In this work we present a sampling scheme that uses feature-location information to compactly represent the data. Traditional Nyquist sampling leverages compact frequency support to form the representation, but it ignores location when doing so. Instead, our location-oriented method (LOM) uses coarse location estimates to allow a reduced-rate representation of fine-scale data. We apply a model of local symmetry to the fine-scale data, motivated by features in natural signals. We present an analysis of the concepts behind LOM as well as performance results on synthetic and natural signals.