COHERENCE ENHANCING DIFFUSION FILTERING BASED ON THE PHASE CONGRUENCY TENSOR

Many biomedical applications require the enhancement of coherent flow-like curvilinear structures in images. This can be accomplished in a natural way by adopting anisotropic diffusion filtering to local texture analysis by means of the structure tensor. Here we propose a contrast independent anisotropic diffusion filtering of curvilinear structures based on a novel concept: the Phase Congruency Tensor (PCT). The phase angle of the PCT determines the preferred diffusion direction and the diffusity is controlled by PCT-based vesselness measure. We show that the proposed method is largely insensitive to intensity variations along the curve, and provides successful enhancement within noisy regions. The performance of the proposed approach was evaluated by comparing it with a state-of-the-art intensity-based approach on both synthetic and real biological images.