Peter N. Crabtree (Air Force Research Laboratory, Space Vehicles Directorate), Collin Seanor (Air Force Research Laboratory, Space Vehicles Directorate), Jeremy Murray-Krezan (Air Force Research Laboratory, Space Vehicles Directorate), Patrick J. McNicholl (Air Force Research Laboratory, Space Vehicles Directorate)
Keywords: SSA
Abstract:
A variety of image registration techniques have been investigated for applications such as image analysis, fusion, compression, enhancement, and creating mosaics. In particular, robust registration is a key component for successful multi-frame processing aimed at super-resolution or high dynamic range imaging. Image registration techniques are broadly categorized as global (area) or feature-based, and can also be classified as being performed in either the Fourier- or spatial-domain. Spatial domain methods are typically used for applications requiring accurate estimation of sub-pixel motion, such as multi-frame super-resolution based on de-aliasing. However, these techniques often rely on the availability of a priori information (good initial guess), and are therefore limited in terms of the dynamic range of the global motion estimates. A Gaussian pyramid approach is one standard method for extending the region of convergence of spatial domain techniques. On the other hand, Fourier domain-based correlation techniques such as the log-polar FFT method provide fast and reasonably accurate estimates of global shifts, rotation, and uniform scale changes, and tend to perform well over a large range of frame-to-frame motion magnitudes. In this paper we explore several possible hybrid algorithms for robust global registration based on combining the log-polar FFT and spatial-domain techniques. This includes the straightforward use of the log-polar FFT algorithm to generate an initial guess for use by a spatial domain algorithm, as well as the intertwining of the two methods by applying both global correlation and spatial domain registration at each relevant step within the log-polar FFT algorithm. In addition, we explore the benefits of normalized gradient correlation in performing the coarse log-polar FFT registration. The use of normalized gradient correlation, as opposed to phase-only correlation, has recently been proposed for improving the log-polar FFT method in terms of robustness and dynamic range of the uniform scale estimates. The results presented here are based on image sequences captured in the laboratory using a CMOS machine vision camera.
Date of Conference: September 11-14, 2012
Track: Poster