Multi-directional broad learning system for the unsupervised stereo matching method

Zihao Zhang, Niu Ying, Fanman Meng, Tiejun Yang, Chao Fan, Xiaozhen Ren, Ruiqi Wu, Cao Kun, Haocheng Wang

Abstract: Highlights•The broad learning system is first introduced to the unsupervised stereo matching.•Construct a multi-directional broad learning system to generate candidate disparity maps.•The invalid depth redefinition based on local gravity weight method is designed by combining the local color space, matching cost space and geometric space.•The proposed method can show superiority in indoor and outdoor data sets without ground truth disparity map.