A Novel Multi-Fractal Image Description Method for Complex Linear Object

Fractal geometry works well in image singularity analysis. In this paper, we describe a novel representation of linear objects in road crack images using multifractal. Firstly, we give the definition of linear objects and texture around linear objects; and then we use the distribution of texture around linear objects to represent cracks. In addition, we present a multifractal feature extraction method for linear objects. We first estimate the orientation map of textures by phase-grouping method, and then compute the discrete distribution curves of linear features in several homocentric annuluses in different scales. Our results show that the bandwidth between the maximum and zero-value point of multifractal function, Δ(α) is proportion to the complex of texture of linear object.