening-scaling filter coefficients derived from the original scaling filter coefficients and one of sharpening filter coefficients for a sharpening filter and a sharpening strength and for applying the new sharpening-scaling filter coefficients to the scaling filtering means;
wherein the scaling filtering means comprises a single polyphase m-tap finite impulse response (FIR) scaling filter;
wherein the polyphase m-tap finite impulse response (FIR) scaling filter has a plurality of phases; and
wherein the new sharpening-scaling filter coefficients are calculated according to a first order approximation of
[
d
p
,
-
1
d
p
,
0
d
p
,
1
d
p
,
2
]
=
(
[
α
+
256
-
α
0
0
-
α
2
α
+
256
-
α
0
0
-
α
2
α
+
256
-
α
0
0
-
α
α
+
256
]
[
c
p
,
-
1
c
p
,
0
c
p
,
1
c
p
,
2
]
+
[
128
128
128
128
]
)
>>
8
where α is the sharpening strength in a range of −127-+127; c p,i , −1≦i≦2, represent the original scaling filter coefficients for a phase p; and d p,i , −1≦i≦2, represent the new sharpening-scaling filter coefficients for the phase p.