HEASK: Robust homography estimation based on appearance similarity and keypoint correspondences

作者：

Highlights：

• A new loss function is formulated in homography verification.

• The function combines models of keypoint correspondence fitness and appearance similarity.

• Keypoint fitness errors are better characterized with a Laplacian distribution.

• Image similarity is described using the distribution of ECC.

• The algorithm is realized by an improved RANSAC framework.

摘要

•A new loss function is formulated in homography verification.•The function combines models of keypoint correspondence fitness and appearance similarity.•Keypoint fitness errors are better characterized with a Laplacian distribution.•Image similarity is described using the distribution of ECC.•The algorithm is realized by an improved RANSAC framework.

论文关键词：Homography estimation,Keypoint consensus,Appearance similarity,RANSAC

论文评审过程：Received 6 March 2012, Revised 10 April 2013, Accepted 13 May 2013, Available online 20 June 2013.

论文官网地址：https://doi.org/10.1016/j.patcog.2013.05.007