keyword
weak geometric consistency
Weak geometric consistency is a fast spatial verification technique in image retrieval that filters out incorrect local feature matches by checking whether they share simple, uniform transformations such as consistent changes in scale, orientation, or translation. Rather than estimating a complex and computationally demanding global geometric transformation like an affine model or homography, this approach treats transformation parameters independently, typically using low-dimensional voting or histogram binning. By discarding matches whose scale and rotation differences deviate from the dominant consensus across an image pair, weak geometric consistency can be efficiently incorporated directly into indexing structures, providing an effective balance between matching accuracy and search scalability.
1 item
