System and method for recompressing a digital image using requantization
View Patent ↗A system and method for recompressing a JPEG image using requantization is described. DCT transform coefficients for a compressed JPEG image are statistically analyzed to generate a stored numeric error representation. Local minimizers are determined from the stored numeric error representation with selective rounding. A new quantization matrix Q 1 is selected from the local minimizers bounded by a target quantization matrix Q bound . The compressed JPEG image is recompressed using the new quantization matrix Q 1 .
1. A system for recompressing a digital image using requantization, wherein the digital image comprises compressed image data quantized per an initial quantization matrix, comprising:
a matrix module statistically analyzing DCT transform coefficients for the compressed digital image to generate a stored numeric error representation, determining local minimizers from the stored numeric error representation with selective rounding, and selecting a new quantization matrix Q 1 from the local minimizers bounded by a target quantization matrix Q bound ; and
an image recompression module recompressing the compressed digital image using the new quantization matrix Q 1 .
2. A method for recompressing a digital image using requantization, wherein the digital image comprises compressed image data quantized per an initial quantization matrix, comprising:
statistically analyzing DCT transform coefficients for the compressed digital image to generate a stored numeric error representation;
determining local minimizers from the stored numeric error representation with selective rounding;
selecting a new quantization matrix Q 1 from the local minimizers bounded by a target quantization matrix Q bound ; and
recompressing the compressed digital image using the new quantization matrix Q 1 .
3. A system for constructing a new quantization matrix for use in recompressing a digital image, wherein the digital image comprises compressed image data quantized per an initial quantization matrix, comprising:
an intermediate constant module determining an intermediate constant k computed in accordance with the equation:
k
=
⌊
q
bound
q
0
⌋
;
a matrix module defining a new quantization matrix Q 1 , comprising a plurality of quantizers q 1 , from an initial quantization matrix Q 0 , comprising a plurality of quantizers q 0 , each such quantizer q 1 computed in accordance with the equation:
q
1
=
{
q
0
,
if
k
=
0
;
kq
0
,
if
k
is
odd
;
kq
0
+
1
,
if
k
>
0
and
k
is
even
.
;
and
a recompression module recompressing a digital image using the new quantization matrix Q 1 .
4. A system according to claim 3 , further comprising:
a compression module compressing the digital image using a lossy digital image compression mode.
5. A system for constructing a new quantization matrix for use in recompressing a digital image with selective rounding, comprising:
an intermediate constant module determining an intermediate constant k computed in accordance with the equation:
k
=
⌊
q
bound
q
0
⌋
;
a matrix module defining a new quantization matrix Q 1 , comprising a plurality of quantizers q 1 , from an initial quantization matrix Q 0 , comprising a plurality of quantizers q 0 , each such quantizer q 1 computed in accordance with the equation:
q
1
=
{
q
0
,
if
k
=
0
;
kq
0
,
if
k
>
0.
;
and
a recompression module recompressing a digital image using the new quantization matrix Q 1 while rounding each half integer value towards zero.
6. A system according to claim 5 , further comprising:
a compression module compressing the digital image using a lossy digital image compression mode.
7. A system for recompressing a digital image using requantization, comprising:
a digital image stored as compressed image data quantized per an initial quantization matrix Q 0 comprising a plurality of initial quantizers q 0 ;
a quantization matrix generator defining a target quantization matrix Q bound comprising a plurality of target quantizers q bound , computing an intermediate constant k in accordance with the equation:
k
=
⌊
q
bound
q
0
⌋
,
and constructing a new quantization matrix Q 1 comprising a plurality of quantizers q 1 from the initial quantization matrix Q 0 by computing each such quantizer q 1 in accordance with the equation:
q
1
=
{
q
0
,
if
k
=
0
;
kq
0
,
if
k
is
odd
;
kq
0
+
1
,
if
k
>
0
and
k
is
even
.
;
and
an image recompression module requantizing the compressed image data using the new quantization matrix Q 1 .
8. A system according to claim 7 , wherein the digital image comprises a color image, further comprising:
a component separation submodule separating the color image into separate color components; and
a quantization submodule requantizing each such color component using the new quantization matrix Q 1 .
9. A system according to claim 7 , wherein the digital image comprises a grayscale image, further comprising:
a quantization submodule requantizing the grayscale image using the new quantization matrix Q 1 .
10. A system according to claim 7 , further comprising:
a partition submodule partitioning the digital image;
a transformation submodule transforming each partition using a DCT; and
a quantization submodule quantizing the transformed partitions using the quantization matrix Q 0 .
11. A system according to claim 7 , wherein the digital image is compressed using one of baseline sequential encoding and DCT-encoding.
12. A system for recompressing a digital image using requantization with selective rounding, comprising:
a digital image stored as compressed image data quantized per an initial quantization matrix Q 0 comprising a plurality of initial quantizers q 0 ;
a quantization matrix generator defining a target quantization matrix Q bound comprising a plurality of target quantizers q bound , computing an intermediate constant k in accordance with the equation:
k
=
⌊
q
bound
q
0
⌋
,
and constructing a new quantization matrix Q 1 comprising a plurality of quantizers q 1 from the initial quantization matrix Q 0 by computing each such quantizer q 1 in accordance with the equation:
q
1
=
{
q
0
,
if
k
=
0
;
kq
0
,
if
k
>
0.
;
(
3
)
an image recompression module requantizing the compressed image data using the new quantization matrix Q 1 by rounding each half integer value towards zero.
13. A system according to claim 12 , wherein the digital image comprises a color image, further comprising:
a component separation submodule separating the color image into separate color components; and
a quantization submodule requantizing each such color component using the new quantization matrix Q 1 .
14. A system according to claim 12 , wherein the digital image comprises a grayscale image, further comprising:
a quantization submodule requantizing the grayscale image using the new quantization matrix Q 1 .
15. A system according to claim 12 , further comprising:
a partition submodule partitioning the digital image;
a transformation submodule transforming each partition using a DCT; and
a quantization submodule quantizing the transformed partitions using the quantization matrix Q 0 .
16. A system according to claim 12 , wherein the digital image is compressed using one of baseline sequential encoding and DCT-encoding.