A modified self-correction subpixel registration algorithm for passive millimeter wave images

For multi-frame image super-resolution, an accurate subpixel registration algorithm is necessary. In this paper, a modified self-correction registration algorithm for passive millimeter wave images is proposed for accurate robust registration. This algorithm uses the feature region to realize registration other than the whole image for computationally efficient. After obtained the feature points by Harris corner detector, the normalized cross-correlation feature matching is done for feature region pairs extraction. Then, the least-squares method and the extension of phase correlation algorithm are employed to accomplish the pixel and subpixel registration of regions respectively. At last, the spatial cluster is introduced for self-correction to ensure the robustness and accuracy. Simulation results show that the algorithm has good robustness and the accuracy has reached to 0.01 pixels.