Eddy current stimulated thermography is an emerging technique for non-destructive testing and evaluation of conductive materials. However, quantitative estimation of the depth of subsurface defects in metallic materials by thermography techniques remains challenging due to significant lateral thermal diffusion. This work presents the application of eddy current pulse compression thermography to detect surface and subsurface defects with various depths in an aluminum sample. Kernel Principal Component analysis and Low Rank Sparse modelling were used to enhance the defective area, and cross point feature was exploited to quantitatively evaluate the defects depth. Based on experimental results, it is shown that the crossing point feature has a monotonic relationship with surface and subsurface defects depth, and it can also indicate whether the defect is within or beyond the eddy current skin depth. In addition, the comparison study between aluminum and composites in terms of impulse response and proposed features are also presented.