This paper investigates the scale selection problem for nonlinear diffusion scale-spaces. This topic comprises the notions of localization scale selection and scale space discretization. For the former, we present a new approach. It aims at maximizing the image content's presence by finding the scale that has a maximum correlation with the noise-free image. For the latter, we propose to adapt the optimal diffusion stopping time criterion of Mrázek and Navara in such a way that it may identify multiple scales of importance.Scale-space theory provides the computer vision and image processing communities with a powerful tool for multiscale signal representation. Such representation can be used e.g. in image segmentation, where the scale-space framework is used to include scale in the corresponding feature detectors (Lindeberg 1998), or directly for segmentation (Pratikakis 1998;Mukhopadhyay and Chanda 2003;Vanhamel et al. 2003;Petrovic et al. 2004;Katartzis et al. 2005;Bresson et al. 2005). Yet, scale-spaces contain a very large amount of information, most of it being redundant. Considering the increasing size of the actual image data, a reduced variant of this kind of signal representation is needed. Deriving a compact version of the scale-space is desirable also for computational accuracy reasons: by processing an image only at certain scales, known to contain the most important information concerning image embedded features, we eliminate the additional spurious data which are introduced when processing a larger number of scales.In scale-space theory, the original image is embedded in a family of gradually smoother versions. Scale-space filters using nonlinear PDE-based diffusion processes are particularly interesting since they avoid blurring and dislocation of important features. In this work, we focus on edge-affected diffusion processes in which the diffusion is locally adaptive aiming to favor intra-region instead of inter-region smoothing, thus overcoming the dislocation of region boundaries. The mathematical form of this type of processes (Catté et al. 1992;Whitaker and Gerig 1994) is given by: ∂ t u (i) (t) = div g(|∇ σ r u(t)|)∇u (i) (t) , ∀i = 1, 2, . . . , M and ∀t ∈ R + .(1)
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.