We propose a variational method dedicated to jitter correction of large fluorescence scanner images. Our method consists in minimizing a global energy functional to estimate a dense displacement field representing the spatially-varying jitter. The computational approach is based on a half-quadratic splitting of the energy functional, which decouples the realignment data term and the dedicated differential-based regularizer. The resulting problem amounts to alternatively solving two convex and nonconvex optimization subproblems with appropriate algorithms. Experimental results on artificial and large real fluorescence images demonstrate that our method is not only capable to handle large displacements but is also efficient in terms of subpixel precision without inducing additional intensity artifacts.