Wavelet Descriptor for Closed Curves Detection in Complex Background

In the problem of designated a well-defined wavelet describing the closed curve, it is critical that we define a unique starting point for the wavelet representatives of closed curve. In this paper, we design a one-dimensional periodized wavelet transformation for closed curves detection in complex background. The uniqueness property facilitates the quantitative analysis of the unique properties of the one-to-one mapping between the changes of one-dimensional discrete periodized wavelet transformation and the starting point of original sample close curve. We propose uniqueness wavelet descriptor (UWD) by using the unique properties of a new shape descriptor. Robustness of the UWD in complex background is analyzed. By enhancing local shape feature, some experiments show adaptive property of the UWD for ideal starting point determination. Our experiments show that the UWD can provide an optimal pattern classification in complex background.