RADIAL CENTROID CURVE FOR AFFINE INVARIANT FEATURES EXTRACTION

摘要

The extraction of affine invariant features plays an important role in many fields of computer vision. Contour-based wavelet methods are unapplicable to objects with several separable components. In this paper, a method is proposed by converting the object into a closed curve, which is called radial centroid curve (RCC). Point on this curve is the centroid of radial line segment from centroid of the object. It is shown that the RCC derived from the affine transformed object is the same affine transformed version as that of the original object. An affine invariant function (AIF) is constructed by applying stationary wavelet transform (SWT) to the derived RCC. This scheme is applicable to objects with several separable components. Several experiments have been conducted to evaluate the performance of the proposed method.