The importance of seismic diffractions has been recognized in subsurface imaging. Diffractions might contain useful information about small-scale geological structures and heterogeneities, e.g. faults, pinch-outs, thin lenses, fractures etc and can be very helpful for prospecting and development of fractured reservoirs. One of the most challenging task in diffraction imaging is the separation of diffractions from reflections. In this work, we present a new diffraction filter based on the projected (first) Fresnel zone (PFZ) to enhance diffraction separation and imaging. We found that the difference in the width of the PFZ appears to be a very consistent parameter for diffraction separation and thus, can be used as a robust diffraction filter. We derived the width of the PFZ from wavefront attributes, which are determined using the common-reflection-surface method. Applications to a simple gradient model as well as a complex synthetic data set validate our approach.