Photographic underexposure correction using a neural network
A method for image capture includes determining an exposure range and setting at least one camera parameter to capture an underexposed image outside the exposure range. The underexposed image is processed using a neural network to recover image details. Image defects due to camera or object motion blur can be reduced.
1. A method for image capture, the method comprising:
determining for a camera one or more camera parameters predicted to provide a correct exposure for a scene;
setting, on the camera based on the determining, at least one camera parameter of the one or more camera parameters to provide an underexposure for the scene;
capturing, by the camera due to the setting of the at least one camera parameter, an underexposed image of the scene; and
processing the underexposed image using a neural network alone to recover image details captured within the underexposed image.
2. The method of claim 1 , wherein the at least one camera parameter is aperture size.
3. The method of claim 1 , wherein the at least one camera parameter is image sensor sensitivity.
4. The method of claim 1 , wherein the at least one camera parameter is shutter speed.
5. The method of claim 1 , further comprising providing, by the underexposed image, motion or object blur reduction over what would be obtained in a correctly exposed image of the scene.
6. The method of claim 1 , further comprising providing, by the underexposed image, improved long range image capture over what would be obtained in a correctly exposed image of the scene.
7. The method of claim 1 , further comprising providing, by the underexposed image, frame rate increase over what would be obtained in a correctly exposed image of the scene.
8. The method of claim 1 , further comprising providing, by the underexposed image, improved burst mode capture over what would be obtained in a correctly exposed image of the scene.
9. The method of claim 1 , further comprising providing, by the underexposed image, super image stabilization over what would be obtained in a correctly exposed image of the scene.
10. The method of claim 1 , further comprising providing, by the underexposed image, increased dynamic range imaging over what would be obtained in a correctly exposed image of the scene.
11. The method of claim 1 , further comprising providing, by the underexposed image, light level reduction over what would be obtained in a correctly exposed image of the scene.
12. The method of claim 1 , further comprising providing, by the underexposed image, radiation reduction over what would be obtained in a correctly exposed image of the scene.
13. The method of claim 1 , wherein:
the processing the underexposed image using the neural network comprises generated an improved image derived from the underexposed image; and
the method further comprises digitally compressing the improved image.
14. The method of claim 1 , wherein the setting the at least one camera parameter comprises setting an f-stop two or more above what is predicted to provide a correct exposure for the scene.
15. The method of claim 1 , wherein the setting the at least one camera parameter comprises setting an ISO to one half or less of what is predicted to provide a correct exposure for the scene.
16. The method of claim 1 , wherein the setting the at least one camera parameter comprises setting a shutter speed to two or more times what is predicted to provide a correct exposure for the scene.
17. A method for image improvement, the method comprising:
determining analog settings for a sensor system having a sensor type, wherein the analog settings are predicted to provide an underexposure for a scene;
applying the analog settings to the sensor system;
capturing, by the sensor system due to the applying, an underexposed image of the scene; and
using a neural network trained on previous image data provided by the sensor type to alone reduce noise in and recover image details from the underexposed image and, thereby, produce an improved image of the scene derived from the underexposed image.
18. The method of claim 17 , further comprising providing, by the analog settings, an improved neural network training response over what would be obtained in connection with settings corresponding to a correctly exposed image of the scene.
19. The method of claim 17 , wherein the analog settings correspond to one or more of aperture size, shutter speed, and sensor sensitivity.
20. The method of claim 17 , further comprising digitally compressing the improved image.