Parameter-Free Selective Segmentation With Convex Variational Methods

Abstract

Selective segmentation methods involve incorporating user input to partition an image into a foreground and background. Often these methods are sensitive to some aspect of the user input in a counter intuitive manner, making their use in practice difficult. The most robust methods often involve laborious refinement on the part of the user, and sometimes editing/supervision. The proposed method reduces the burden on the user by simplifying the requirements on the input. Specifically, the fitting term does not depend on a distance function and so no selection parameter is introduced. Instead, we consider how the user input relates to some general intensity fitting term to ensure the approach is less sensitive to the decisions or intuition of the user. We give comparisons to existing approaches to show the advantages of the new selective segmentation model.

A Generalized Asymmetric Dual-Front Model for Active Contours and Image Segmentation

“…Eventually, we evaluate the isotropic dual-front model, the geodesic distance thresholding model and asymmetric dualfront model on 30 CT images [67]. In this experiment, the initial contour in each image is a circle centred at an interior point that is farthest to the boundary of the the ground truth region.…”

Geodesic Paths for Image Segmentation with Implicit Region-based Homogeneity Enhancement

“…In contrast, the proposed DualCut-Asy model is able to find satisfactory segmentation in both CT images, due to the benefits from the implicit regionbased homogeneity terms and the edge asymmetric features. Finally, we present the quantitative comparison results over 86 CT images 3 [13]. For each tested image, we artificially add Gaussian white noise with mean 0 and normalized variance 0.05.…”

“…The image segmentations can be obtained by exploiting a level set-based curve evolution formulation. This is also the case for the selective segmentation models [13], [14] using scribbles to extract image feature statistical priors. For both models, the image segmentations are achieved using a convex relaxation method [15].…”

Agglomerative Clustering with Threshold Optimization via Extreme Value Theory

“…Another automatically liked clustering approach named Swarm intelligence [16] still requires parameters and the abundance values of a dataset. Some researchers developed autonomous parameter selection for image segmentation techniques, and argue that they can be viewed as a type of clustering [17][18][19][20][21]. Image segmentation is a very special problem with added constraints, and the approaches have failed to generalize to general clustering problems even in other vision applications.…”