Integrating edge and surface information for range image segmentation

<title>Fast extraction of planar surfaces from range images</title>

“…In a scan line z(x, yo), Yo E 'N, the points belonging to S are z = Ax + Byo + C = Ax + B0 (2) that clearly form a straight line segment in the x -z plane. Different line segments of S have the same slope A but different intercepts B0 which are dependent on Yo.…”

“…The used paradigms range from split-and-merge [11,16,13], clustering [6,7] to relaxation [9]. While most of these algorithms are region-based, an integration of edge and surface information could give more accurate segmentation results [2]. Despite the restriction to the planar regions, the known algorithms are still quite time-consuming, for example:…”

Saliency Sequential Surface Organization for Free-Form Object Recognition

“…Jain and Nadabar [21] describe a Markov Random Field (MRF) approach that integrates both region and edge based segmentations. Bhandarkar and Siebert [22] exploit the synergy between the processes of detecting surface discontinuities and surface regions by using the geometrical properties of the detected surface regions. Homogeneous surface regions are extracted using a combination of clustering in parameter space and region growing.…”

“…After such points of discontinuity have been detected, the edges are "grown" to ensure topological consistency. For a sampling of work in edge-based range image segmentation see [14,15,22,24].…”

Iterative thresholding for segmentation of cells from noisy images

“…These region‐based algorithms are usually more time‐consuming. The combination of region and edge‐based methods are reported by Bhandarkar & Siebert (1992) and Chu & Aggarwal (1990). In addition to the edge‐ and region‐based image segmentation algorithms, clustering algorithms and neural network approaches for segmentation of images are investigated extensively in the literature (Nguyen & Cohen, 1990; Celenk, 1991; Lin et al ., 1992; Pappas, 1992; Wu et al ., 1996b).…”