Abstract

We present and compare two multi-scale shape matching schemes: Chi-square distance based scheme and pyramid matching mode based scheme. We define a shape as a set of points. Multi-scale shape matching includes two steps: multi-scale feature extraction and point correspondence. We define a hybrid feature for every point by combining a global multi-scale shape context feature and a local variation feature. The two schemes have a difference in the computation of multi-scale shape context feature distance: the Chi-square distance based scheme directly sums up weighted Chi-square distances at different scales while the pyramid matching mode based scheme utilizes a multi-scale pyramid matching mode. Experimental results based on Frenkel and Kimia databases show that: (1) the pyramid matching mode based scheme can achieve robust and often better performance than the Chi-square distance based scheme; (2) the proposed two multi-scale schemes can achieve averagely better results than the single scale schemes.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call