System and method for scene-adaptive denoise scheduling and efficient deghosting
A method includes generating alignment maps for a first image frame having a first exposure level and a second image frame having a second exposure level different than the first exposure level. The method also includes generating, for the second image frame and a third image frame having a third exposure level different than the first and second exposure levels, shadow maps, saturation maps, and multi-exposure (ME) motion maps based on the alignment maps. The method further includes determining, based on the shadow maps, saturation maps, and ME motion maps, whether to perform machine learning-based denoising and, if so, on which image frame(s) to perform the machine learning-based denoising. In addition, the method includes updating at least one saturation map and at least one ME motion map for at least one of the second and third image frames depending on the image frame(s) on which the denoising is to be performed.
1. A method comprising:
generating alignment maps for a first image frame having a first exposure level and a second image frame having a second exposure level different than the first exposure level;
generating, for the second image frame and a third image frame having a third exposure level different than the first and second exposure levels, shadow maps, saturation maps, and multi-exposure (ME) motion maps based on the alignment maps;
determining, based on the shadow maps, the saturation maps, and the ME motion maps, whether to perform machine learning-based denoising and, if so, on which image frame or frames to perform the machine learning-based denoising; and
updating at least one of the saturation maps and at least one of the ME motion maps for at least one of the second and third image frames depending on the image frame or frames on which the machine learning-based denoising is to be performed.
2. The method of claim 1 , wherein:
the first exposure level is an EV-4 exposure level;
the second exposure level is an EV-2 exposure level; and
the third exposure level is an EV0 exposure level.
3. The method of claim 1 , further comprising:
performing the machine learning-based denoising on at least one of the first image frame, the second image frame, and the third image frame.
4. The method of claim 1 , further comprising:
performing a deghosting operation that differs based on whether or not the machine learning-based denoising is performed.
5. The method of claim 4 , wherein performing the deghosting operation comprises:
in response to determining that the machine learning-based denoising is not performed:
generating motion maps based on a single EV0 reference deghost image and multiple EV0 non-reference deghost images; and
performing a blending operation based on the motion maps; or
in response to determining that the machine learning-based denoising is performed:
generating motion maps based on image frames that are input to the machine learning-based denoising and multiple EV0 image frames; and
performing a blending operation based on the motion maps, the multiple EV0 image frames, and an output of the machine learning-based denoising.
6. The method of claim 5 , wherein the blending operation comprises a single-exposure (SE) blending operation followed by an ME blending operation.
7. The method of claim 1 , wherein generating the ME motion map for each of the second image frame and the third image frame comprises:
warping the first image frame and the second image frame;
converting the warped first image frame, the warped second image frame, and the third image frame to a luma-chroma (YUV) domain;
generating the ME motion map for the third image frame based on differences between the third image frame and the warped second image frame;
generating a blended image frame based on the third image frame and the warped second image frame; and
generating the ME motion map for the second image frame based on differences between the warped first image frame and the blended image frame.
8. An electronic device comprising:
at least one imaging sensor configured to capture multiple image frames of a scene; and
at least one processing device configured to:
generate alignment maps for a first image frame having a first exposure level and a second image frame having a second exposure level different than the first exposure level;
generate, for the second image frame and a third image frame having a third exposure level different than the first and second exposure levels, shadow maps, saturation maps, and multi-exposure (ME) motion maps based on the alignment maps;
determine, based on the shadow maps, the saturation maps, and the ME motion maps, whether to perform machine learning-based denoising and, if so, on which image frame or frames to perform the machine learning-based denoising; and
update at least one of the saturation maps and at least one of the ME motion maps for at least one of the second and third image frames depending on the image frame or frames on which the machine learning-based denoising is to be performed.
9. The electronic device of claim 8 , wherein:
the first exposure level is an EV-4 exposure level;
the second exposure level is an EV-2 exposure level; and
the third exposure level is an EV0 exposure level.
10. The electronic device of claim 8 , wherein the at least one processing device is further configured to perform the machine learning-based denoising on at least one of the first image frame, the second image frame, and the third image frame.
11. The electronic device of claim 8 , wherein the at least one processing device is further configured to perform a deghosting operation that differs based on whether or not the machine learning-based denoising is performed.
12. The electronic device of claim 11 , wherein, to perform the deghosting operation, the at least one processing device is configured to:
in response to determining that the machine learning-based denoising is not performed:
generate motion maps based on a single EV0 reference deghost image and multiple EV0 non-reference deghost images; and
perform a blending operation based on the motion maps; or
in response to determining that the machine learning-based denoising is performed:
generate motion maps based on image frames that are input to the machine learning-based denoising and multiple EV0 image frames; and
perform a blending operation based on the motion maps, the multiple EV0 image frames, and an output of the machine learning-based denoising.
13. The electronic device of claim 12 , wherein the blending operation comprises a single-exposure (SE) blending operation followed by an ME blending operation.
14. The electronic device of claim 8 , wherein, to generate the ME motion map for each of the second image frame and the third image frame, the at least one processing device is configured to:
warp the first image frame and the second image frame;
convert the warped first image frame, the warped second image frame, and the third image frame to a luma-chroma (YUV) domain;
generate the ME motion map for the third image frame based on differences between the third image frame and the warped second image frame;
generate a blended image frame based on the third image frame and the warped second image frame; and
generate the ME motion map for the second image frame based on differences between the warped first image frame and the blended image frame.
15. A non-transitory machine-readable medium containing instructions that when executed cause at least one processor of an electronic device to:
generate alignment maps for a first image frame having a first exposure level and a second image frame having a second exposure level different than the first exposure level;
generate, for the second image frame and a third image frame having a third exposure level different than the first and second exposure levels, shadow maps, saturation maps, and multi-exposure (ME) motion maps based on the alignment maps;
determine, based on the shadow maps, the saturation maps, and the ME motion maps, whether to perform machine learning-based denoising and, if so, on which image frame or frames to perform the machine learning-based denoising; and
update at least one of the saturation maps and at least one of the ME motion maps for at least one of the second and third image frames depending on the image frame or frames on which the machine learning-based denoising is to be performed.
16. The non-transitory machine-readable medium of claim 15 , wherein:
the first exposure level is an EV-4 exposure level;
the second exposure level is an EV-2 exposure level; and
the third exposure level is an EV0 exposure level.
17. The non-transitory machine-readable medium of claim 15 , wherein the instructions when executed further cause the at least one processor to perform the machine learning-based denoising on at least one of the first image frame, the second image frame, and the third image frame.
18. The non-transitory machine-readable medium of claim 15 , wherein the instructions when executed further cause the at least one processor to perform a deghosting operation that differs based on whether or not the machine learning-based denoising is performed.
19. The non-transitory machine-readable medium of claim 18 , wherein the instructions that when executed cause the at least one processor to perform the deghosting operation comprise instructions that when executed cause the at least one processor to:
in response to determining that the machine learning-based denoising is not performed:
generate motion maps based on a single EV0 reference deghost image and multiple EV0 non-reference deghost images; and
perform a blending operation based on the motion maps; or
in response to determining that the machine learning-based denoising is performed:
generate motion maps based on image frames that are input to the machine learning-based denoising and multiple EV0 image frames; and
perform a blending operation based on the motion maps, the multiple EV0 image frames, and an output of the machine learning-based denoising.
20. The non-transitory machine-readable medium of claim 19 , wherein the blending operation comprises a single-exposure (SE) blending operation followed by an ME blending operation.