In this study we investigate spatial dimensions of bousing-market dynamics in the City of Milwaukee by modeling the determinants of housing prices. From the 2003 Master Property data file of the city, two sets of owner-occupied single-family houses were randomly selected (one to construct the models, and the other to test the models). Besides conventional housing attributes, remote-sensing information, in particular the fractions of soil and impervious surface representing degraded neighborhood environment conditions, is added to improve the model. Spatial regression and geographically weighted regression approaches are employed to examine spatial dependence and heterogeneity. Results reveal that these spatial models tend to perform better, especially in terms of model performance and predictive accuracy, than the ordinary least squares estimates.