Xia De-shen

Staircase effect alleviation by coupling gradient fidelity term

Image denoising with second order nonlinear PDEs often leads to an undesirable staircase effect, namely, the transformation of smooth regions into piecewise constant ones. In this paper, the similarity in gradient between the noisy images and the restored ones is described and preserved by the gradient fidelity term during the noise removal. The introduction of the Euler equation derived from the gradient fidelity term into nonlinear diffusion PDEs helps to alleviate staircase effect efficiently, while preserving sharp discontinuities in images. The gradient fidelity term is integrable in bounded variation function space, which makes our models outperform fourth order nonlinear PDE-based denoising methods in the preservation of edges and textures. In addition, the necessity of introducing spatial regularization into gradient estimation is theoretically analyzed and experimentally emphasized.

Fast and active texture segmentation based on orientation and local variance

This paper describes a fast and active texture segmentation approach based on the orientation and the local variance. First, a set of feature images are extracted using the orientation and the local variance. To reduce the computational complexity, a separability measurement method, which is used for selecting four feature images with good separability in four orientations, is proposed in this paper. To improve the segmentation, we adopt a nonlinear diffusion filtering to smooth the four feature images. Finally, a variational framework incorporating these features in a level set based, unsupervised segmentation process is adopted. To improve the computational speed, instead of solving the Euler-Lagrange equation, we calculate the energy, with level set representation, to solve the variational framework. Segmentation results of various synthetic and real textured images has demonstrated that our method has good performance and efficiency.