In advanced manufacturing, surface topographical designs with deterministic freeform and embedded structures have proven to contain effective, additive functionalities. These surfaces need to be geometrically characterised regarding the designed form and structures. However, this is problematic since existing characterisation techniques such as polynomial form removal, Gaussian/spline/wavelet filtration, field-based statistical parameterisation, spectral and fractal analysis do not provide satisfying results. In this paper, we, therefore, propose to characterise the complex surfaces in T-spline spaces, i.e. basis spline spaces along with T-junctions, using an efficient T-spline fitting algorithm. Several case studies show that the proposed method is compatible and has notable potentials for the challenging characterisation tasks, including non-Euclidean freeform removal, edge-reserving filtration with multiscale analysis, scattered data interpolation and smoothing, and smart large-data downsampling or compression.