Image sequence change detection via sparse representations

We present a sparsity-based algorithm for image change detection. A series of reference images in conjunction with an orthonormal basis are used to form an overcomplete basis. The mission image can then be represented as a linear combination of reference images and components of the orthonormal basis. We solve for the linear combination that has minimum l¹ norm, encouraging a sparse representation. Components of the solution that lie within the orthonormal basis indicate changes in the mission image. We discuss experiments using this algorithm and the results that were obtained.