Image prediction based on primary color grading model

1. An image prediction method comprising:

receiving a first image v ( 310 ) and a second image s ( 320 ), wherein the second image s ( 320 ) has a different dynamic range than the first image v ( 310 );

receiving metadata ( 330 ) related to a color grading process, wherein the color grading process is used to derive the second image s ( 320 ) from the first image v ( 310 );

wherein the metadata ( 330 ) that relates to the color grading process comprise a color transformation matrix M and slope S, offset O, and power P parameters of a non-linear function;

applying said first image v ( 310 ) and said second image s ( 320 ) into an inverse color grading prediction process to compute an output image {circumflex over (v)} ( 340 ) that represents the predicted value of the first image v ( 310 );

wherein the inverse color grading prediction process comprises a non-linear prediction process which includes the slope S, offset O, and power P parameters and is related to the color transformation matrix M;

wherein the non-linear prediction process comprises applying a linear operator M −1 to the second image s ( 320 ) to generate intermediate values, and applying a non-linear function N −1 to the intermediate values to generate the output image {circumflex over (v)} ( 340 ) according to the following equation:

{circumflex over (v)}=N −1 ( sM −1 ),

wherein the linear operator M −1 comprises an inverse of the color transformation matrix M, and the non-linear function N −1 comprises the slope S, offset O, and power P parameters according to the following equation:

N

-

1

⁡

(

y

)

=

1

S

⁢

(

y

1

/

P

-

O

)

;

and

outputting the prediction parameters ( 350 ) of the inverse color grading process and the output image {circumflex over (v)} ( 340 ).

2. An image prediction method comprising:

receiving a first image v ( 310 ) and a second image s ( 320 ), wherein the second image s ( 320 ) has a different dynamic range than the first image v ( 310 );

applying said first image v ( 310 ) and said second image s ( 320 ) into a non-linear prediction process which includes slope S, offset O, and power P parameters of a non-linear function and an inverse M −1 of a color transformation matrix M to compute an output image {circumflex over (v)} ( 340 ) that represents the predicted value of the first image v ( 310 );

wherein the non-linear prediction process comprises applying a linear operator M −1 to the second image s ( 320 ) to generate intermediate values, and applying a non-linear function N −1 to the intermediate values to generate the output image {circumflex over (v)} ( 340 ) according to the following equation:

{circumflex over (v)}=N −1 ( sM −1 ),

wherein the non-linear function N −1 comprises parameters related to slope S, offset O, and power P parameters according to the following equation:

N

-

1

⁡

(

y

)

=

1

S

⁢

(

y

1

/

P

-

O

)

;

in response to the first image v ( 310 ) and the second image s ( 320 ), generating the prediction parameters of the model M −1 , S, O, and P by solving an optimization problem to minimize a prediction error ε to determine a parameter vector c according to the following equation:

min

c

⁢

ɛ

=



V

-

V

^



2

where

⁢

:

V

^

=

NS

-

1

⁡

(

c

,

S

)

,

NS −1 ( ) represents a non-linear function with inputs of the second image s ( 320 ) and the parameter vector c, and

the parameter vector c comprises the prediction parameters of the model M −1 , S, o, and P; and

outputting the prediction parameters of the model M −1 , S, O, and P and the output image {circumflex over (v)} ( 340 ).

3. The method of claim 1 or claim 2 wherein the first image v ( 310 ) comprises a visual dynamic range (VDR) image and the second image s ( 320 ) comprises a standard dynamic range (SDR) image.

4. The method of claim 2 , wherein the prediction process comprises:

computing a first result, wherein computing the first result comprises:

raising the result of a vector multiplication between a first vector and a second vector to the power of one over a j-th power parameter;

wherein the first vector represents the values of the color components of the i-th pixel from the second image s ( 320 ), and

wherein the second vector represents the j-th row of the color transformation matrix M; and

expressing each j-th color component of the i-th output pixel as the ratio of a numerator divided by the j-th slope parameter;

wherein the numerator comprises the j-th offset parameter subtracted from the first result.

5. The method of claim 4 wherein the output {circumflex over (v)} ( 340 ) has n total color components, wherein n comprises a positive integer number.

6. The method of claim 5 wherein n is equal to three (3).

7. The method of claim 1 , wherein the inverse color grading prediction process further comprises multiplying the color transformation matrix with the output of a color grading function s ( 320 ).

8. The method of claim 1 wherein the color grading process comprises a non-linear function having slope S, offset O, and power P parameters.

9. The method of claim 8 , wherein the output s ( 320 ) of the function is expressed in terms of slope, offset, and power parameters, and the input v ( 310 ).

10. The method of claim 9 wherein the output of the function comprises an intermediate sum raised to the power parameter P, wherein the intermediate sum comprises the offset parameter O added to a result of multiplying the input v ( 310 ) of the function with the slope parameter S.

11. An image decoding method comprising:

receiving a first image s ( 690 ) with a first dynamic range;

receiving metadata ( 645 ) related to a color grading process, wherein the color grading process is used to derive the first image s ( 690 ) from a second image that has a second dynamic range that is different than the first dynamic range;

wherein the metadata ( 645 ) that relates to the color grading process comprise a color transformation matrix M and slope S, offset O, and power P parameters of a non-linear function;

applying the first image s ( 690 ) and the metadata ( 645 ) into an inverse color grading prediction process ( 650 ) to compute an output image {circumflex over (v)} ( 680 ) that represents the predicted value of the second image;

wherein the inverse color grading prediction process ( 650 ) comprises a non-linear prediction process which includes the slope S, offset O, and power P parameters and is related to the color transformation matrix M;

wherein the non-linear prediction process comprises applying a linear operator M −1 to the first image s ( 690 ) to generate intermediate values, and applying a non-linear function N −1 to the intermediate values to generate the output image {circumflex over (v)} ( 680 ) according to the following equation:

{circumflex over (v)}=N −1 ( sM −1 ),

wherein the linear operator M −1 comprises an inverse of the color transformation matrix M, and the non-linear function N −1 comprises the slope S, offset O, and power P parameters according to the following equation:

N

-

1

⁡

(

y

)

=

1

S

⁢

(

y

1

/

P

-

O

)

.

12. The method of claim 11 wherein the first image s ( 690 ) comprises a standard dynamic range (SDR) image and the second image comprises a visual dynamic range (VDR) image.

13. The method of claim 11 , wherein the inverse color grading prediction process ( 650 ) comprises:

computing a first result, wherein computing the first result comprises:

raising the result of a vector multiplication between a first vector and a second vector to the power of one over a j-th power parameter;

wherein the first vector represents the values of the color components of the i-th pixel from the second image and

wherein the second vector represents the j-th row of a color transformation matrix; and

expressing each j-th color component of the i-th output pixel as the ratio of a numerator divided by the j-th slope parameter;

wherein the numerator comprises the j-th offset parameter subtracted from the first result.

14. An apparatus comprising a processor and configured to perform any one of the methods recited in claims 1 , 2 , or 11 .

15. A non-transitory computer-readable storage medium having stored thereon computer-executable instruction for executing a method using at least a processor in accordance with any one of the claims 1 , 2 , or 11 .