A Novel Method for Redundant Feature Rejection in Correspondence Problem

Feature vector plays an important role in many computer vision applications such as image registration, face recognition, object tracking etc. In many cases, the dimension of the extracted feature vectors using algorithms such as SIFT (Scale-Invariant Feature Transform), SURF (Speeded Up Robust Features), PFH (Point Feature Histogram), FPFH (Fast Point Feature Histogram) etc. can be very large. At the same time, a large number of extracted features is not necessarily conducive to better correspondence. It can make the process very slow and unsuitable for real time use. A large number of feature vectors increases the computational burden in subsequent processing. The feature vectors may be redundant most of the time. Features extracted should not be unique but also as mutually distinct as possible for best results in its application. In this paper a novel method is proposed to control the number of feature vectors which can be applied with various feature extraction algorithms. It seeks to ensure uniqueness and distinction of the selected features. A controllable number of feature vectors is very desirable in many situations and makes this approach very relevant for real time problems.