The paper presents a method for globally optimal stitching of tiled 3D microscopic image acquisitions. The method uses phase correlation and fast Fourier transform to compute translational offsets between images, allowing for accurate alignment of overlapping tiles. It then globally optimizes the registration by minimizing the sum of all pairwise transfer errors using least squares. This approach avoids error propagation that can occur with consecutive registration steps. Additionally, a non-linear blending technique is applied to compensate for brightness differences between tiles, resulting in artifact-free stitched images. The method works on both 2D and 3D images and does not require prior knowledge of the tile configuration. The implementation is available as an ImageJ plugin distributed as part of the Fiji project. The method is efficient, fully multi-threaded, and applicable to various types of microscopy image data. It has been tested on a variety of tiled microscopy data, including histological 2D images and 3D confocal stacks, with successful results. The globally optimized registration produces high-quality stitched images with minimal displacement and no discarded tiles. The method is open-source and widely available, making it a valuable tool for the analysis of large biological specimens with high-resolution imaging techniques.The paper presents a method for globally optimal stitching of tiled 3D microscopic image acquisitions. The method uses phase correlation and fast Fourier transform to compute translational offsets between images, allowing for accurate alignment of overlapping tiles. It then globally optimizes the registration by minimizing the sum of all pairwise transfer errors using least squares. This approach avoids error propagation that can occur with consecutive registration steps. Additionally, a non-linear blending technique is applied to compensate for brightness differences between tiles, resulting in artifact-free stitched images. The method works on both 2D and 3D images and does not require prior knowledge of the tile configuration. The implementation is available as an ImageJ plugin distributed as part of the Fiji project. The method is efficient, fully multi-threaded, and applicable to various types of microscopy image data. It has been tested on a variety of tiled microscopy data, including histological 2D images and 3D confocal stacks, with successful results. The globally optimized registration produces high-quality stitched images with minimal displacement and no discarded tiles. The method is open-source and widely available, making it a valuable tool for the analysis of large biological specimens with high-resolution imaging techniques.