Systems and methods for dynamic calibration of array cameras

• US 10,250,871 B2
• Filed: 09/29/2015
• Issued: 04/02/2019
• Est. Priority Date: 09/29/2014
• Status: Active Grant

• Current Assignee: FotoNation Limited (Adeia Inc.)
• Original Assignee: FotoNation Limited (Adeia Inc.)
• Inventors: Ciurea, Florian, Lelescu, Dan, Chatterjee, Priyam
• Primary Examiner(s): Ye, Lin
• Assistant Examiner(s): Nguyen, Chan T
• Application Number:
  US15/515,489
• Publication Number:
  US 20170244960A1
• Time in Patent Office: 1,281 Days
• Field of Search: 3482181
• CPC Class Codes: G06T 2207/10012 Stereo images
• CPC Class Codes: G06T 2207/10052 Images from lightfield camera
• CPC Class Codes: G06T 5/80 Geometric correction
• CPC Class Codes: G06T 7/80 Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
• CPC Class Codes: H04N 17/002 for television cameras
• CPC Class Codes: H04N 23/90 Arrangement of cameras or camera modules, e.g. multiple cameras in TV studios or sports stadiums

Abstract

Systems and methods for dynamically calibrating an array camera to accommodate variations in geometry that can occur throughout its operational life are disclosed. The dynamic calibration processes can include acquiring a set of images of a scene and identifying corresponding features within the images. Geometric calibration data can be used to rectify the images and determine residual vectors for the geometric calibration data at locations where corresponding features are observed. The residual vectors can then be used to determine updated geometric calibration data for the camera array. In several embodiments, the residual vectors are used to generate a residual vector calibration data field that updates the geometric calibration data. In many embodiments, the residual vectors are used to select a set of geometric calibration from amongst a number of different sets of geometric calibration data that is the best fit for the current geometry of the camera array.

Claims

1. A method of dynamically generating geometric calibration data for an array of cameras, comprising:

acquiring a set of images of a scene using a plurality of cameras, where the set of images comprises a reference image and an alternate view image;

detecting features in the set of images using a processor directed by an image processing application;

identifying within the alternate view image features corresponding to features detected within the reference image using a processor directed by an image processing application;

rectifying the set of images based upon a set of geometric calibration data using a processor directed by an image processing application;

determining residual vectors for geometric calibration data at locations where features are observed within the alternate view image based upon observed shifts in locations of features identified as corresponding in the reference image and the alternate view image using a processor directed by an image processing application;

determining updated geometric calibration data for a camera that captured the alternate view image based upon the residual vectors using a processor directed by an image processing application, wherein determining updated geometric calibration data comprises;

using at least an interpolation process to generate a residual vector calibration field from the residual vectors;

mapping the residual vector calibration field to a set of basis vectors; and

generating a denoised residual vector calibration field using a linear combination of less than the complete set of basis vectors; and

rectifying an image captured by the camera that captured the alternate view image based upon the updated geometric calibration data using a processor directed by an image processing application.

2. The method of claim 1, wherein determining residual vectors for geometric calibration data at locations where features are observed within the alternate view image comprises:

estimating depths of features within the alternate view image identified as corresponding to features detected within the reference image based upon components of the observed shifts in locations of features identified as corresponding in the reference image and the alternate view image along epipolar lines;

determining scene dependent geometric corrections to apply to the observed shifts in locations of features identified as corresponding in the reference image and the alternate view image based upon the estimated depths of the corresponding features; and

applying the scene dependent geometric corrections to the observed shifts in locations of features identified as corresponding in the reference image and the alternate view image to obtain residual vectors for geometric calibration data at locations where features are observed within the alternate view image.

3. The method of claim 1, wherein determining updated geometric calibration data for a camera that captured the alternate view image based upon the residual vectors further comprises using an extrapolation process in the generation of the residual vector calibration field from the residual vectors.

4. The method of claim 1, further comprising applying the residual vector calibration field to the set of geometric calibration data with respect to the camera that captured the alternate view image.

5. The method of claim 1, wherein the set of basis vectors is learned from a training data set of residual vector calibration fields.

6. The method of claim 5, wherein the set of basis vectors is learned from a training data set of residual vector calibration fields using Principal Component Analysis.

7. The method of claim 1, wherein determining updated geometric calibration data for a camera that captured the alternate view image further comprises selecting an updated set of geometric calibration data from amongst a plurality of sets of geometric calibration data based upon at least the residual vectors for geometric calibration data at locations where features are observed within the alternate view image.

8. The method of claim 1, further comprising:

acquiring an additional set of images of a scene using the plurality of cameras; and

determining residual vectors for the geometric calibration data using the additional set of images;

wherein determining updated geometric calibration data for a camera that captured the alternate view image based upon the residual vectors also comprises utilizing the residual vectors for the geometric calibration data determined using the additional set of images.

9. The method of claim 8, further comprising:

detecting at least one region within a field of view of a camera that does not satisfy a feature density threshold;

wherein the additional set of images of a scene is acquired in response to detecting that at least one region within a field of view of a camera does not satisfy the feature density threshold.

10. The method of claim 9, wherein utilizing the residual vectors determined using the additional set of images further comprises utilizing the residual vectors determined using the additional set of images to determine updated geometric calibration data with respect to the at least one region within the field of view of the camera in which the density threshold was not satisfied.

11. The method of claim 10, further comprising providing prompts via a user interface using a processor directed by an image processing application, where the prompts direct orientation of the camera array to shift locations of features identified as corresponding in the reference image and the alternate view image into the at least one region within the field of view of a camera that does not satisfy a feature density threshold during acquisition of the additional set of images.

12. A camera array, comprising:

at least one array of cameras comprising a plurality of cameras;

a processor; and

memory containing an image processing application;

wherein the image processing application directs the processor to;

acquire a set of images of a scene using the plurality of cameras, where the set of images comprises a reference image and an alternate view image;

detect features in the set of images;

identify within the alternate view image features corresponding to features detected within the reference image;

rectify the set of images based upon a set of geometric calibration data;

determine residual vectors for geometric calibration data at locations where features are observed within the alternate view image based upon observed shifts in locations of features identified as corresponding in the reference image and the alternate view image;

determine updated geometric calibration data for a camera that captured the alternate view image based upon the residual vectors, wherein determining updated geometric calibration data comprises;

using at least an interpolation process to generate a residual vector calibration field from the residual vectors;

mapping the residual vector calibration field to a set of basis vectors; and

generating a denoised residual vector calibration field using a linear combination of less than the complete set of basis vectors; and

rectify an image captured by the camera that captured the alternate view image based upon the updated geometric calibration data.

13. The camera array of claim 12, wherein determining residual vectors for geometric calibration data at locations where features are observed within the alternate view image comprises:

estimating depths of features within the alternate view image identified as corresponding to features detected within the reference image based upon components of the observed shifts in locations of features identified as corresponding in the reference image and the alternate view image along epipolar lines;

determining scene dependent geometric corrections to apply to the observed shifts in locations of features identified as corresponding in the reference image and the alternate view image based upon the estimated depths of the corresponding features; and

applying the scene dependent geometric corrections to the observed shifts in locations of features identified as corresponding in the reference image and the alternate view image to obtain residual vectors for geometric calibration data at locations where features are observed within the alternate view image.

14. The camera array of claim 12, wherein determining updated geometric calibration data for a camera that captured the alternate view image based upon the residual vectors further comprises using an extrapolation process in the generation of the residual vector calibration field from the residual vectors.

15. The camera array of claim 12, wherein the image processing application further directs the processor to apply the residual vector calibration field to the set of geometric calibration data with respect to the camera that captured the alternate view image.

16. The camera array of claim 12, wherein the set of basis vectors is learned from a training data set of residual vector calibration fields.

17. The camera array of claim 16, wherein the set of basis vectors is learned from a training data set of residual vector calibration fields using Principal Component Analysis.

18. The camera array of claim 12, wherein determining updated geometric calibration data for a camera that captured the alternate view image further comprises selecting an updated set of geometric calibration data from amongst a plurality of sets of geometric calibration data based upon at least the residual vectors for geometric calibration data at locations where features are observed within the alternate view image.

19. The camera array of claim 12, wherein the image processing application further directs the processor to:

acquire an additional set of images of a scene using the plurality of cameras; and

determine residual vectors for the geometric calibration data using the additional set of images;

wherein determining updated geometric calibration data for a camera that captured the alternate view image based upon the residual vectors also comprises utilizing the residual vectors for the geometric calibration data determined using the additional set of images.

20. The camera array of claim 19, wherein the image processing application further directs the processor to:

detect at least one region within a field of view of a camera that does not satisfy a feature density threshold;

wherein the additional set of images of a scene is acquired in response to detecting that at least one region within a field of view of a camera does not satisfy the feature density threshold.

21. The camera array of claim 20, wherein utilizing the residual vectors determined using the additional set of images further comprises utilizing the residual vectors determined using the additional set of images to determine updated geometric calibration data with respect to the at least one region within the field of view of the camera in which the density threshold was not satisfied.

22. A method of dynamically generating geometric calibration data for an array of cameras, comprising:

acquiring a set of images of a scene using a plurality of cameras, where the set of images comprises a reference image and an alternate view image;

detecting features in the set of images using a processor directed by an image processing application;

identifying within the alternate view image features corresponding to features detected within the reference image using a processor directed by an image processing application;

rectifying the set of images based upon a set of geometric calibration data using a processor directed by an image processing application;

determining residual vectors for geometric calibration data at locations where features are observed within the alternate view image based upon observed shifts in locations of features identified as corresponding in the reference image and the alternate view image using a processor directed by an image processing application;

determining updated geometric calibration data for a camera that captured the alternate view image based upon the residual vectors using a processor directed by an image processing application;

rectifying an image captured by the camera that captured the alternate view image based upon the updated geometric calibration data using a processor directed by an image processing application;

acquiring an additional set of images of a scene using the plurality of cameras;

determining residual vectors for the geometric calibration data using the additional set of images, wherein determining updated geometric calibration data for a camera that captured the alternate view image based upon the residual vectors also comprises utilizing the residual vectors for the geometric calibration data determined using the additional set of images;

detecting at least one region within a field of view of a camera that does not satisfy a feature density threshold;

wherein the additional set of images of a scene is acquired in response to detecting that at least one region within a field of view of a camera does not satisfy the feature density threshold.

wherein utilizing the residual vectors determined using the additional set of images further comprises utilizing the residual vectors determined using the additional set of images to determine updated geometric calibration data with respect to the at least one region within the field of view of the camera in which the density threshold was not satisfied; and

providing prompts via a user interface using a processor directed by an image processing application, where the prompts direct orientation of the camera array to shift locations of features identified as corresponding in the reference image and the alternate view image into the at least one region within the field of view of a camera that does not satisfy a feature density threshold during acquisition of the additional set of images.