Recursive Noise Reduction for Digital Radiography Images

Abstract

Modern digital radiography images are large, suggesting that a denoising algorithm should be computationally efficient. In this paper, we propose a homomorphic recursive noise reduction approach. The proposed method was compared to anisotropic diffusion on a database consisting of 47 clinical radiography images, and was subjectively deemed better at noise reduction. The proposed approach was shown to require 50 % fewer iterations to produce subjectively the same results. Objective comparison of the algorithms through a structural similarity index and peak signal-to-noise ratio showed that the proposed method outperforms anisotropic diffusion for various signal-to-noise ratios.

Keywords

• digital radiography
• homomorphic filtering
• image noise reduction
• recursive filtering