This research proposes a framework for reconstructing an image of high dynamic range and high spatial resolution from a sequence of multisampled images. In the framework, there are two key issues: robust motion estimation and appropriate measurement discarding for reconstruction. An approach to increase the robustness of motion estimation for a multisampled image sequence is presented. By extracting the luminance component from each captured image in the sequence, and then enhancing the low intensity regions, conventional motion estimation methods can be applied with good results. The research also presents a novel discarding method of invalid measurements. The proposed spatio-temporal discarding method prevents overdiscarding in the case of extreme exposure differences. Experimental results show the effectiveness of the proposed framework.