Texture-Based Watershed 3D Medical Image Segmentation Based on Fuzzy Region Growing Approach

In this paper, a hybridization technique for multi-dimensional image segmentation algorithm is proposed. This algorithm is the combination of both the region growing and texture-based morphological algorithm of watersheds. An edge-preserving statistical noise reduction method is utilized as a preprocessing phase to calculate a perfect estimate of the image gradient. After that, a preliminary segmentation of the images into primitive regions is generated by employing the region growing. Then, watershed is applied. There are some drawbacks in medical mage study, when watershed is employed after the region growing. The main drawbacks are: over segmentation, sensitivity to noise and incorrect identification of thin or low signal to noise ratio structures. The main issue of over segmentation is controlled by texture local binary pattern (LBP). In addition, this paper has presented experimental outcomes achieved with two-dimensional/three-dimensional (2-D/3-D) magnetic resonance images. Numerical justification of the experimental outcomes is presented and demonstrated the efficiency of the algorithm for segmenting the medical image.