Assessment of segmentation methods for pore detection in cellular concrete images

In the last years the use of cellular concretes has been extended due to the rise in the ratio strength/weight reached. Porosity is a property that must be taken into account because it is associated directly to the performance of a cellular concrete. The mercury porosimetry and vacuum saturation are test used to concrete porosity. However, these tests are expensive, and it requires a careful preparation of samples. Another way to determine porosity and pore distribution over concrete is reconstruction using high-resolution images from microscopy. As an alternative, in this work we compare traditional edge detection methods and fractional derivate method to detect the pores in images taken from a flat sample of cellular concrete. The experiments show that the method based on fractional derivate is more accurate to detect the pores, which is the first step to estimate total porosity of cellular concrete through non-specialized images.