US 12,406,466 B1
Gaze correction of multi-view images
Eric Sommerlade, Oxford (GB); and Michael G. Robinson, Boulder, CO (US)
Assigned to RealD Spark, LLC, Boulder, CO (US)
Filed by RealD Spark, LLC, Boulder, CO (US)
Filed on Nov. 20, 2023, as Appl. No. 18/514,040.
Application 18/514,040 is a continuation of application No. 17/703,039, filed on Mar. 24, 2022, granted, now 11,854,243.
Application 17/703,039 is a continuation of application No. 16/937,382, filed on Jul. 23, 2020, granted, now 11,317,081, issued on Apr. 26, 2022.
Application 16/937,382 is a continuation of application No. 16/405,415, filed on May 7, 2019, granted, now 10,750,160, issued on Aug. 18, 2020.
Application 16/405,415 is a continuation of application No. 15/397,951, filed on Jan. 4, 2017, granted, now 10,321,123, issued on Jun. 11, 2019.
Claims priority of provisional application 62/274,897, filed on Jan. 5, 2016.
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 3/01 (2006.01); G06V 10/46 (2022.01); G06V 10/62 (2022.01); G06V 40/18 (2022.01); H04N 7/14 (2006.01); H04N 7/15 (2006.01); H04N 13/344 (2018.01); H04N 13/117 (2018.01)
CPC G06V 10/62 (2022.01) [G06F 3/011 (2013.01); G06F 3/013 (2013.01); G06V 10/462 (2022.01); G06V 40/193 (2022.01); H04N 7/147 (2013.01); H04N 7/15 (2013.01); H04N 13/344 (2018.05); H04N 13/117 (2018.05)] 20 Claims
OG exemplary drawing
 
1. A method of adjusting multi-view images of a head to modify a perceived gaze, the method comprising:
identifying, in each image of the multi-view images, captured image patches including a left image patch containing a left eye of the head and a right image patch containing a right eye of the head;
deriving a feature vector from local image descriptors of the right image patch or the left image patch;
determining a displacement vector field representing a transformation of the right image patch or the left image patch, based on the derived feature vector, and based on reference data including one or more predefined relationships between displacement vector fields and feature vectors; and
adjusting one or more of the multi-view images by transforming one or both of the right image patch and the left image patch based on the determined displacement vector field.