A novel hybrid homomorphic fuzzy filter for speckle noise reduction

Purpose: The presence of speckle noise and artifacts make delineation of transthoracic echocardiographic (TTE) images quite difficult and challenging; the edges are to be preserved while removing noise. To address issues, a novel speckle reduction technique is being proposed and analyzed. Methods: Three fuzzy filters based on median and moving average concepts are experimented in homomorphic domain and the best one is further fine tuned for applications on TTE images. The proposed hybrid homomorphic Fuzzy (HHF) filter is the sequential integration of homomorphic fuzzy filter with anisotropic diffusion. The denoising characteristics of HHF filter are compared with ten existing techniques tested in homomorphic and other seven in non-homomorphic domain using seven different performance parameters along with visual quality assessment. Experimentations are performed on TTE images acquired in two parasternal and three apical views, since image in one view may not very precisely speak of underlying valvular abnormality. Results: The edge preservation capability is increased by many fold (around 8 times) upon integration of homomorphic fuzzy filter with the anisotropic diffusion filtering technique. Beta metric, figure of merit and structure similarity indices are all greater than 0.97 for proposed HHF filter. Conclusions: The performance of proposed HHF filter is superior in comparison to seventeen state-of-art denoising techniques in terms of all seven performance parameters. Noise is reduced with the edges and structure of TTE images well preserved. © 2014 Korean Society of Medical and Biological Engineering and Springer.