One of the most widely used denoising techniques with well preserving edge features in medical imaging is
anisotropic diffusion filtering. It is based on the iterative solving of the diffusion equation that takes into account only the
heat propagation by conduction1. At living beings, blood perfusion represents another important way to transfer the heat.
In this case, we used a new equation modeling the heat propagation, mainly Pennes equation2, processing both
conductive and convective heat components. Unlike heat, white noise, which accompanies the signal detected by a
thermal sensor, is not a solution to this equation. The new filtering method consists in an iterative solving of bio-heat
equation by using Crank-Nicolson convergent algorithm. Bazan’s criterion for stopping iterations agrees well with the
behavior of this filter. The filter was tested on some theoretical models of images simulating different signals and noises,
and on many thermal images of healthy people or patients suffering from different types of thyroid disorders. One image
processing of a patient suffering from papillary carcinoma is shown at different time moments. The noise is rapid
attenuated, and it is possible to assess the contour shape or to locate more outbreaks in a certain area, if any.