Improved coherence-enhancing diffusion with hybrid diffusion tensor

Coherence-enhancing diffusion (CED), based on analysis of oriented structures, has been extensively used in image processing. This diffusion filtering can keep some junctions and close broken linear structures, but it also destroys crease points and deforms nonlinear structures. In this paper, we proposed an improved algorithm for CED based on analysis of structure tensor and Hessian matrix. This approach can not only denoise and enhance linear structures such as edges, but also preserve nonlinear structures such as creases. Experiments with fingerprint image show that the improved CED outperforms classical CED.