US 12,079,929 B2
Three-dimensional (3D) model generation from two-dimensional (2D) images
Nima Sarshar, Miami, FL (US); and Max Hwang, Hayward, CA (US)
Assigned to NexTech AR Solutions Inc., Toronto (CA)
Filed by NexTech AR Solutions Corp., Toronto (CA)
Filed on Mar. 29, 2022, as Appl. No. 17/707,877.
Claims priority of provisional application 63/243,185, filed on Sep. 12, 2021.
Prior Publication US 2023/0083607 A1, Mar. 16, 2023
Int. Cl. G06T 17/10 (2006.01); G06F 30/10 (2020.01); G06T 7/10 (2017.01); G06T 7/62 (2017.01); G06T 7/70 (2017.01); G06T 7/90 (2017.01); G06T 15/04 (2011.01); G06T 17/00 (2006.01); G06T 17/20 (2006.01); G06T 17/30 (2006.01); G06V 10/74 (2022.01); G06V 20/64 (2022.01); G06F 30/12 (2020.01); G06N 20/20 (2019.01)
CPC G06T 17/00 (2013.01) [G06F 30/10 (2020.01); G06T 7/10 (2017.01); G06T 7/62 (2017.01); G06T 7/70 (2017.01); G06T 7/90 (2017.01); G06T 15/04 (2013.01); G06T 17/10 (2013.01); G06T 17/20 (2013.01); G06T 17/30 (2013.01); G06V 10/761 (2022.01); G06V 20/64 (2022.01); G06F 30/12 (2020.01); G06N 20/20 (2019.01); G06T 2207/20044 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/30196 (2013.01)] 18 Claims
OG exemplary drawing
 
1. A method comprising:
receiving an object image, the object image comprising a two-dimensional (2D) array of color pixels that depict an object;
generating an object mask based on the object image, the object mask comprising a set of pixels from the object image, the set of pixels comprising pixels corresponding to the object in the object image;
generating an object skeleton for the object based on the object mask and the object image, the object skeleton comprising a subset of the set of pixels that represent a structure of the object in the object mask, wherein generating the object skeleton comprises identifying, based on the object image and the object mask, a set of pixels of the 2D array of pixels that are substantially equidistant to more than one closest border pixel;
generating a plurality of pixel partitions based on the object skeleton, each pixel partition comprising a plurality of pixels from the object image corresponding to a parallel cross-section of the object; and
generating a three-dimensional (3D) object model based on the pixel partitions.