Display management for high dynamic range images

1. A method for display management, the method comprising:

accessing a first image in a first dynamic range and a first spatial resolution, the first image comprising intensity pixels and chroma;

applying a tone-mapping function to the intensity pixels of the first image to generate corresponding intensity pixels (I′) of a tone-mapped image in a target dynamic range, wherein the target dynamic range is different than the first dynamic range;

computing a saturation metric for pixels of the first image based on values of the chroma pixels in the first image;

computing saturation-mapping-intensity (SMI) scaler based on corresponding pixel intensity values in the first image and the tone-mapped image, wherein computing the SMU scaler comprises computing:

SaturationScale

=

f

S

⁡

(

I

′

)

f

S

⁡

(

I

)

,

wherein SaturationScale denotes the SMI scaler, I denotes an intensity pixel value in the first image, I ′ denotes a corresponding intensity value in the tone-mapped image, and the f s (I) function comprises a monotonically increasing function of intensity;

computing a tone-mapping-saturation (TMS) scaler based on a first function of the saturation metric of the chroma pixels in the first image;

computing a saturation-mapping-saturation scaler (SMS) based on a second function of the saturation metric of the chroma pixels in the first image;

multiplying the TMS scaler with the intensity pixels of the tone-mapped image to generate output intensity pixels (I″);

multiplying the SMS and SMI scalers with the chroma pixels of the first image to generate intermediate chroma pixels; and

generating an output image comprising the output intensity pixels and output chroma pixels, wherein the output chroma pixels comprise the intermediate chroma pixels.

2. The method of claim 1 , wherein generating the TMS scaler ToneMapS comprises computing:

S =√{square root over ( T 2 +P 2 )},

ToneMap S =(1+ S 3 *c 1 )*(1+ S*c 2 ),

wherein S denotes the saturation metric for chroma pixel values T and P, c 1 is a variable computed based on differences between the first dynamic range and the target dynamic range, and C 2 is calculated based on metadata.

3. The method of claim 1 , wherein

SaturationScale

=

(

I

′

-

1

)

3

+

1

(

I

-

1

)

3

+

1

.

4. The method of claim 1 wherein generating the SMS scaler SaturationMapS comprises computing:

S =√{square root over ( T 2 +P 2 )},

ToneMap S =(1+ S 3 *c 1 )*(1+ S*c 2 ),

wherein S denotes the saturation metric for chroma pixel values T and P, c 1 is a variable computed based on differences between the first dynamic range and the target dynamic range, and c 2 is calculated based on metadata.

5. The method of claim 1 , wherein generating the intermediate chroma pixels comprises:

ct=T*SMI*SMS

cp=P*SMI*SMS

where T and P denote values of the chroma pixels of the first image, and ct and cp denote the intermediate chroma pixels.

6. The method of claim 1 , wherein generating the output chroma pixels further comprises:

receiving trim metadata as part of the first image;

receiving a region of interest in the first image;

for pixels in the first image belonging to the region of interest:

computing a hue value of the intermediate chroma pixels;

computing a saturation-map-hue (SMH) scaler based on the hue value and the trim metadata;

computing a hue-map value (r1) based on the hue value and the trim metadata; and

generating the output chroma pixels based on the hue-map value, the SMH scaler and the intermediate chroma pixels.

7. The method of claim 6 , wherein generating the output chroma pixels comprises computing:

r 2=1− r 1 2 /2;

ct=ct*SMH;

cp=cp*SMH;

ct′=r 2* ct−r 1* cp;

Cp′=r 1* ct+r 2* cp,

wherein r1 denotes the hue-map value, and Ct′ and Cp′ denote the output chroma pixels.

8. The method of claim 1 , wherein the intensity pixels of the first image are generated as

I B =α*I F +(1− a )* I,

wherein α comprises a value between 0 and 1, I denotes an intensity pixel value of an input image and I F denotes a corresponding intensity pixel value of a filtered base layer image generated using the input image, pyramid down-sampling, and pyramid up-sampling.

9. The method of claim 8 , wherein given a scaler 13 between 0 and 1, generating intensity pixel values of the output image comprises:

I′=I′ B +( I−I B )*β,

where I′ B denotes pixel intensity of a tone-mapped version of I B .

10. The method of claim 8 , wherein pyramid down-sampling generates a pyramid representation of the input image using N pyramid layer images, where N is at least two, wherein the first pyramid layer image has a lower spatial resolution than the spatial resolution of the input image, and the k-th layer pyramid image, for k=2, . . . , N, has a lower spatial resolution than the k-1-th layer pyramid image.

11. The method of claim 10 , wherein generating the intensity pixel values of the filtered base layer image comprises:

generating a set of edge-filter coefficient images Ima(1) and Imb(1) ( 710 ) based on the N pyramid layer images;

up-sampling the set of edge-filter coefficient images Ima(1) and Imb(1) to match the spatial resolution of the input image and generate a set of up-sampled edge-filter coefficient images ImaU(1) and ImbU(1) ( 712 ); and

generating the intensity pixel values of the filtered base layer as:

I F =ImaU (1)* I+ImbU (1).

12. The method of claim 11 , wherein generating Ima(1) and Imb(1) comprises:

generating Ima(N) and Imb(N) using an edge filter and P(N), wherein P(N) denotes the N-layer pyramid image with the lowest spatial resolution;

generating ImaU(N) and ImbU(N) by up-scaling Ima(N) and Imb(N) to match the spatial resolution of the (N−1) pyramid layer image;

for i=N−1 to 2:

generating S(i)=ImaU(i+1)*P(i)+ImbU(i+1), wherein P(i) denotes the i-th layer pyramid image;

generating Ima(i) and Imb(i) using the edge filter, S(i) and P(i); and

computing S(1)=ImaU(2)*P(1)+ImbU(2); and

generating Ima(1) and Imb(1) using the edge filter, S(1) and P(1).

13. The method of claim 12 , wherein generating Ima(i) and Imb(i) using the edge filter, and S(i) and P(i) comprises:

generating images P2(i)=P 2 (i) and PS(i)=P(i)*S(i);

convolve P2(i) with filter H to generate P2out;

convolve P(i) with the filter H to generate Pout;

convolve S(i) with the filter H to generate Sout;

convolve PS(i) with filter H to generate PSout;

compute:

P var= P 2out−Pout 2 ,

PS cov=PSout−Sout*Pout,

Ima ( i )= PS cov/( P var+ PW ( i )),

Imb ( i )=Sout− Ima ( i )*Pout,

wherein PW(i) denotes a set of weights received via metadata.

14. An apparatus comprising a processor and configured to perform the method recited in claim 1 .

15. A non-transitory computer-readable storage medium having stored thereon computer-executable instructions for executing a method with one or more processors in accordance with claim 1 .