This paper presents methods and results of small-area population estimation using a combined Light Detection And Ranging (LiDAR), Landsat Thematic Mapper (TM) and parcel dataset for a study area in Denton, Texas, USA. A normalized digital surface model (nDSM) was created from a digital surface model (DSM) and a digital elevation model (DEM) built from LiDAR point data. Residential and commercial parcels were selected from parcel data and used as a mask to remove non-residential and non-commercial pixels from the nDSM. Classification results of residential areas from Landsat TM images acquired on two dates were used to further refine the nDSM. Using continuous and random census blocks as samples, building count, building area and building volume were calculated from the nDSM through mathematical morphological operations, zonal statistics, data conversion and spatial joining in a geographic information system (GIS). Combined with census 2000 data, a total of 10 ordinary least squares (OLS) regression models and geographically weighted regression (GWR) models were built and applied to the census blocks in the study area. Finally, accuracy assessments were carried out. The results show that the sign and magnitude of the relative estimation errors at the census-block level lead to underestimation of the total population in the study area. Possible reasons for the relatively low accuracies and problems for further investigation are also discussed.