Finding Homologous Lines
Arthur Ardeshir Goshtasby · 2017
This chapter relates the transformation parameters for registration of two images to the parameters of homologous lines in the images. It describes methods for determining homologous lines in images of a scene. Then, the chapter gives examples of image registration using homologous lines in images. The chapter summarizes the number of homologous lines required to determine the parameters of various transformation models to register two images of a planar scene. The weakness of clustering is that parameters resulting from most line matches are selected without verification of the correctness of the obtained parameters. Rather than adding a verification step to the clustering-based correspondence algorithm, one can use a RANSAC-based algorithm, which already has a built-in verification step, to find homologous lines in images. The chapter also short reviews vanishing point detection methods, details implementation of one vanishing point detection method and demonstrates its use in line grouping.