Synthetic bracketing for exposure correction
Systems and methods are disclosed related to synthetic bracketing for exposure correction. A deep learning based method and system produces a set of differently exposed images from a single input image. The images in the set may be combined to produce an output image with improved global and local exposure compared with the input image. An image encoder applies learned parameters to each input image to generate a set of image features including local exposure estimates for each of two or more regions of the input image and a low resolution latent representation of the input image. A decoder receives the local exposure estimates, the latent representation, and target enhancements that are processed to generate synthesized transformations. When applied to the input image, the synthesized transformations produce the set of transformed images. Each transformed image is a version of the input image synthesized to correspond to a respective target enhancement.
1. A computer-implemented method, comprising:
encoding an input image into a latent representation for the input image and, for each of two or more regions of the input image, a respective local exposure estimate;
computing synthesized transformations using the local exposure estimates, latent representation, and target enhancement levels; and
applying the synthesized transformations to the input image to produce a set of exposure transformed images, wherein each exposure transformed image in the set is associated with a different one of the target enhancement levels.
2. The computer-implemented method of claim 1 , wherein the target enhancement levels are generated from the latent representation and the local exposure estimates.
3. The computer-implemented method of claim 1 , wherein the target enhancement levels are generated from the latent representation and enhancement signals provided by a user.
4. The computer-implemented method of claim 3 , wherein the enhancement signals include at least one of an indication of whether the input image is acquired with denoising on or off, a level of exposure correction to be applied, a brightness level control, a minimum luminance boosting amount, a noise reduction control, a color saturation control, a black point control, a local edge enhancement control, or a chroma control.
5. The computer-implemented method of claim 1 , further comprising combining the set of exposure transformed images to produce an output image with corrected exposure.
6. The computer-implemented method of claim 5 , wherein the input image is a frame in a video sequence, and further comprising repeating the encoding, computing, applying, and combining for additional input images included in the video sequence to produce additional output images with corrected exposures.
7. The computer-implemented method of claim 5 , wherein the combining is performed based on the target enhancement levels.
8. The computer-implemented method of claim 1 , wherein the encoding and computing are performed according to parameters that are learned using a training dataset comprising training input images and associated training target enhancements and ground truth sets of exposure transformed images.
9. The computer-implemented method of claim 8 , further comprising adjusting the parameters based on differences between the ground truth sets of exposure transformed images and sets of exposure transformed images.
10. The computer-implemented method of claim 9 , further comprising adjusting the parameters based on differences between a latent exposure channel included in the synthesized transformations and exposure and saturation of the training input images.
11. The computer-implemented method of claim 5 , wherein at least one of the steps of encoding, computing, or applying is performed on a server or in a data center to generate the output image, and the output image is streamed to a user device.
12. The computer-implemented method of claim 1 , wherein at least one of the steps of encoding, computing, or applying is performed within a cloud computing environment.
13. The computer-implemented method of claim 1 , wherein at least one of the steps of encoding, computing, or applying is performed for training, testing, or certifying a neural network employed in a machine, robot, or autonomous vehicle.
14. The computer-implemented method of claim 1 , wherein at least one of the steps of encoding, computing, or applying is performed on a virtual machine comprising a portion of a graphics processing unit.
15. A system, comprising:
a memory that stores an input image; and
a processor that is connected to the memory, wherein the processor is configured to:
encode the input image into a latent representation for the input image and, for each of two or more regions of the input image, a respective local exposure estimate;
compute synthesized transformations using the local exposure estimates, latent representation, and target enhancement levels; and
apply the synthesized transformations to the input image to produce a set of exposure transformed images, wherein each exposure transformed image in the set is associated with a different one of the target enhancement levels.
16. The system of claim 15 , wherein the target enhancement levels are generated from the latent representation and the local exposure estimates.
17. The system of claim 15 , wherein the target enhancement levels are generated from the latent representation and enhancement signals provided by a user.
18. The system of claim 15 , wherein the processor is further configured to combine the set of exposure transformed images to produce an output image with corrected exposure.
19. A non-transitory computer-readable media storing computer instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of:
encoding an input image into a latent representation for the input image and, for each of two or more regions of the input image, a respective local exposure estimate;
computing synthesized transformations using the local exposure estimates, latent representation, and target enhancement levels; and
applying the synthesized transformations to the input image to produce a set of exposure transformed images, wherein each exposure transformed image in the set is associated with a different one of the target enhancement levels.
20. The non-transitory computer-readable media of claim 19 , further comprising combining the set of exposure transformed images to produce an output image with corrected exposure.