Abstract:
"Ultrasound images are corrupted by a multiplicative noise – the speckle –, which
makes hard high-level image analysis. In order to solve the difficulty of designing a
filter for an effective speckle removing, we propose a new approach for de-noising
images while preserving important features. This method combines a data misfit
function based on Loupas et al . model and a Weighted Total Variation (WTV) function
as a multiplicative factor in the cost functional. The de-noising process is performed
using a multiplicative regularization method through an adaptive window whose
shapes, sizes and orientations vary with the image structure. Instead of performing
the smoothing uniformly, the process is achieved in preferred orientations, more in
homogeneous areas than in detailed ones to preserve region boundaries while
reducing speckle noise within regions. Quantitative results on synthetic and real
images have demonstrated the efficiency and the robustness of the proposed method
compared to well-established and state-of-the-art methods. The speckle is removed
while edges and structural details of the image are preserved."
