Thermal Image Filtering by Bi-dimensional Empirical Mode Decomposition

The abnormal function of cells can be detected by anatomic or physiological registrations. Most of modern approaches, as ultrasound, RMN or CT, show anatomic parametric modifications of tissues or organs. They highlight areas with a larger diameter 1 cm. In the case of skin or superficial cancers, local temperature is different, and it can be put out by thermal imager. Medical imaging is a leading role in modern diagnosis for abnormal or normal tissues or organs. Some information has to be improved for a better diagnosis by reducing or removing some unwanted information like noise affecting image texture. The traditional technologies for medical image enhancement use spatial or frequency domain methods, but whole image processing will hide both partial and specific information for human signals. A particular kind of medical images is represented by thermal imaging. Recently, these images were used for skin or superficial cancers diagnosis, but very clear outlines of certain alleged affected areas need to be shown. Histogram equalization cannot highlights the edges and control the effects of enhancement. A new filtering method was introduced by Huang by using the empirical mode decomposition, EMD. An improved filtering method for thermal images, based on EMD, is presented in this paper, and permits to analyze nonlinear and non-stationary data by the adaptive decomposition into intrinsic mode surfaces. The results, evaluated by SNR ratios, are compared with other filtering methods.