This paper considers the problem of Through-the-Wall Radar Imaging (TWRI) from multiple views using Compressed Sensing (CS). The scene reconstruction problem is reformulated in terms of finding a sparse representation of the target locations, consistent with the observations. In contrast to the common approach of first applying image formation to each view and then fusing the single-view images, observations from the different views are combined together into a composite measurement vector and a new dictionary is constructed accordingly. A sparse image representation of the scene is then obtained from the composite measurement vector and the new dictionary using 1 -norm minimization. Experimental results demonstrate that the proposed approach using various standoff distances and perspectives achieves a better performance in terms of detecting targets compared to the alternative approach of image formation followed by fusion.