Subjective speckle suppression for 3D measurement using one-dimensional numerical filtering

Laser speckle projection is a reliable method to generate statistical illumination patterns for 3D reconstruction purposes as in stereo photogrammetry. This type of pattern has several advantages compared to incoherent methods. However, the biggest disadvantage is given by the coherent noise, the so-called "subjective speckles," developing when a coherently illuminated surface is imaged by a lens system. Some experimental techniques have been published already, being costly in measurement time or leading to loss in light intensity and/or depth of field. In this work we want to present numerical one-dimensional filtering techniques that reduce this kind of noise and increase the performance of 3D reconstruction, while no experimental changes to the classical speckle projection technique have to be made. Therefore, a model describing the expectable contrast reduction is derived, the dependency between filter orientation and setup geometry is investigated, and results from simulations and real experiments are shown. It is found that for small filter sizes the results can be improved independent of the filter, but that in the general case a vertical orientation of the filter towards the setup geometry is most useful.