Abstract
This paper presents a novel approach to planar shape recognition using wavelets. There are two stages, called representation stage and recognition stage, in the proposed method. In the representation stage, in order to extract the features of a shape, a set of wavelet basis are investigated, and wavelet decomposition strategy from the orientation function of the boundary curve of the shape are then performed. The representation of a shape is achieved with a collection of multiscale feature set (MFS) which consists of the scale parameters of wavelet function, the positions where the dominant feature take place, a similarity measure, etc. In the recognition stage, the test shape is compared with various model shapes stored in a database by computing the distance of their MFSs, and the one with the minimum distance is chosen as the correct matching of the test shape. Experimental results obtained with the proposed scheme are encouraging which demonstrates the effectiveness and robustness of the approach.
Original language | English |
---|---|
Pages (from-to) | 201-216 |
Number of pages | 16 |
Journal | Computers and Mathematics with Applications |
Volume | 39 |
Issue number | 5-6 |
DOIs | |
State | Published - 08 02 2000 |
Keywords
- Multiscale features
- Shape recognition
- Wavelet transform