Rigid Image Registration Technology Based On Fractional Fourier Transform

As a kind of image processing techniques, image registration referring to the alignment or superposition of two images obtained under different conditions, has been widely used in the field of medical diagnosis, computer vision and so on.The image registration algorithm based on transform domain with low computational cost, good resistance to noise and ease of implementation has been applied widely in different fields. Traditional algorithms based on transform domain mainly refer to Fourier transform. The concept of transform order has been introduced into Fractional Fourier transform(FRFT)—a generalized form of the traditional Fourier transform. FRFT not only maintains good properties of Fourier transform, but also has its own advantages, which contains time domain and frequency domain information at the same time. In this paper, regarding the rigid image registration based on Fourier transform as the breakthrough point, it puts focus on the application of fractional Fourier transform in rigid image registration, and explores deeply the image registration in rotation and translational transform. The main researches are following:1. It first elaborates the development of image registration, and states its research significance and research status at home and abroad. Then it summarizes some mainly used algorithms about image registration. The definition, property and mainly-used discrete algorithms of one-dimensional fractional Fourier transform are introduced and analyzed before the definition and good properties of the two-dimensional fractional Fourier transform are brought up are proposed.2. A kind of image registration algorithm based on fractional Fourier transform and phase correlation technology is proposed. The study has deduced the time shift and rotation invariance—two properties of the two-dimensional fractional Fourier transform, relying on which images of translational and rotational transform are resisted respectively. Compared with traditional image registration algorithm based on Fourier transform, simulation results prove that this algorithm has better results for images with more serious noise pollution.3. A kind of rigid image registration technique based on fractional Fourier transform and simplified particle swarm optimization algorithm(SPSO) is put forward in this paper. Based on the time shift and rotation invariance of two-dimensional fractional Fourier transform, this paper introduces a more efficient simplified particle swarm optimization algorithm(SPSO), which uses the improved peak signal-to-noise ratio(MPSNR) as similarity criterion to research on images with both translational and rotational transform. Compared with other image registration algorithm based on fractional Fourier transform, this algorithm has certain improvement on accuracy and efficiency. Compared with the registration algorithm based on mutual information, this algorithm has better noise resistance.