A Similarity-Based Approach for Shape Classification Using Region Decomposition
Measuring the similarity of two shapes is an important task in human vision systems in order to either recognize or classify the objects. For obtaining reliable results, a high discriminative shape descriptor should be extracted by considering both global
- PDF / 720,822 Bytes
- 11 Pages / 439.37 x 666.142 pts Page_size
- 7 Downloads / 247 Views
Abstract. Measuring the similarity of two shapes is an important task in human vision systems in order to either recognize or classify the objects. For obtaining reliable results, a high discriminative shape descriptor should be extracted by considering both global and local information of the shape. Taking into account, this work introduces a centroid-based tree-structured (CENTREES) shape descriptor invariant to rotation and scale. Extracting the CENTREES descriptor is started by computing the central of mass of a binary shape, assigned as the root node of tree. The entire shape is then decomposed into b sub-shapes by voting each pixel point according to an angle between point and major principal axis relative to a centroid. In the same way, the central of mass of the sub-shapes are calculated and these locations are considered as level-1 nodes. These processes are repeated for a predetermined number of levels. For each node corresponding to sub-shapes, parameters invariant to translation, rotation and scale are extracted. A vector of all parameters is considered as descriptor. A feature-based template matching with X2 distance function is used to measure shape dissimilarity. The evaluation of our descriptor is conducted using MPEG-7 dataset. The results justify that the CENTREES is one of reliable shape descriptors for shape similarity. Keywords: Shape descriptor matching
Region decomposition
Centroid
Shape
1 Introduction Shape classification plays an important part in a visual object recognition. To do such a task, technically, the various shapes should be known previously considered as templates. Given an unknown shape, our goal is to classify this shape into one of the shape template classes by computing the similarity value of their descriptors, namely shape matching. The similarity computation can be done using some similarity measurements [1] such as Manhattan, Euclidean, X2, and other distance metrics. The key issue of a shape matching problem is to find an effective shape representation(i.e. shape descriptor) [2]. There are mainly two categories of shape descriptors: contour-based [3–5] and region-based methods [6–10]. These two kinds of method can also be divided into full shape descriptor (e.g. global features) [6–8], © Springer International Publishing Switzerland 2016 D.-S. Huang and K.-H. Jo (Eds.): ICIC 2016, Part II, LNCS 9772, pp. 279–289, 2016. DOI: 10.1007/978-3-319-42294-7_24
280
Wahyono et al.
part-based shape representation (e.g. local features) [9, 11], and combining global and local information [12, 13]. Strategies by decomposing a shape into several sub-regions and extracting local properties were suggested in [9, 11]. Zhang [11] decomposed a shape contour under multiple scales based on a visual perception. For obtaining a contour, they applied morphological operations. Each segment of contour is described by a shape context descriptor. Author [9], rather than extracting features from an edge shape that unsolved occlusion problem, the authors used all points of the shape. Fir
Data Loading...
