[A novel method for rotated images registration based on wavelet transform].

Abstract

UNLABELLED: This paper presents a novel approach to image registration which addresses the problem of rotation correction. METHOD: 1. 2-DWT of medical images. 2. Maxima of wavelet coefficients (MWC) were used as feature points. 3. Starting from last level of decomposition and then iteratively for each level going up, find the best match between MWC of input image and rotated MWC of reference image. The best rotation is refined iteratively and with increasing accuracy. RESULTS: The performance of the method was compared to that of the traditional methods and found it to be superior. CONCLUSIONS: Images can be fast registrated with high accuracy and less sensitivity to noise.