An anisotropic diffusion system with nonlinear time-delay structure tensor for image enhancement and segmentation

Anisotropic diffusion model has been established as one of the most significant image processing techniques, because it not only allows for space-variant filtering, but also realizes real direction-specific anisotropic smoothing of image structures. In this paper, an anisotropic nonlinear diffusion system is proposed for image enhancement and segmentation by taking into account the structure tensor constructed from the time-delay regularization and a nonlinear isotropic diffusion equation. Meanwhile, the diffusion system makes full use of the source term which increases the contrast and makes the image be less susceptible to noise, even in high level noise images. The source term also causes the restored image to tend to the binary image, which has proven to be an effective strategy for image segmentation. We analyze the proposed model in the view of the existence and uniqueness of the weak solution by using Galerkin's method. Some other theoretical analysis such as the maximum property is also discussed in the paper. With the appropriate and adaptive parameters, the anisotropic diffusion system achieves a better trade-off between computation time and image processing task. Furthermore, various image experiments illustrate that the proposed algorithm achieves much better efficiency and universality.