摘要

In this paper, a new set of Fourier descriptors for color image recognition, the GCFD3, is proposed. These descriptors are computed using the color Clifford Fourier transform (CFT) of Batard et al. (2010). The GCFD3 are invariant under translation and rotation and avoid a marginal processing of color images. Unlike the GCFD1 defined in (Mennesson et al., 2011), the goal is to obtain invariant descriptors under bivector change (a color plane parameterizing the CFT) and to reduce the size of the descriptors vector required for image recognition. These descriptors are tested in a color image recognition framework on different well-known image databases, showing their capabilities to identify color objects.

  • 出版日期2012