This paper presents a faster one-dimensional (1-D) phaseonly correlation (POC)-based method for estimations of translations, rotation, and scaling in images. The proposed method is to project twodimensional (2-D) images horizontally and vertically onto 1-D signals, and uses 1-D POCs of the 1-D signals to estimate the translations in images. Combined with a log-polar transform, the proposed method is extended to scaling and rotation estimations. Compared with conventional 2-D and 1-D POC-based methods, the proposed method performs in a lower computational cost. Experimental results demonstrate that the proposed method is capable of estimating large translations, rotation and scaling in images, and its accuracy is comparable to those of the conventional POC-based methods. The experimental results also show that the computational cost of the proposed method is much lower than those of the conventional POC-based methods.
|Number of pages||11|
|Journal||IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences|
|Publication status||Published - 2014|
- Image registration
- Log-polar transform
- Phase-only correlation