Feedback coherent anisotropic diffusion for high resolution image enhancement

Anisotropic diffusion has proved efficiency in filtering wide range of images. However, its application to improve SNR in MR images was limited due to the minute details existing in many medical image modalities e.g., MR and CT images. Therefore, these types of images require sophisticated sensitive filtering methods. In this paper we propose a modified variational model based on the coherent diffusion for enhancing SNR. The difference between the filtered and the degraded image may contain some residual coherent information. This coherent information means that the applied diffusion parameters in these areas were overestimated. In this paper, the coherence in the difference image is used as a feedback to re-estimate diffusivities in places where coherent leakage was detected. Results show that information associated with resolved-object structure, even minute structures, were substantially preserved.