Techniques are described for passive three-dimensional image sensing based on referential image blurring. For example, a filter mask is integrated with a lens assembly to provide one or more normal imaging bandpass (NIB) regions and one or more reference imaging bandpass (RIB) regions, the regions being optically distinguishable and corresponding to different focal lengths and/or different focal paths. As light rays from a scene object pass through the different regions of the filter mask, a sensor can detect first and second images responsive to those light rays focused through the NIB region and the RIB region, respectively (according to their respective focal lengths and/or respective focal paths). An amount of blurring between the images can be measured and correlated to an object distance for the scene object. Some embodiments project additional reference illumination to enhance blurring detection in the form of reference illumination flooding and/or spotted illumination.