Building skeleton models via 3-D medial surface/axis thinning algorithms
CVGIP: Graphical Models and Image Processing
Pattern Recognition Letters
A Skeletal Measure of 2D Shape Similarity
IWVF-4 Proceedings of the 4th International Workshop on Visual Form
Qualitative similarity measures-The case of two-dimensional outlines
Computer Vision and Image Understanding
Skeleton comparisons: the junction neighbourhood histogram
Proceedings of the 11th ACM symposium on Document engineering
Hi-index | 0.00 |
This paper is about off-line handwritten text comparison of historic documents. The long-term motivation is the support of palaeographic research, in particular to back up decisions as to whether two handwritings can be ascribed to the same author. In this paper, a first fundamental step is presented for extracting relevant structures from handwritten texts. Such structures are represented by skeletons, due to their resemblance to original writing movements. Core result is an approach to the simplification of skeleton structures. While skeletons represent constitutive structures for a wide variety of subsequent algorithms, simplification algorithms usually focus on pruning branches off the skeleton instead of simplifying the skeleton as a whole. By contrast, our approach reduces the amount of elements in a skeleton based on a global error level, reducing the skeleton's complexity while keeping its structure as close to the original exemplar as possible. The results are much easier to analyse while relevant information is maintained.