Sequency-ordered generalized Walsh-Fourier Transform based shape description and retrieval

Abstract

Sequency-ordered generalized Walsh-Fourier transform (SGWFT) is a new orthogonal transform family. SGWFT shows good properties, and is similar to Discrete Fourier Transformation in many ways. SGWFT has lower complexity with less number of multiplications required than the DFT, which makes it has a better performance in potential applications. In this paper, SGWFT is used to derive new shape descriptor (SGWFD) for shape image retrieval. The new descriptor uses a complex shape signature to express sampled shape, and then applies SGWFT to the signature. The resulting transform coefficients are used as corresponding shape features to form SGWFD. The image retrieval performance of the proposed descriptor is evaluated on the Swedish leaf database using standard measurement, and the experimental results show that the proposed SGWFD outperforms Fourier descriptor.