Deep SDR-HDR conversion

1. A method, comprising:

collecting multiple training images, wherein the training images include a training standard dynamic range (SDR) image and a training high dynamic range (HDR) image;

defining a set of color grading actions for use by a neural network; and

training the neural network to convert an input SDR image into an HDR image using the defined set of color grading actions and the training images, wherein the trained neural network is configured to convert the SDR image into the HDR image by applying one or more color grading actions from the set of color grading actions and generating an output that includes an indication of the one or more color grading actions;

wherein training the neural network includes generating an intermediate HDR image using the input SDR image, and measuring how close the intermediate HDR image is to the training HDR image.

2. The method of claim 1 , further comprising:

receiving a user input to modify the one or more color grading actions; and

generating a modified HDR image based on the user input.

3. The method of claim 1 , wherein training the neural network includes applying a first color grading action from the set of color grading actions to the training SDR image and wherein the first color grading action is selected based on the training HDR image.

4. The method of claim 1 , wherein the neural network is configured to extract contextual features or color features from the training SDR image.

5. The method of claim 1 , wherein the set of color grading actions includes at least one of adjusting brightness, adjusting contrast, adjusting color saturation or adjusting exposure.

6. The method of claim 1 , further comprising:

converting the SDR image into the HDR image, wherein converting includes applying the one or more color grading actions from the set of color grading actions.

7. The method of claim 6 , wherein converting further includes generating the output that includes the indication of the one or more color grading actions.

8. The method of claim 1 , wherein the training images further include a first training SDR image and a first training HDR image, the method further comprising:

generating a low dimensional latent representation of the first training SDR image;

generating a first HDR image by decoding the low dimensional latent representation of the first training SDR image; and

generating an SDR to HDR conversion model using the first HDR image, the first training SDR image and the first training HDR image, wherein the SDR to HDR conversion model is configured to convert a first input SDR image into a first output HDR image and trained to reconstruct a rolled off highlight in the first output HDR image that was not visible in the first input SDR image.

9. The method of claim 8 , wherein the rolled off highlight is reconstructed in a region of the first output HDR image that corresponds to a region of the first input SDR image that includes a clipped light level caused by a dynamic range of luminosity of the first input SDR image.

10. The method of claim 8 , wherein the first input SDR image uses a first dynamic range of luminosity and the first output HDR image uses a second dynamic range of luminosity, and

wherein the SDR to HDR conversion model is configured to convert the first input SDR image into the first output HDR image by mapping a first value defined relative to the first dynamic range of luminosity to a second value defined relative to the second dynamic range of luminosity.

11. The method of claim 8 , wherein the first input SDR image uses a first color gamut and the first output HDR image uses a second color gamut, and

wherein the SDR to HDR conversion model is configured to convert the first input SDR image into the first output HDR image by mapping a first value defined relative to the first color gamut to a second value defined relative to the second color gamut.

12. A method, comprising:

collecting multiple training images, wherein the training images include a training standard dynamic range (SDR) image and a training high dynamic range (HDR) image;

defining a set of color grading actions for use by a neural network; and

training the neural network to convert an input SDR image into an HDR image using the defined set of color grading actions and the training images, wherein the trained neural network is configured to convert the SDR image into the HDR image by applying one or more color grading actions from the set of color grading actions and generating an output that includes an indication of the one or more color grading actions;

wherein training the neural network to convert the input SDR image into the HDR image includes training the neural network to maximize a pixel level distance between the training SDR image and the training HDR image.

13. The method of claim 12 , further comprising:

generating an intermediate HDR image; and

measuring a similarity between the intermediate HDR image and the training HDR image.

14. The method of claim 12 , wherein training the neural network includes applying a first color grading action from the set of color grading actions to the training SDR image, and wherein the first color grading action is selected based on the training HDR image.

15. The method of claim 12 , further comprising:

converting the SDR image into the HDR image, wherein converting includes applying the one or more color grading actions from the set of color grading actions.

16. The method of claim 15 , wherein converting further includes generating the output that includes the indication of the one or more color grading actions.

17. The method of claim 12 , wherein the training images further include a first training SDR image and a first training HDR image, the method further comprising:

generating a low dimensional latent representation of the first training SDR image;

generating a first HDR image by decoding the low dimensional latent representation of the first training SDR image; and

generating an SDR to HDR conversion model using the first HDR image, the first training SDR image and the first training HDR image, wherein the SDR to HDR conversion model is configured to convert a first input SDR image into a first output HDR image and trained to reconstruct a rolled off highlight in the first output HDR image that was not visible in the first input SDR image.

18. The method of claim 17 , wherein the rolled off highlight is reconstructed in a region of the first output HDR image that corresponds to a region of the first input SDR image that includes a clipped light level caused by a dynamic range of luminosity of the first input SDR image.

19. The method of claim 17 , wherein the first input SDR image uses a first dynamic range of luminosity and the first output HDR image uses a second dynamic range of luminosity, and

wherein the SDR to HDR conversion model is configured to convert the first input SDR image into the first output HDR image by mapping a first value defined relative to the first dynamic range of luminosity to a second value defined relative to the second dynamic range of luminosity.

20. A method, comprising:

collecting multiple training images, wherein the training images include a training standard dynamic range (SDR) image and a training high dynamic range (HDR) image;

defining a set of color grading actions for use by a neural network; and

training the neural network to convert an input SDR image into an HDR image using the defined set of color grading actions and the training images, wherein the trained neural network is configured to convert the SDR image into the HDR image by applying one or more color grading actions from the set of color grading actions and generating an output that includes an indication of the one or more color grading actions;

wherein the neural network includes a fully-convolutional asynchronous advantage actor-critic (A3C) network architecture.

21. The method of claim 20 , wherein training the neural network includes applying a first color grading action from the set of color grading actions to the training SDR image, and wherein the first color grading action is selected based on the training HDR image.

22. The method of claim 20 , further comprising:

converting the SDR image into the HDR image, wherein converting includes applying the one or more color grading actions from the set of color grading actions.

23. The method of claim 22 , wherein converting further includes generating the output that includes the indication of the one or more color grading actions.