InSAR image co-registration based on phase correlation and log polar transforms

The paper presents an improved subpixel image registration algorithm for InSAR image based on phase correlation and log polar transform. An improved registration technique is proposed that phase unwrapping is not necessary when two images differ only by no more than one pixel translational shift. The set of experiments based on real images demonstrate that the proposed algorithm can estimate image translation with 1/100-pixel accuracy, which leads to 0.5 degree accuracy in rotation estimation and to 1/1000 scale accuracy in scale factor estimation. The experiments also show a robust to noise.