Use of Autostitch for automatic stitching of microscope images Bin Ma a, * , Timo Zimmermann b , Manfred Rohde c , Simon Winkelbach d , Feng He a , Werner Lindenmaier a , Kurt E.J. Dittmar a a Division of Molecular Biotechnology, German Research Centre of Biotechnology, Mascheroder Weg 1, D-38124 Braunschweig, Germany b Advanced Light Microscopy Facility, EMBL Heidelberg, Heidelberg, Germany c Department of Microbial Pathogencity, German Research Centre of Biotechnology, Braunschweig, Germany d Institute for Robotics and Process Control, Technical University of Braunschweig, Braunschweig, Germany Received 9 June 2006; received in revised form 26 July 2006; accepted 28 July 2006 Abstract Image stitching is the process of combining multiple images to produce a panorama or larger image. In many biomedical studies, including those of cancer and infection, the use of this approach is highly desirable in order to acquire large areas of certain structures or whole sections, while retaining microscopic resolution. In this study, we describe the application of Autostitch, viz. software that is normally used for the generation of panoramas in photography, in the seamless stitching of microscope images. First, we tested this software on image sets manually acquired by normal light microscopy and compared the performance with a manual stitching approach performed with Paint Shop Pro. Secondly, this software was applied to an image stack acquired by an automatic microscope. The stitching results were then compared with that generated by a self- programmed rectangular tiling macro integrated in Image J. Thirdly, this program was applied in the image stitching of images from electron microscopy. Thus, the automatic stitching program described here may find applications in convenient image stitching and virtual microscopy in the biomedical research. # 2006 Elsevier Ltd. All rights reserved. Keywords: Autostitch; Image stitching; Automatic microscopy; Image masaicing; Virtual microscopy 1. Introduction Image stitching is the process of combining multiple images to produce a panorama or larger image (Shum and Szeliski, 1997; Chen and Klette, 1999; Zomet et al., 2006) and has found many applications in the acquisition of high resolution images. For example, for bright-field or fluorescence microscopy, analysis of the whole section of several centimetres at high resolution cannot be performed, even at a low power objective and even if cameras with high resolution are applied. Image stitching is a powerful approach for solving this problem by creating a single composite image by overlapping multiple images acquire from different parts of the section with high resolution. The image sets needed for the generation of the large single images can be acquired either manually with a normal microscope or with an automatic microscope with a motorised x, y stage. If small numbers of images are handled, the image sets can be acquired manually and stitched together either manually with normal image processing software or with automatic stitching software. In this case, manual stitching can produce good results, although it is difficult and time consuming. If large numbers of sections are handled, manual stitching becomes almost impossible and automatic mosaicing must be applied. In addition, the acquisition of data must be performed by a microscope system with motorised x, y stage. If stage moves along the x, y direction precisely, the mosaicing can be performed through Tile Scan, i.e. by arrangement of the images one by one in their natural position. This approach can be applied in confocal microscopy to generate a large single image. However, in practice, the motorised stage does not move as exactly as we had expected and this leads to misalignment of adjacent images resulting in the failure of Tile Mosaicing to produce a seamless image. Computer software is often used to interpolate the final image if the component images are not in precise alignment. Therefore, mosaicing with blending, i.e. assembling a set of overlapped images to generate seamless images, must be performed (Bhosle et al., 2002). www.elsevier.com/locate/micron Micron 38 (2007) 492–499 * Corresponding author. Tel.: +49 531 6181 293; fax: +49 531 6181 202. E-mail address: bma@gbf.de (B. Ma). 0968-4328/$ – see front matter # 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.micron.2006.07.027