System, devices and/or processes for adaptive image resolution scaling

1. A method comprising:

obtaining an image frame comprising an image of an object captured at an imaging device from a first signal path;

obtaining one or more signals from a second signal path separate from the first signal path;

identifying portions of the image frame obtained from the first signal path having diminished signal information based, at least in part, on one or more signals indicative of geometric distortions in the identified portions in the one or more signals obtained from the second signal path, the one or more signals indicative of geometric distortions being based, at least in part, on a model of a physical lens to capture the obtained image frame or a model of a computed target lens, or a combination thereof; and

reducing computations in processing of the at least one of the identified portions of the image frame obtained from the first signal path.

2. The method of claim 1 , wherein reducing computations in processing of the at least one of the identified portions further comprises reducing a spatial sampling in at least one of the identified portions.

3. The method of claim 1 , wherein reducing computations in processing of the at least one of the identified portions comprises reducing computations in a denoising operation, a banding removal operation or an adjustment to compression parameters, or a combination thereof.

4. The method of claim 1 , wherein the one or more signals obtained from the second signal path comprise auto-focus parameters, parameters indicative of a lens model, parameters indicative of a geometric distortion model or depth sensor parameters, or a combination thereof.

5. The method of claim 1 , wherein the model of a computed target lens based on application of a bokeh effect, or a combination thereof.

6. The method of claim 1 , wherein:

obtaining the image frame further comprises decoding a compressed signal formatted in blocks; and

identifying portions of the obtained image frame having diminished signal information comprises determining a highest spatial frequency associated with the blocks.

7. The method of claim 1 , wherein:

the obtained image frame is partitioned into tiles; and

identifying portions of the obtained image frame having diminished signal information comprises identifying tiles having the diminished signal information.

8. The method of claim 7 , and further comprises:

generating an image quality map based, at least in part on, application of a lens model an estimated distance of an object from focus; and

computing a scale factor for at least one of the tiles based, at least in part, on the image quality map.

9. The method of claim 1 , and further comprising:

reducing a spatial sampling in the at least one of the identified portions;

obtaining samples of the at least one of the identified portions according to the reduced spatial sampling; and

including the obtained samples in an activation input tensor to be applied to a neural network.

10. The method of claim 9 , and further comprising executing the neural network to apply a denoising operation, feature detection operation, feature classification operation or image enhancement operation, or a combination thereof, based, at least in part, on the activation input tensor.

11. An apparatus comprising:

an image capture device in a first signal path to obtain an image frame comprising an image of an object captured at the image capture device; and

one or more processors to:

obtain one or more signals from a second signal path separate from the first signal path;

identify portions of the image frame obtained from the first signal path having diminished signal information based, at least in part, on one or more signals obtained from the second signal path indicative of geometric distortions in the identified portions, the one or more signals indicative of geometric distortions to be based, at least in part, on a model of a physical lens to capture the image frame or a model of a computed target lens, or a combination thereof; and

reduce computations in processing of the at least one of the identified portions of the image frame obtained from the first signal path.

12. The apparatus of claim 11 , wherein the one or more processors are further to reduce computations in processing of the at least one of the identified portions based, at least in part, on a reduction in a spatial sampling in at least one of the identified portions.

13. The apparatus of claim 11 , wherein the signals indicative of limited spatial frequency comprise auto-focus parameters, parameters indicative of a lens model, parameters indicative of a geometric distortion model or depth sensor parameters, or a combination thereof.

14. The apparatus of claim 11 , wherein:

the image frame obtained from the first signal path is to be partitioned into tiles; and

identification of portions of the image having distortions comprises identifying tiles having the diminished signal information.

15. The apparatus of claim 11 , wherein the one or more processors are further to:

reduce a spatial sampling in the at least one of the identified portions;

obtain samples of the at least one of the identified portions according to the reduced spatial sampling; and

include the obtained samples in an activation input tensor to be applied to a neural network.

16. An article comprising:

a non-transitory storage medium comprising computer-readable instructions stored thereon which are executable by one or more processors of a computing device to:

obtain an image frame provided by an image capture device in a first signal path;

obtain one or more signals from a second signal path separate from the first signal path;

identify portions of the image frame obtained from the first signal path having diminished signal information based, at least in part, on the one or more signals obtained from the second signal path indicative of identifying geometric distortions in the identified portions, the one or more signals indicative of geometric distortions being based, at least in part, on a model of a physical lens to capture the obtained image frame or a model of a computed target lens, or a combination thereof; and

reduce computations in processing of the at least one of the identified portions of the image frame obtained from the first signal path.

17. The article of claim 16 , wherein the instructions are further executable by the one or more processors to reduce computations in processing of the at least one of the identified portions based, at least in part, on a reduction in a spatial sampling in at least one of the identified portions.

18. The article of claim 16 , wherein the signals indicative of limited spatial frequency comprise auto-focus parameters, parameters indicative of a lens model, parameters indicative of a geometric distortion model or depth sensor parameters, or a combination thereof.

19. The article of claim 16 , wherein the instructions are further executable by the one or more processors to:

obtain the image frame based, at least in part, on a decoded compressed signal formatted in blocks; and

identify portions of the obtained image frame having diminished signal information based, at least in part, on a determination of a highest spatial frequency associated with the blocks.

20. The article of claim 19 , wherein:

the obtained image frame is partitioned into tiles; and

the instructions are further executable by the one or more processors to:

identify portions of the obtained image having diminished signal information comprises identifying tiles having the diminished signal information;

generate an image quality map based, at least in part on, application of a lens model an estimated distance of an object from focus; and

compute a scale factor for at least one of the tiles based, at least in part, on the image quality map.