Automatic Image Registration of Multimodal Remotely Sensed Data With Global Shearlet Features

Abstract

Automatic image registration is the process of aligning two or more images of approximately the same scene with minimal human assistance. Wavelet-based automatic registration methods are standard, but sometimes are not robust to the choice of initial conditions. That is, if the images to be registered are too far apart relative to the initial guess of the algorithm, the registration algorithm does not converge or has poor accuracy, and is thus not robust. These problems occur because wavelet techniques primarily identify isotropic textural features and are less effective at identifying linear and curvilinear edge features. We integrate the recently developed mathematical construction of shearlets, which is more effective at identifying sparse anisotropic edges, with an existing automatic wavelet-based registration algorithm. Our shearlet features algorithm produces more distinct features than wavelet features algorithms; the separation of edges from textures is even stronger than with wavelets.Our algorithm computes shearlet and wavelet features for the images to be registered, then performs least squares minimization on these features to compute a registration transformation. Our algorithm is two-staged and multiresolution in nature. First, a cascade of shearlet features is used to provide a robust, though approximate, registration. This is then refined by registering with a cascade of wavelet features. Experiments across a variety of image classes show an improved robustness to initial conditions, when compared to wavelet features alone.

Agile multi-scale decompositions for automatic image registration

“…15,16 These algorithms were based on the observation that while wavelet-based registration was usually accurate and efficient, even for large classes of multi-modal images, it was often not robust to the choice of initial registration guess. By examining the features produced in the wavelet registration algorithm currently used at NASA, 7 it was hypothesized that this was due to the speckled and diffuse features often produced by wavelet algorithms.…”

“…This observation is in accordance with the theory of wavelets. In particular, wavelets are known not to be optimally sparse for an important class of images, namely cartoon-like images, which are composed of smooth images with smooth edges; see 16 for a fuller exposition on the mathematical limitations of wavelets for image registration, and 10 for a more general discussion.…”

“…Many remotely sensed images fall to some extent into this regime, which suggested the use of such anisotropic dictionaries in the registration of remotely sensed images. Recent work 15,16 proposed both a stand-alone shearlet feature registration algorithm, and a novel two-stage registration algorithm using shearlets for the first stage of registration, and wavelets to refine. The hybrid algorithm acquired a robust, though occasionally somewhat inaccurate, first stage shearlet registration, which was refined with the less robust, but sometimes more precise, wavelet registration.…”

Optimal Weight Design Approach for the Geometrically-Constrained Matching of Satellite Stereo Images

“…Similarly, the normalization values (scale and offset) for the geodetic coordinates of the ground point will be used in the linearized model in Equation (10). These normalization values are provided in the RPC files of stereo images.…”

“…It should be noted that all coordinates in Equation (10) are denormalized. The relationship between the vector of unknown parameters and observations in the geometric constraint, according to Equation (10), can be established:…”