Distance transformations in digital images
Computer Vision, Graphics, and Image Processing
Hierarchical Decomposition and Axial Shape Description
IEEE Transactions on Pattern Analysis and Machine Intelligence
Journal of Visual Languages and Computing
| Hi-index | 0.00 |
A shape decomposition method for non-elongated figures is proposed. The method allows to obtain structural descriptions that are widely invariant with respect to non-significant shape changes occurring in rotated or noisy instances of a same figure. The detection of the significant parts composing a figure is based on a suitable definition of shape primitives and is performed by exploiting the information associated to the skeleton pixels. In this process the regions having higher perceptive relevance are first identified and extracted from the image. Then, starting from this initial decomposition, the remaining parts of the figure are detected together with the structural relations among them. The proposed decomposition scheme is particularly appropriate for building structural descriptions in terms of attributed relational graphs. The experimental results obtained by using a large set of figures confirmed the robustness of the proposed approach and the stability of the achievable decompositions.