US 11,657,530 B2
Stereo matching method and apparatus of images
Yong Seok Heo, Seoul (KR); Jae Cheol Jeong, Suwon-si (KR); and Su Yeon Jeon, Daegu (KR)
Assigned to AJOU UNIVERSITY INDUSTRY—ACADEMIC COOPERATION FOUNDATION, Suwon-si (KR)
Filed by AJOU UNIVERSITY INDUSTRY—ACADEMIC COOPERATION FOUNDATION, Suwon-si (KR)
Filed on May 12, 2022, as Appl. No. 17/743,008.
Claims priority of application No. 10-2021-0071453 (KR), filed on Jun. 2, 2021.
Prior Publication US 2022/0392094 A1, Dec. 8, 2022
Int. Cl. G06T 7/593 (2017.01)
CPC G06T 7/593 (2017.01) [G06T 2207/10012 (2013.01); G06T 2207/20021 (2013.01); G06T 2207/20084 (2013.01)] 8 Claims
OG exemplary drawing
 
1. A stereo matching method of images performed by a system implemented by a computer, comprising steps of:
receiving a pair of images obtained at different time points;
generating a feature map by extracting features of each pixel of the pair of images;
generating sequentially cost volumes for each of partial time points based on the feature map and generating a feature map for an entire disparity by fusing the cost volumes using 2D convolution; and
generating a final disparity map by refining the generated feature map,
wherein in the generating of the feature map, the entire disparity is divided into M partial disparities and a feature map for each partial disparity is generated by a sequential feature fusion (SFF) module,
wherein M feature maps corresponding to each M partial disparity are generated from each M SFF, and
wherein the M features maps have the same size, and as a result, each cost volume is the same.