This body of work demonstrates applications of signal processing, specifically image processing, in the fields of neuroimaging and art conservation. Understanding of any kind of data relies on identification of relevant and meaningful features to describe and represent them. In this dissertation, various feature extraction procedures were applied, tested and compared with the goals of: (a) understanding neural changes following an intervention for late-stage stroke recovery and the correlates of behavior on the basis of brain imaging; (b) automatically identifying neural differences between subgroups of early-stage stroke population with subtle (non-clinical) deficits using brain imaging; (c) delineating neural correlates of changes in cognitive and inflammatory outcomes associated with surgical older adults based on their brain scans; (d) automating characterization of ink strokes found in Vincent van Gogh’s art work. These problems were modeled as cases of classification or estimation/prediction with the help of data-inspired as well as data-driven approaches. The core of this thesis, however, revolved around investigation of feature extraction methods to enhance the notion of connectivity in the brain based on functional neuroimaging. Functional connectivity is a critical measure utilized to describe not only the normal population but also pathological groups. Typically characterized by Pearson’s correlation, the aim was to augment this definition of functional connectivity by incorporating alternative metrics from time-, frequency-, wavelet-domains, linear, non-linear measures, similarity and dissimilarity measures. Findings revealed that for differentiation between population groups (validated by machine learning classification) and association between brain and behavior (validated by standard statistical regression), a multi-metric definition could be more comprehensive and enhance the understanding of functional connectivity in the brain. The potential implications of this work are discussed and include: (i) alternative measures may offer complementary information when combined, thus, augmenting the notion of functional connectivity; (ii) the patterns of functional connectivity, as currently understood by canonical brain networks, may vary on the basis of the measure used to define it.