Fast Matching of Planar Shapes in Sub-cubic Runtime

Design and perceptual validation of performance measures for salient object segmentation

“…Typically, the matching cost is based on two components: 1) dissimilarity of local properties of matched points, e.g., tangent orientations, and 2) dissimilarity of matched curve segments, i.e., the cost of deforming one curve segment (stretching, bending or compressing) to match the other curve segment. Equipped with a translation-, rotation-and scale-invariant cost function, these measures have proven effective in image database search applications and for clustering of shape databases [21]. Optimization is often based on cyclic string correction [16] methods and its vari- Note that although the mapping is monotonic, it is not necessarily strictly monotonic, and thus need not be 1:1.…”

Structured Learning and Prediction in Computer Vision

“…Outerplanar MRFs have been used by Batra et al [11] to approximately solve hard random field instances. Even in the case of polynomial-time solvable instances the computational complexity can be reduced by exploiting planarity, as shown by Schmidt et al [132,133]. Note that planarity is distinct from treewidth in that a planar graph may have high treewidth.…”

IS-Match: Partial Shape Matching by Efficiently Solving an Order Preserving Assignment Problem

“…Shape clustering methods were recently presented by Schmidt, et al 21) who clustered 40 shapes of 4 classes using dynamic time warping matching and kmeans clustering. Yankov and Keogh 31) clustered shapes for grouping together objects in large collections by a manifold clustering approach.…”