Roughness from Alpha Stable Images Applied to Fossil Images

Among the parameters used to study the texture of an object's surface is the roughness descriptor. The digital image of the object surface is used to provide roughness descriptors mainly when the surface studied is deformable or damaged by a roughness meter. This work proposes a new roughness descriptor from the image. The particularity of this descriptor is that it is adapted to surfaces containing significant fluctuation (enormous variation). This descriptor is constructed from the estimator of the spectral density of alpha stable processes. To compare our method to other methods developed in the literature, we measured the sensitivity of the descriptors to random Gaussian noise added to each image. In this work, some images of fossils representing significant variation are the subject of the images studied. This study shows that our descriptor is more sensitive than other descriptors for this kind of image (the sensitivity value is higher). Thus, the originality of this work lies in proposing a descriptor adapted to images with large fluctuations. This work falls within the framework of developing adapted methods considering different surface structures in order to provide descriptors that are more precise.