Prior-based Hierarchical Segmentation Highlighting Structures of Interest

Abstract

Image segmentation is the process of partitioning an image into a set of meaningful regions according to some criteria. Hierarchical segmentation has emerged as a major trend in this regard as it favors the emergence of important regions at different scales. On the other hand, many methods allow us to have prior information on the position of structures of interest in the images. In this paper, we present a versatile hierarchical segmentation method that takes into account any prior spatial information and outputs a hierarchical segmentation that emphasizes the contours or regions of interest while preserving the important structures in the image. Several applications are presented that illustrate the method versatility and efficiency.

Watershed-Based Attribute Profiles With Semantic Prior Knowledge for Remote Sensing Image Analysis

“…Then, those labeled markers guide the construction of a hierarchical segmentation by preventing regions of different classes to be merged, and by propagating the labeled markers to unlabeled regions. Another related approach, proposed in [19], uses prior knowledge to keep the regions of interest from being merged early in the hierarchy, i.e., the details in the regions of interest are preserved at high levels of the hierarchy. This later idea is also explored in the Watershed-AP framework, in which we aim to filter out the regions with low probability of belonging to a given groundtruth semantic class.…”

Watershed-Based Attribute Profiles for Pixel Classification of Remote Sensing Data

“…Then, those labeled markers guide the construction of a hierarchical segmentation by preventing regions of different classes to be merged, and by propagating the labeled markers to unlabeled regions. Another related approach, proposed in [12], uses prior knowledge to keep the regions of interest from being merged early in the hierarchy, i.e., the details in the regions of interest are preserved at high levels of the hierarchy.…”

“…In the steps 4 and 5, a gradient (V, E, w G ) of I is computed and then combined with (V, E, w P ) as a multiplication of edge weights, similarly to [15]. We note that the proposed method is related to ones introduced in [12,16], the main difference being the type of hierarchy under consideration and how the original data is combined with the prior knowledge.…”

Characterizing Images by the Gromov-Hausdorff Distances Between Derived Hierarchies

“…elongation, surfaces equilibrium, contrast) for segmentation tasks [8]. It can also take spatial prior information into account in its making [10]. If we consider our image as a topographic relief, flooding this image leads to watershed lines, i.e.…”