A New Image Feature Matching Method Based on Wavelet-Structure Moment Invariants

摘要

This paper introduces a new method for recognizing images that may have been translated, scaled and/or rotated. The new method is based on a combination of the structure moment invariants and the wavelet transformation. It is found that after applying the structure moment function to the image, the image features are highlighted. On the other hand, using cubic B-spine wavelet transformation, both global features and local features can be extracted effectively. Experiments demonstrate that the new method outperforms the existing methods such as the improved Hu moments method and the structure moments method. The new method has been successfully applied in a Surface Mount Technology (SMT) system in electronics manufacturing.