Multiscale Fourier descriptor for shape classification

Abstract

The description of the object shape is an important characteristic of the image. In image processing and pattern recognition, several different shape descriptors are used. In human visual perception, the shapes are processed in multiple resolutions. Therefore multiscale shape representation is essential in the shape based image classification and retrieval. In the description of the object shape, the multiresolution representation provides also additional accuracy to the shape classification.In this paper we introduce a new descriptor for shape classification. This descriptor is called multiscale Fourier descriptor, and it combines the benefits of Fourier descriptor and multiscale shape representation. This descriptor is formed by applying Fourier transform to the coefficients of wavelet transform of the object boundary. In this way the Fourier descriptor can be presented in multiple resolutions.We make classification experiments using three image databases. The classification results of our method are compared to those of Fourier descriptors.

Shape classification via image-based multiscale description

“…Regionbased methods can overcome these limitations. The most common boundary-based shape descriptors are Fourier descriptors [1][2][3][4], wavelet descriptors [5], wavelet-Fourier descriptors [6][7][8] and curvature scale space (CSS) [9].…”

“…However, the matching schema was more complicated than for Fourier descriptors. Kunttu et al [6][7][8] introduce multiscale Fourier descriptors using wavelet and Fourier transforms. The multiscale contour Fourier descriptors are obtained by applying the Fourier transform to the coefficients of the multiscale complex wavelet transform.…”

Shape classification using multiscale Fourier-based description in 2-D space

“…The most common boundarybased shape descriptors are chain codes [1], wavelet descriptors [2], Curvature Scale Space (CSS) [3] and Fourier descriptors [4]. Fourier descriptors can also be combined with the wavelet transform to smooth the boundary in one dimension to produce multiscale Fourier descriptors [5]. Common region-based shape descriptors are moments [4] [6] and Generic Fourier Descriptors (GFDs) [7].…”

Use of Dynamic Time Warping for Object Shape Classification through Signature

“…Besides, Kunttu et al, 2003 positioned the concept of multi-scale Fourier descriptor for shape classification that includes an idea of curvature Fourier, radius Fourier, contour Fourier (Kauppinen et al, 1995), and A-invariant methods for Fourier-based shape representation (Arbter et al, 1990). Moreover, (Kunttu et al, 2003) added that contour Fourier and Ainvariant methods were best approaches in shape classification. The Contour Fourier method transforms the Fourier directly for the complex coordinate function of the object boundary in which both positive and negative frequency axis descriptors are taken.…”

“…In human visual perception, the shapes are processed in a multiple resolutions and therefore multi-scale shape representations is essential (Kunttu et al, 2003) in the shape based image classification and retrieval. This multiresolution representation provides additional accuracy to the shape classification.…”