Machine learning of real-time image capture parameters

Patent metadata

Field	Value
Publication number	US-9826149-B2
Application number	US-201514670642-A
Country	US
Kind code	B2
Filing date	Mar 27, 2015
Priority date	Mar 27, 2015
Publication date	Nov 21, 2017
Grant date	Nov 21, 2017

Abstract

Methods, apparatuses and systems may provide for operating a machine learning device by obtaining training image data, conducting an offline prediction analysis of the training image data with respect to one or more real-time parameters of an image capture device, and generating one or more parameter detection models based on the offline prediction analysis. Additionally, methods, apparatuses and systems may provide for operating the image capture device by obtaining a candidate image associated with the image capture device, determining that the candidate image corresponds to a particular type of scene represented in a parameter prediction model, and adjusting one or more real-time parameters of the image capture device based at least in part on one or more parameter values associated with the particular type of scene.

First claim

We claim:

1. An image capture device comprising:
a sensor module to capture a candidate image including a default image characteristic level; and
a target override apparatus including:
a predictor to determine that the candidate image includes a scene that is similar to a particular type of scene represented in a parameter prediction model based on training image data, wherein the training image data is based on a reference image including an image characteristic target level for the particular type of scene; and
a parameter controller to adjust one or more real-time parameters of the image capture device based on the determination that the candidate image includes a scene that is similar to the particular type of scene and one or more parameter values associated with the image characteristic target level for the particular type of scene to generate an optimized image corresponding to the candidate image that includes the image characteristic target level.

2. The image capture device of claim 1 , wherein the predictor is to compute the one or more parameter values from one or more neural network nodes in the parameter prediction model.

3. The image capture device of claim 1 , wherein the one or more real-time parameters are to include one or more of an exposure control brightness parameter, a tone mapping brightness parameter, a focus parameter or a white balance parameter.

4. The image capture device of claim 1 , wherein the target override apparatus further includes a segmentation component to segment the candidate image into a plurality of regions based on luminance intensity.

5. The image capture device of claim 1 , wherein the target override apparatus further includes:
a feature calculator to identify a set of candidate features based on the candidate image; and
a feature selector to select a set of analysis features from the set of candidate features, wherein the one or more real-time parameters are to be adjusted further based on the set of analysis features.

6. The image capture device of claim 5 , wherein the predictor is to select the parameter prediction model from a plurality of parameter prediction models based on a level of correspondence between the set of analysis features and the selected parameter prediction model.

7. At least one non-transitory computer readable storage medium comprising a set of instructions which, when executed by an image capture device, cause the image capture device to:
obtain a candidate image associated with the image capture device including a default image characteristic level;
determine that the candidate image includes a scene that is similar to a particular type of scene represented in a parameter prediction model based on training image data, wherein the training image data is based on a reference image including an image characteristic target level for the particular type of scene; and
adjust one or more real-time parameters of the image capture device based on the determination that the candidate image includes a scene that is similar to the particular type of scene and one or more parameter values associated with the image characteristic target level for the particular type of scene to generate an optimized image corresponding to the candidate image that includes the image characteristic target level.

8. The at least one computer readable storage medium of claim 7 , wherein the instructions, when executed, cause the image capture device to compute the one or more parameter values from one or more neural network nodes in the parameter prediction model.

9. The at least one computer readable storage medium of claim 7 , wherein the one or more real-time parameters are to include one or more of an exposure control brightness parameter, a tone mapping brightness parameter, a focus parameter or a white balance parameter.

10. The at least one computer readable storage medium of claim 7 , wherein the instructions, when executed, cause the image capture device to segment the candidate image into a plurality of regions based on luminance intensity.

11. The at least one computer readable storage medium of claim 7 , wherein the instructions, when executed, cause the image capture device to:
identify a set of candidate features based on the candidate image; and
select a set of analysis features from the set of candidate features, wherein the one or more real-time parameters are to be adjusted further based on the set of analysis features.

12. The at least one computer readable storage medium of claim 11 , wherein the instructions, when executed, cause the image capture device to select the parameter prediction model from a plurality of parameter prediction models based on a level of correspondence between the set of analysis features and the selected parameter prediction model.

13. A method of operating a machine learning device, comprising:
obtaining training image data, wherein the training image data is based on a reference image including an image characteristic target level for a particular type of scene to be represented in a parameter prediction model;
conducting an offline prediction analysis of the training image data with respect to one or more real-time parameters of an image capture device; and
generating one or more parameter prediction models based on the offline prediction analysis, wherein a determination is to be made that a candidate image including a default image characteristic level includes a scene that is similar to the particular type of scene based on the training image data, and wherein an adjustment is to be made to the one or more real-time parameters of the image capture device based on the determination that the candidate image includes the scene that is similar to the particular type of scene and one or more parameter values associated with the image characteristic target level for the particular type of scene to generate an optimized image corresponding to the candidate image that includes the image characteristic target level.

14. The method of claim 13 , wherein each parameter prediction model includes a plurality of neural network nodes having parameter values.

15. The method of claim 13 , wherein the offline prediction analysis is conducted with respect to one or more of an exposure control brightness parameter, a tone mapping brightness parameter, a focus parameter or a white balance parameter.

16. The method of claim 13 , wherein the training image data includes reference images of particular types of scenes and candidate images associated with the image capture device.

17. The method of claim 13 , further including:
segmenting images in the training image data into a plurality of regions based on luminance intensity; and
extracting ground truth data from the training image data on a per-region basis, wherein the offline prediction analysis is conducted based at least in part on the ground truth data.

18. The method of claim 13 , further including:
identifying a set of candidate features based on the training image data; and
selecting a set of analysis features from the set of candidate features, wherein the offline prediction analysis is conducted based at least in part on the set of analysis features.

19. At least one computer readable storage medium comprising a set of instructions which, when executed by an machine learning device, cause the machine learning device to:
obtain training image data, wherein the training image data is based on a reference image including an image characteristic target level for a particular type of scene to be rep

Assignees

• Intel Corp

Inventors

• Chalom Edmond
• Rudoy Dmitry
• Levy Noam
• Essaky Sankaran Harish
• Nikkanen Jarno
• Maatta Joni-Matti
• Voronov German

Classifications

• G06V10/82 Primary

  using neural networks · CPC title

• G06V10/771

  Feature selection, e.g. selecting representative features from a multi-dimensional feature space · CPC title

• G06V10/764

  using classification, e.g. of video objects · CPC title

• H04N23/80

  Camera processing pipelines; Components thereof · CPC title

• G06F18/2113

  by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation · CPC title