Progressive compression is one of the most effective multi-resolution schemes for sending large 3D meshes in a network environment. However, if we change the 3D meshes, i.e., adding watermarks or changing the pose of models, the deformed 3D mesh model should be compressed again from scratch and it may take too much time especially for large mesh models. In this paper, we greatly reduce the compression time for deformed models, which we call recompression, when connectivity-preserving deformation is applied to mesh models. We first compress a large mesh progressively using a wavelet-based method, and then extract vertices and wavelet coefficients affected by deformed area using our propagation algorithm. Direct modification acts on the required elements using original matrix filters without additional postprocessing, boosting compression performance. Our method increases the recompression speed up to 17 fold compared with the original speed, even if the entire mesh is deformed. We also show that the speedup increases up to 33 fold when local deformation acts on the local regions comprising 5% of mesh.