Speckle suppression using adaptive frequency compounding in ultrasound nondestructive evaluation of coarse-grained material

Coarse-grained engineering materials exhibiting heterogeneous and anisotropic microstructure are used extensively in many industrial sectors. Inspecting such materials using ultrasound suffers from backscattered signals from grain boundaries resulting in strong speckle noise, which reduces image contrast and increases difficulty in image interpretation. In this paper, we propose an algorithm to adaptively compound images which are acquired in different frequency channels (FC), in order to remove or minimize speckle. Most existing image compounding algorithms are point-wise and based on order statistics estimation, which implicitly assume speckle is spatially and spectrally uncorrelated, leading to limited success in speckle suppression. The proposed method applies spatially variant weightings which are calculated to account for both spatial correlation and inter-FC correlation of speckle noise and flaw echo signals. To validate this method, experiments on two highly scattering samples were conducted: an Inconel 625 block with two side-drilled holes and a steel weld block with an implanted tilted flaw, simulating lack of fusion. Speckle point was found to be less spatially and spectrally correlated than flaw echo, yielding smaller compounding output. By applying adaptive compounding algorithms, flaw-to-noise ratio (FNR) of the Inconel block has achieved a ~45 dB enhancement compared with minimization compounding, whilst ~60dB FNR improvement for the steel weld block is demonstrated. It is concluded that the proposed algorithm can significantly suppresses speckle and enhance flaw characterization without spatial resolution loss.