A Novel Image Registration Algorithm Using Wavelet Transform and Matrix-Multiply Discrete Fourier Transform
Cui Zhou,Gui Zhang,Zefa Yang,Jinghong Zhou
DOI: https://doi.org/10.1109/lgrs.2020.3031335
IF: 5.343
2022-01-01
IEEE Geoscience and Remote Sensing Letters
Abstract:Nearly all existing image registration algorithms are based on the full information (e.g., intensity and/or features) carried by the images being registered. Such a full-image-based strategy can achieve high-precision registration (e.g., at the subpixel level) but at the expense of long computation times, especially for high spatial resolution remote-sensing (RS) images. This letter presents a novel subimage-based registration method that is expected to dramatically reduce the time consumption but still meet subpixel accuracy requirements. The core idea behind this method is that the transform pattern of the reference and sensed images under a global transformation is theoretically similar to that of subimages derived from the full image, and vice versa. Hence, this method first divides the original reference image into subimages with nearly equal sizes, and extracts the subimage with the maximum wavelet coefficient summation to serve as a new reference image using the wavelet transform technique. Then, the two-step methods based on the phase correlation and matrix-multiply discrete Fourier transform (DFT) algorithms are applied to register the sensed image to the reference subimage. Finally, the proposed method is tested with simulated data sets. The results indicate that this method can perform subpixel registration as achieved by full-image-based methods but with the benefit of much shorter computation times (e.g., approximately 89% compared with a widely used rapid registration algorithm named enhanced phase correlation).
imaging science & photographic technology,remote sensing,engineering, electrical & electronic,geochemistry & geophysics