This thesis proposes the application of space-frequency transformations to the domain of pose estimation in images. This idea is explored using the Wavelet Transform with illustrative applications in pose estimation for face images, and images of planar scenes. The approach is based on examining the spatial frequency components in an image, to allow the inherent scene symmetry balance to be recovered. For face images with restricted pose variation (looking left or right), an algorithm is proposed to maximise this symmetry in order to transform the image into a fronto-parallel pose. This scheme is further employed to identify the optimal frontal facial pose from a video sequence to automate facial capture processes. These features are an important pre-requisite in facial recognition and expression classification systems. The under lying principles of this spatial-frequency approach are examined with respect to images with planar scenes. Using the Continuous Wavelet Transform, full perspective planar transformations are estimated within a featureless framework. Restoring central symmetry to the wavelet transformed images in an iterative optimisation scheme removes this perspective pose. This advances upon existing spatial approaches that require segmentation and feature matching, and frequency only techniques that are limited to affine transformation recovery. To evaluate the proposed techniques, the pose of a database of subjects portraying varying yaw orientations is estimated and the accuracy is measured against the captured ground truth information. Additionally, full perspective homographies for synthesised and imaged textured planes are estimated. Experimental results are presented for both situations that compare favourably with existing techniques in the literature.