Detecting and mapping localized near surface inhomogeneities is an important problem in a variety of applications such as engineering, site investigation, environmental studies and archaeology. When an inhomogeneity has a linear character the use of a point diffractor model could result in a poor image. To improve the image of near surface linear inhomogeneities we propose an imaging technique based on a new representation of diffraction travel-time surface in 3D space. Our approach suggests that the linear diffractor may be situated along any azimuth and horizontal location in 3D space and its imaging is performed by stacking seismic energy along all possible linear travel time diffraction surfaces. This approach involves a multipath summation which is performed for all possible velocity values within a specific range and therefore does not require prior information of the subsurface velocities. This paper summarises the theory and shows a real data example in which the diffraction section produced represents a reliable image of a linear diffractor.