Image filtering using weighted curvature-preserving PDE

The tensor-driven curvature-preserving partial differential equation has been an outstanding anisotropic diffusion filtering model. In this paper, a weighted modification is proposed, which equals to weighted averaging of different Line Integral convolutions utilizing local image directional information to adaptively design weight coefficients for different integral curves. Finally, the efficiency of the new approach is tested on a commonly-used standard test image database, in terms of speed and filtering performent.