Optimization of image encoding using perceptual weighting

1. A method of encoding transform domain coefficients for encoding an image, the method comprising:

iteratively,

quantizing the transform domain coefficients to generate quantized transform domain coefficients using quantization step sizes, and

entropy encoding the quantized transform domain coefficients using entropy encoding parameters,

wherein the quantization step sizes and the encoding parameters are selected to minimize a rate-distortion cost function,

until an incremental improvement in the cost is less than a threshold value,

and wherein the rate-distortion cost function includes a perceptual weighting factor applied to a quantization error, the perceptual weighting factor adjusting the rate-distortion cost function to apply greater weight to quantization error occurring in coefficients quantized by smaller quantization step sizes than to quantization error occurring at coefficients quantized by larger quantization step sizes.

2. The method claimed in claim 1 , wherein the perceptual weighting factor compensates for non-uniformity in the quantization step sizes within the rate-distortion cost function.

3. The method claimed in claim 1 , wherein the perceptual weighting factor is based on a ratio of a minimum quantization step size to a quantization step size used for a particular iteration.

4. The method claimed in claim 3 , wherein the quantization steps sizes are defined by a default quantization table, and wherein the perceptual weighting factor is given by the expression

w

i

=

qd

min

qd

i

wherein qd min is the smallest quantization step size in the default quantization table, and qd i is the i th quantization step size in the default quantization table.

5. The method claimed in claim 1 , wherein the encoding parameters comprise run-size pairs, and wherein the rate-distortion cost function includes an expression of cost attributable to quantization error relating to a particular run-size pair over a run r, expressed as

∑

j

=

i

-

r

i

-

1

⁢

w

j

2

⁢

C

j

2

+

w

i

2

⁢



C

i

-

q

i

·

K

i



2

wherein index i represents a current state within 64 possible states, index j represents a previous state prior to the run r, C i is the AC coefficient for the current state, q i is a quantization coefficient corresponding to C i , K i is a quantization index corresponding to C i , and w is the perceptual weighting factor corresponding to the like-indexed AC coefficient.

6. The method claimed in claim 1 , wherein the encoding parameters comprises run-level coding parameters to encode the quantized transform domain coefficients in a variable-length entropy encoding process.

7. The method claimed in claim 1 , wherein the encoding parameters comprise run-index pairs to encode the quantized transform domain coefficients in a variable-length entropy encoding process.

8. A system for encoding transform domain coefficients for encoding an image, the system comprising:

memory;

a processor; and

an encoding application that, when executed, configures the processor to:

iteratively,

quantize the transform domain coefficients to generate quantized transform domain coefficients using quantization step sizes, and

entropy encode the quantized transform domain coefficients using entropy encoding parameters,

wherein the quantization step sizes and the encoding parameters are selected to minimize a rate-distortion cost function,

until an incremental improvement in the cost is less than a threshold value,

and wherein the rate-distortion cost function includes a perceptual weighting factor applied to a quantization error, the perceptual weighting factor adjusting the rate-distortion cost function to apply greater weight to quantization error occurring in coefficients quantized by smaller quantization step sizes than to quantization error occurring at coefficients quantized by larger quantization step sizes.

9. The system claimed in claim 8 , wherein the perceptual weighting factor compensates for non-uniformity in the quantization step sizes within the rate-distortion cost function.

10. The system claimed in claim 8 , wherein the perceptual weighting factor is based on a ratio of a minimum quantization step size to a quantization step size used for a particular iteration.

11. The system claimed in claim 10 , wherein the quantization steps sizes are defined by a default quantization table, and wherein the perceptual weighting factor is given by the expression

w

i

=

qd

min

qd

i

wherein qd min is the smallest quantization step size in the default quantization table, and qd i is the i th quantization step size in the default quantization table.

12. The system claimed in claim 8 , wherein the encoding parameters comprise run-size pairs, and wherein the rate-distortion cost function includes an expression of cost attributable to quantization error relating to a particular run-size pair over a run r, expressed as

∑

j

=

i

-

r

i

-

1

⁢

w

j

2

⁢

C

j

2

+

w

i

2

⁢



C

i

-

q

i

·

K

i



2

wherein index i represents a current state within 64 possible states, index j represents a previous state prior to the run r, C i is the AC coefficient for the current state, q i is a quantization coefficient corresponding to C i , K i is a quantization index corresponding to C i , and w is the perceptual weighting factor corresponding to the like-indexed AC coefficient.

13. The system claimed in claim 8 , wherein the encoding parameters comprises run-level coding parameters to encode the quantized transform domain coefficients in a variable-length entropy encoding process.

14. The system claimed in claim 8 , wherein the encoding parameters comprise run-index pairs to encode the quantized transform domain coefficients in a variable-length entropy encoding process.

15. A non-transitory computer-readable medium storing computer-readable instructions for encoding transform domain coefficients for encoding an image, the computer-readable instructions comprising:

instructions for iteratively,

quantizing the transform domain coefficients to generate quantized transform domain coefficients using quantization step sizes, and

entropy encoding the quantized transform domain coefficients using entropy encoding parameters,

wherein the quantization step sizes and the encoding parameters are selected to minimize a rate-distortion cost function,

until an incremental improvement in the cost is less than a threshold value,

and wherein the rate-distortion cost function includes a perceptual weighting factor applied to a quantization error, the perceptual weighting factor adjusting the rate-distortion cost function to apply greater weight to quantization error occurring in coefficients quantized by smaller quantization step sizes than to quantization error occurring at coefficients quantized by larger quantization step sizes.

16. The non-transitory computer-readable medium claimed in claim 15 , wherein the perceptual weighting factor compensates for non-uniformity in the quantization step sizes within the rate-distortion cost function.

17. The non-transitory computer-readable medium claimed in claim 15 , wherein the perceptual weighting factor is based on a ratio of a minimum quantization step size to a quantization step size used for a particular iteration.

18. The non-transitory computer-readable medium claimed in claim 17 , wherein the quantization steps sizes are defined by a default quantization table, and wherein the perceptual weighting factor is given by the expression

w

i

=

qd

min

qd

i

wherein qd min is the smallest quantization step size in the default quantization table, and qd i is the i th quantization step size in the default quantization table.

19. The non-transitory computer-readable medium claimed in claim 15 , wherein the encoding parameters comprise run-size pairs, and wherein the rate-distortion cost function includes an expression of cost attributable to quantization error relating to a particular run-size pair over a run r, expressed as

∑

j

=

i

-

r

i

-

1

⁢

w

j

2

⁢

C

j

2

+

w

i

2

⁢



C

i

-

q

i

·

K

i



2

wherein index i represents a current state within 64 possible states, index j represents a previous state prior to the run r, C i is the AC coefficient for the current state, q i is a quantization coefficient corresponding to C i , K i is a quantization index corresponding to C i , and w is the perceptual weighting factor corresponding to the like-indexed AC coefficient.

20. The non-transitory computer-readable medium claimed in claim 15 , wherein the encoding parameters comprise run-level coding parameters to encode the quantized transform domain coefficients in a variable-length entropy encoding process.