Abstract

   In this paper we propose an improved method for image filtering with slow time variation obtained from some successive frames of a thermovision camera investigating the heat radiated by human beings. In the first step, a single image is obtained as a result of the mean value evaluation from median filtering of all temporal frames. The standard deviation of each pixel is computed as well. The second stage consisted of a spatial filtering of the image from the first step by using the anisotropic diffusion filtering. New relations for the weighted coefficients of the diffusion matrix were proposed from the local statistic estimators. This method was tested on infrared images of different patients. Both subjective and objective evaluations of a patient image suffering from papillary thyroid cancer showed good performances for improving the signal to noise ratio and location of tumor.