Abstract

Fourier Descriptors (FD) generation depends heavily on the input shape signature and is a core component in traditional Content Based Image Retrieval (CBIR) systems. This paper presents a novel basic shape classifier developed using Complex Coordinates (CC) FD. A spatial domain normalisation of the FD is achieved by overlaying a Fourier Synthesised Boundary (FSB) against its original Boundary Points (BP). This process creates Intersection Points (IP). A new shape signature is formed using a ratio representing the number of IP over the number of BP. The shape signature coined as Spatially Normalised Fourier Shape Signature (SNFSS), varies from 0 to 1 with increasing number of FD used, and exhibit key trends for the detection of basic shapes like circle and regular polygons. The trends are proven experimentally to be invariant to scale and rotation, as well as being robust to noise.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call