Abstract:
We propose a new method for image filtering and enhancement based on the use of partial differential equations (PDE) framework and of image fusion techniques. The restoration process is defined iteratively as a succession of diffusion and fusion steps; in each step the degraded image is first independently processed using a directional diffusion PDE tuned to different sets of parameters and then the results are combined through fusion. Fusion takes place at an intermediate results level through a weighted averaging/selection fusion rule. An experimental setup involving both synthetic and real images is used to illustrate the increased efficiency of the method for noise filtering with small scale detail, edge and junction preservation.
