Mean of Median Absolute Derivation Technique for Speckle Noise Variance Estimation in Computerised Tomography Images

Abstract

The accurate estimation of noise variance in an image is the first important stage in image filtering using adaptive filters. In this paper, a new technique for the estimation of speckle noise present in Computerised Tomography (CT) lung image was developed. The development of mean of median absolute derivation technique based on the estimated mean of speckle noise present in CT images is presented. From the result of the simulations, the new technique gave a reasonably accurate estimate of variance of speckle noise present in CT Images. Ten samples of 85x73 CT images corrupted by speckle noise level ranging from 10% to 30% where used as test images. Also, the new technique gave the lowest average speckle noise variance estimation error of 2.53% compared to 12.53% for the Median of Median Absolute Derivative Technique, 18.18% for the Transfer function technique and 37.14% for the Mode of Variance Technique. The simulation software used in the paper is Matrix Laboratory (MATLAB2012). http://dx.doi.org/10.4314/njt.v34i2.22