The use of computerized image processing techniques for the extraction of clinically important diagnostic information and the enhancement of medical digital medical images is becoming increasingly more important. A major problem in medical imaging is the inability to adequately separate interfering physiological processes or anatomical features from the physiological process or anatomical structure under investigation. We have developed and reported a new linear filter which has the unique ability to segment a feature of interest from an interfering feature in a sequence of images of the same anatomical site, even when the gray levels of the pixels representing the interfering process vary through the sequence of images. We call this technique """"""""eigenimage filtering"""""""". The purpose of this project is to continue the development and investigation of this filtering technique to include multiple interfering processes and its application to Magnetic Resonance Imaging. After formulating the filtered image contrast criterion in the form of a contrast ratio, we use Rayleigh's principle to recast the ratio in form of a generalized eigenvalue problem. According to Rayleigh's principle, the eigenvector associated with the dominate eigenvalue of the generalized eigenvalue problem is the vector that will maximize the contrast ratio. The arguments of this vector are then used as the weighting factors in the linear filter. The applications to Magnetic Resonance images will include scene segmentation with emphasis on feature selection. A specific application will be the development of volume calculations from magnetic resonance images in which partial volume effects will be taken into account. The ability to separate and segment features of interest from those that interfere with their observation will increase the availability of quantitative data which, if properly interpreted, can significantly improve the diagnostic information available to the clinician.