Towards Invariant Interest Point Detection of an Object

Abstract

Detection of some interest points on an object is useful for many applications, such as local shape description of the object, recognition of the object in clutter environment etc. The same object present in different images can have some geometric and photometric transformations with respect to one another. The detection method should be robust to all these transformations. We describe relative scale Harris method for interest point detection. This method is robust to linear geometric transformations. A threshold selection method is also described for invariance to intensity change, partial occlusion and clattered environment. Unlike multi-scale methods our method is fast enough to be suitable for real time applications.