computer vision - Decompose homography matrix to rotation and translation -

how can correct r , t h when have 4 solutions. understanding can eliminate two.

is there no way correct solutions if have points on single plane?

thanks help.

there 1 approach best select solution 2 possible solutions. when decompose homography,after eliminating 2 possible solutions 4,you 2 set of

rotations, translations , normals.

normals nothing possible camera_normals wrt scene of first image(when estimate homography between 2 images). having known camera_normal wrt planar scene when first image captured, can choose solution corresponding closest normal n1,n2 computing dot product dot(camera_normal,n1) , dot(camera_normal,n2), whichever larger.

thanks

---

edit: camera_normal explanation

planar homography assumes features used computing homography between tow images in plane. camera_normal direction of z-axis of camera in plane's frame.