Geometric Image Correspondence Verification by Dense Pixel Matching
This addresses the problem of improving image retrieval accuracy for applications like visual localization, though it is incremental as it builds on existing dense matching networks.
The paper tackles dense pixel correspondence verification for image retrieval by proposing a geometric verification approach that re-ranks retrieved images using dense matching and a novel similarity function, achieving favorable comparisons to state-of-the-art methods.
This paper addresses the problem of determining dense pixel correspondences between two images and its application to geometric correspondence verification in image retrieval. The main contribution is a geometric correspondence verification approach for re-ranking a shortlist of retrieved database images based on their dense pair-wise matching with the query image at a pixel level. We determine a set of cyclically consistent dense pixel matches between the pair of images and evaluate local similarity of matched pixels using neural network based image descriptors. Final re-ranking is based on a novel similarity function, which fuses the local similarity metric with a global similarity metric and a geometric consistency measure computed for the matched pixels. For dense matching our approach utilizes a modified version of a recently proposed dense geometric correspondence network (DGC-Net), which we also improve by optimizing the architecture. The proposed model and similarity metric compare favourably to the state-of-the-art image retrieval methods. In addition, we apply our method to the problem of long-term visual localization demonstrating promising results and generalization across datasets.