Image Segmentation Evaluation by Techniques of Comparing Clusterings

Abstract

Abstract. The task considered in this paper is performance evaluation of region segmentation algorithms in the ground truth (GT) based paradigm. Given a machine segmentation and a GT reference, performance measures are needed. We propose to consider the image segmentation problem as one of data clustering and, as a consequence, to use measures for comparing clusterings developed in statistics and machine learning. By doing so, we obtain a variety of performance measures which have not been used before in computer vision. In particular, some of these measures have the highly desired property of being a metric. Experimental results are reported on both synthetic and real data to validate the measures and compare them with others.

Distance Measures for Image Segmentation Evaluation

“…As we will see later, some of the measures even have the highly desired property of being a metric. Note that this paper is a substantially extended version of [5]. The extension includes a new distance measure based on bipartite graph matching, more detailed discussion of the distance measures and their properties, and additional comparison work (Sections 4 and 5.3).…”

Nested Partitions Properties for Spatial Content Image Retrieval

“…Among the most promising metrics which particularly have desirable properties one can name the Earth Mover's Distance (EMD) (Rubner, 2000), Region Matching Distance (RMD) (Hjaltason et al, 2003), variation of information (Meila, 2003), Van Dongen distance (Jiang et al, 2005), and partition metric introduced in (Mashtalir et al, 2006) and extended in (Kinoshenko et al, 2007).…”