A New Shape Matching Measure for Nonlinear Distorted Object Recognition

In this paper, we present a new approach for hand-written character and digit recognitions based on shape descriptor and the Hausdorff Context. We start at finding the corresponding points between two shapes by using a modified shape context. We then use these correspondences as key geometric points for shape alignment with the Thin Plate Spline (TPS) model. After the transformation has been applied completely, the distance between two shapes is computed by a new distance measure, the Hausdorff Context. We achieve a very high recognition rate of 98.7% on 268 images.

• Type: Conference
• Year: 2003
• Where: DICTA
• Authors: Sanun Srisuk, Marut Tamsri, Rerkchai Fooprateepsiri, Pipat Sookavatana, Khamron Sunat
• Keywords: DICTA 2003, DICTA 2007, Hausdorff Context, Modified Shape Context, Shape Descriptor