A Multi-scale Image Matching Recognition Algorithm Based On Phase Correlation

The image matching recognition method of phase correlation is based on the shift characteristics of the Fourier transform. The traditional image matching recognition algorithm has significant influence of the template size upon its matching accuracy and has weak resilience to noises except for the gauss noise. Addressing these shortcomings, we proposed a multi-scale matching recognition method based on phase correlation, combined with wavelet transform and edge detection. The algorithm, processed the reference image and the template image in different scales with such steps: decomposition, denoising, reconstruction, edge detection and the Fourier transform, phase correlation. Hence, it overcome the dependence upon template size effectively and improve the reliability and the resilience of various noises. Finally, we verified the algorithm with a real ground shooting image as the reference image and an intercepted part as the template image. The results have shown that the proposed approach is better than the traditional image matching method.