Correlative microscopy, especially light and electron microscopy (CLEM), enables the study of cells and subcellular elements in complementary ways, provided a reliable registration between images is efficiently achievable. We propose a general automatic registration method. Due to large discrepancies in appearance, field-of-view, resolution and position, a pre-alignment stage is required before any 3D fine registration stage. We define an intensity-based method for both stages, which leverages a common representation of the two involved image modalities. We report experimental results on different real datasets of 3D correlative microscopy, demonstrating time efficiency and overlay accuracy.