US 12,394,070 B2
Multi-level optical flow estimation framework for stereo pairs of images based on spatial partitioning
Niloufar Pourian, Los Gatos, CA (US); and Oscar Nestares, San Jose, CA (US)
Assigned to Intel Corporation, Santa Clara, CA (US)
Filed by Intel Corporation, Santa Clara, CA (US)
Filed on Sep. 23, 2020, as Appl. No. 17/029,896.
Prior Publication US 2021/0004969 A1, Jan. 7, 2021
Int. Cl. G06T 7/269 (2017.01)
CPC G06T 7/269 (2017.01) [G06T 2207/10012 (2013.01); G06T 2207/20021 (2013.01); G06T 2207/20081 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A system comprising:
a memory to store at least a portion of an image pair, the image pair including a first image and a second image;
computer readable instructions; and
at least one programmable circuit to be programmed by the computer readable instructions to:
apply a first machine learning model to a first input volume associated with a first partition of the first image and a corresponding second partition of the second image to generate first optical flow data associated with a full resolution of the image pair;
apply a second machine learning model to a second input volume different from the first input volume, the second input volume associated with a downsampled version of the first image and a downsampled version of the second image to generate second optical flow data associated with a lower resolution than the full resolution of the image pair; and
simultaneously apply at least a portion of the first optical flow data associated with the full resolution and at least a portion of the second optical flow data associated with the lower resolution to a merge operation, the merge operation to generate an optical flow map for the image pair.