New anisotropic diffusion operator in images filtering

Abstract

The anisotropic di usion lters have become in the fundamental bases to address the medical images noise problem. The main attributes of these lters are: the noise removal e ectiveness and the preservation of the information belonging to the edges that delimit the objects of an image. Due to these excellent attributes, through this article, a comparative study is proposed between a new di usion operator and the Lorentz operator, proposed by the pioneers of anisotropic di usion. For this, a strategy consisting of two phases is designed. In the rst, called operator construction, the composition of functions is used to generate a new di usion operator that meets with the conditions reported for this kind of the mathematical object. In the second phase, denominated ltering, a synthetic cardiac images database, based on computed tomography, is ltered using the aforementioned operators. According with the value obtained for the peak of the signal-to-noise ratio, the new operator shows similar performance to the Lorentz operator. The implementation of this new operator contributes to the generation of new knowledge in digital image processing context.