Structure extraction of images using anisotropic diffusion with directional second neighbour derivative operator

Abstract

The aim of structure extraction is to decompose an image into prominent structures and textures. In this paper, we present a new structure extraction method which has two main steps. First, high-frequency components due to the texture information in the original image are alleviated by a pre-smoothing filter. The result is then processed by a new anisotropic diffusion algorithm which uses a second neighbour derivative (SND) operator instead of the first neighbour derivative operator. We have demonstrated that the SND operator is better suited for applications such as texture smoothing. We have also presented a detailed study of the proposed method including the selection of the pre-smoothing filter, the number of iterations, and the scale parameter in the anisotropic diffusion algorithm. We have conducted experiments to compare the performance of the proposed method with those state-of-the-art structure extraction algorithms in a wide range of image editing applications such as: superpixel segmentation, texture transfer, contrast enhancement, and pencil drawing. We show that while the running speed of the proposed method is the fastest, its performance is competitive to other methods.