Abstract:

To improve the speed of scene matching based on invariant moments, a wavelet transform is used to compress searching space before matching, and then, to reduce the computational complexity of similarity measure at each point to be matched, the moments computation of each sub image is simplified by using ten sum tables in terms of the calculation characteristic of moments during matching. By integrating these two speed up methods, a fast moment based scene matching algorithm is proposed, which decreases the computational cost greatly. Simulation results show that the proposed method takes less time with good precision compared with the pure wavelet transform matching.