Abstract
Abstract. In this paper, we present a computationally efficient technique for edge preserving in medical image smoothing, which is developed on the basis of dynamic programming multi-quadratic procedure. Additionally, we propose a new non-convex type of pair-wise potential functions, allow more flexibility to set a priori preferences, using different penalties for various ranges of differences between the values of adjacent image elements. The procedure of image analysis, based on the new data models, significantly expands the class of applied problems, and can take into account the presence of heterogeneities and discontinuities in the source data, while retaining high computational efficiency of the dynamic programming procedure and Kalman filterinterpolator. Comparative study shows, that our algorithm has high accuracy to speed ratio, especially in the case of high-resolution medical images.
Highlights
Low–level image processing is a functional step in almost every medical image analysis system
Wiener filtering (WF), non – linear Total Variation (TV), Anisotropic Diffusion Filter (ADF), Fast Bilateral Filter (FBF), the Bayesian least squares with Gaussians Scale Mixture (BLS-GSM) and proposed algorithms were tested
In our experiments we compare proposed approaches with the other filters for MR images as Wiener filtering (WF), non – linear total variation (TV), anisotropic diffusion filter (ADF), fast bilateral filter (FBF), the Bayesian least squares with Gaussians Scale Mixture (BLS-GSM)
Summary
Low–level image processing is a functional step in almost every medical image analysis system It is a prerequisite for proper data interpretation, diagnosis and suggestion the corresponding treatments. Medical images (example: Magnetic Resonance, Ultrasound, Computed Tomography, X–Ray), may be corrupted by a disruptive noise during acquisition and transmission process and the essential requirement for every noise reduction procedure is to preserve local image features for an accurate and effective diagnosis. Many edge-preserving denoising methods for medical images have been proposed in the literature. We develop new non–iterative parametric procedure for edge preserving in image smoothing. This procedure can effectively remove Gaussian noise as well as Rician noise (Dekker and Sijbers, 2014), typical for MR images, with high quality. That our algorithm has the best accuracy to speed ratio, especially in the case of highresolution images
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
