This paper presents a robust and inverse consistent method for image registration, which is essential for accurate alignment of neuroimaging data. The method is designed to handle differences such as jaw movement, differential MR distortions, and true anatomical changes. It guarantees inverse consistency, can deal with different intensity scales, and automatically estimates a sensitivity parameter to detect outlier regions. The resulting registrations are highly accurate due to their ability to ignore outlier regions and show superior robustness to noise, intensity scaling, and outliers compared to state-of-the-art tools like FLIRT or SPM. The method is based on robust statistics and inspired by Nestares and Heeger (2000). It uses a symmetric setup where both images are transformed halfway towards each other, ensuring symmetry in the registration process. The method incorporates a global intensity scale parameter to adjust for different intensity scalings, which is particularly useful in longitudinal data. It also uses a Tukey's biweight function to handle outliers and robustly estimate parameters using iteratively reweighted least squares. The algorithm involves constructing a Gaussian pyramid for multiresolution processing, an initial alignment using moments, and iterative refinement at different resolution levels. The registration process ensures that both images are resampled into a halfway space to maintain symmetry and avoid bias. The method is tested on synthetic and real data, demonstrating its effectiveness in motion correction and longitudinal analysis. The software implementing this method is publicly available as part of the FreeSurfer package.This paper presents a robust and inverse consistent method for image registration, which is essential for accurate alignment of neuroimaging data. The method is designed to handle differences such as jaw movement, differential MR distortions, and true anatomical changes. It guarantees inverse consistency, can deal with different intensity scales, and automatically estimates a sensitivity parameter to detect outlier regions. The resulting registrations are highly accurate due to their ability to ignore outlier regions and show superior robustness to noise, intensity scaling, and outliers compared to state-of-the-art tools like FLIRT or SPM. The method is based on robust statistics and inspired by Nestares and Heeger (2000). It uses a symmetric setup where both images are transformed halfway towards each other, ensuring symmetry in the registration process. The method incorporates a global intensity scale parameter to adjust for different intensity scalings, which is particularly useful in longitudinal data. It also uses a Tukey's biweight function to handle outliers and robustly estimate parameters using iteratively reweighted least squares. The algorithm involves constructing a Gaussian pyramid for multiresolution processing, an initial alignment using moments, and iterative refinement at different resolution levels. The registration process ensures that both images are resampled into a halfway space to maintain symmetry and avoid bias. The method is tested on synthetic and real data, demonstrating its effectiveness in motion correction and longitudinal analysis. The software implementing this method is publicly available as part of the FreeSurfer package.