Yu-Ning Chang and Herng-Hua Chang

DOI: http://dx.doi.org/10.15379/2409-3394.2015.02.01.4

Published: 29 April 2015

Noise is one of the main sources of quality deterioration not only for visual inspection but also in computerized processing in magnetic resonance (MR) image analysis such as tissue classification, segmentation and registration. Accordingly, noise removal in brain MR images is important for a wide variety of subsequent processing applications. Most existing denoising algorithms require laborious tuning of parameters that are often sensitive to specific image features and textures. Automation of these parameters through artificial intelligence techniques will be highly beneficial. This paper attempts to systematically investigate significant attributes from popular image features and textures to facilitate subsequent automation process. In our approach, a total number of 39 image attributes are considered that are based on three categories: 1) Image statistics. 2) Gray-level co-occurrence matrix (GLCM). 3) Tamura texture features. To obtain the ranking of discrimination in these texture features, a T-test is applied to each individual image features computed in every image based on noise levels, intensity distributions, and anatomical geometries. Preliminary results indicated that the order of significance in the texture features approximately varies in noise, slice, and normality. For distinguishing between noise levels, the features of contrast, standard deviation, angular second moment, and entropy from the GLCM class performed best. For distinguishing between slice positions, the features of mean and variance from the basic statistics class and the coarseness feature from the Tamuraclass outperformed other features.

MRI, image feature, image texture, noise reduction.