An Efficient Method for Brain Image Preprocessing with Anisotropic Diffusion Filter & Tumor Segmentation

Abstract

The brain is central controlling system in human body. The internal structure and functioning of brain change due to any kind of enlargement of the brain cells which is diagnosed as a brain tumor. Early detection of tumor is critical for better treatment options. In this paper, brain magnetic resonance (MR) image is preprocessed and watershed algorithm is applied to separate the image objects and to segment the tumor. Image preprocessing includes anisotropic diffusion filter (ADF), skull stripping and contrast enhancement steps. ADF eliminates noise content from MR images and preserves the edges of existing objects, followed by skull stripping to remove the nonbrain tissues and contrast enhancement to improve visual quality. A morphological opening operation is performed to sharpen the tumor region. In the segmented images, the statistical features and textural features are extracted. A set of performance measure is used to evaluate the system performance and comparison is made with the existing literature. The proposed system performs efficiently with high PSNR value of 17.0195 and lesser computation time of 2.04 s.