In this talk, I will discuss some recent and current work at the Image Guidance Laboratories, Department of Neurosurgery, Stanford University in image registration, segmentation, and visualization for surgical planning and guidance. Image registration topics will include the application of point-based registration error theory to the placement of fiducial markers for cranial image-guided surgery (IGS) and the design of optically tracked surgical instruments, a tracked A-mode ultrasound system for noninvasively measuring bone surface points and performing accurate image-to-physical registration for cranial IGS, 2D-3D registration for spinal IGS and radiosurgery, and intensity-based nonrigid image registration. We have used the latter to measure intraoperative brain deformation using iMRI images, correct for motion artifact in contrast-enhanced breast MR images, and generate tissue motion models of the liver using resiratory-gated MR image sequences. Nonrigid registration of images to an atlas can be used to automatically segment images. The accuracy of the segmentation depends on whether the atlas is an individual subject image or an average image created from many subjects, and can be improved by combining segmentations obtained from multiple atlases. Finally, I will show a variety of augmented reality visualization approaches using microscopes, endoscopes, and head-mounted displays.