Guided Local Feature Matching with Transformer

GLFNet is proposed to be utilized for the detection and matching of local features among remote-sensing images, with existing sparse feature points being leveraged as guided points. Local a crucial step in applications 3D reconstruction. However, methods that detect image pairs match them separately may fail establish correct matches images significant differences lighting or perspectives. To address this issue, problem reformulated extraction corresponding target image, given from source explicit guidance. The approach designed encourage sharing landmarks by searching regions similar image. For purpose, developed search network. main challenge lies efficiently accurate matches, considering massive number tackle problem, network divided into coarse-level network-based point transformer narrows space fine-level regression produces matches. experimental results on challenging datasets demonstrate method provides robust benefits various applications, including registration, optical flow estimation, visual localization, reconstruction registration. Overall, promising solution offered applications.