Recursive radiography image denoising

Radiography images are large in size, thus denoising algorithm should be computationally efficient. We propose a homomorphic recursive denoising approach. Proposed denoising was used as a preprocessing step for a radiography image processing algorithm. It was shown that the proposed approach requires only two (50 % less) iterations to produce results that are visually equivalent to four iterations of anisotropic diffusion (AD). Proposed method was compared to AD on a database that consists of 47 clinical radiography images. Objective comparison through structural similarity index shows that the proposed method outperforms AD for various signal-to-noise ratios.