<p>The deformation of a soil mass induced by filtration is closely associated with the displacement of filter particles. To gain quantitative particle-scale insights to filter particle settlement, transparent soil imaging was employed during filtration experiments. The experiments considered a finer base layer underlying a coarser filter layer, where filtration was triggered with upward seepage flow. A range of filter-base size ratios that covered the complete spectrum of filtration behaviours were considered, along with various levels of seepage flow to investigate the combined influence of geometric characteristics and hydraulic loading. A novel particle detection algorithm was proposed to obtain geometric information of filter particles from the images acquired for the transparent soil assemblies. The filter particle settlement was quantitatively investigated with the particle detection algorithm, and was found to be positively correlated with the inlet flow velocity and the size ratio of the assembly, which was consistent with expectations. While the filter susceptibility has been conventionally evaluated by the size ratio, variability in local filter susceptibility was observed, which was correlated to the local pore geometry of the filter. Local pore geometry was quantitatively characterised by a newly proposed cumulative linear filter porosity associated with detected filter particles that were obtained from the transparent soil images. The observation of local susceptibility highlights the importance of investigating the internal erosion process at the particle scale, and the unique capability of transparent soil imaging in the quantitative investigation of particle behaviour.</p>