Magnetic Resonance Cholangiopancreatography (MRCP) is an evolving MRI technique designed for the imaging of the biliary tree, a system of narrow ducts that collect bile, produced within the liver, store it in the gall bladder, and deliver it into the small intestine as needed. Current MRCP protocols, used to diagnose problems in this ductal system, generate cluttered and noisy, low resolution, non-isometric volume data, often with significant intensity non-uniformities. This combination of undesirable characteristics presents particular challenges for the application of automated image analysis techniques. This thesis examines the development, characterisation, and testing of novel and efficient pre-segmentation filtering procedures designed to achieve increased robustness and precision in the subsequent segmentation and analysis of the biliary tree from MRCP data. A focused set of image preprocessing algorithms has been developed so as to facilitate the operation of non-complex segmentation and computer assisted diagnosis (CAD) procedures. Most notable in this regard are a number of novel techniques designed to address the key areas of this image processing task. These techniques consist of: • a new histogram preserving approach to inter-image and intervolume intensity non-uniformity correction, • a highly versatile adaptive smoothing filter, implemented as an oriented, scaled and shaped ellipsoid filter mask, • the downhill filter, an efficient new algorithm for morphological reconstruction by dilation, and • a novel approach to the reconstruction of fine branching structures in noisy volume data. Through this combination of flexible and efficient preprocessing algorithms, an effective route towards robust MRCP segmentation and analysis, and routine CAD in the assessment of the biliary tree from MRCP is presented.