Meno:Michal
Priezvisko:Jančoąek
Názov:Feature Detection and Tentative Correspondence Estimation in Wide Baseline Stereo
Vedúci:Doc. RNDr. Andrej Ferko PhD.
Rok:2005
Blok:PG
Kµúčové slová:Baseline Stereo; Tentative Correspondence; Maximally Stable Extremal Regions; MSER; Local Affine Frames; LAF; True Tentative Correspondences; TTC
Abstrakt:The aim of this work is to estimate tentative correspondence in a wide baseline image pairs with a well known method and possibly propose a new method. We made a research on known correspondence methods. This work experimentally shows a verification of our implementation of Maximally Stable Extremal Regions (MSER) detection method and describe our implementation in details. We have also implemented the method to estimate tentative correspondences using Local Affine Frames (LAF). In this work is also proposed a new method called True Tentative Correspondences (TTC) which use MSER as features which were put into correspondence. Input to our algorithm is a widebaseline image pair and output is a set where one element consist of eight tentative correspondences between detected MSER regions, these are the best candidates to compute epipolar geometry between images. There is also proposed a new algorithm to estimate epipolar geometry in image pair using TTC as is mentioned in a future work.

Súbory diplomovej práce:

MasterThesisJancosekFinal.pdf