The goal of the segmentation of brain images is to separate the images in different non-compatible homogenous areas reflecting the numerous anatomical structures. Brain segmentation by magnetic resonance has numerous implications for diagnosing brain disorganizations such as Alzheimer's, Parkinson-related syndrome among others. However, it is not an simple job to automatically segment the MR image. The main motive of this study is to provide a better segmentation approach for the segment of the ROI (Region Of Interest) region from the MRI image by solving the issues that currently exist in the literary works. MRI segmentation is not a trivial task, because acquired MR images are imperfect and are often corrupted by noise and other image artifacts. The variety in technologies for image processing has contributed to the creation in numerous image segmentation techniques. That is because there is no universal approach, nor are all methods necessarily appropriate for a specific form of picture suitable for all pictures. Other approaches still use the gray level histogram, for example, while others integrate detailed spatial picture details for bleeding conditions. Some methods use statistical techniques, but some do incorporate existing information to enhance segmentation efficiency. Some methods utilize probabilistic or fuzzy methods. Yet there are certain inconveniences of all the current processes. Therefore, we have intended to propose a new segmentation approach for the ROI region segmentation. The proposed work comprised of three phases namely preprocessing, edge detection and segmentation. At first, the MRI images are extracted from the database and that each of the input images is enhanced by applying a high pass filter. After completing the preprocessing method, the enhanced canny edge detection (ECED) approach is used to enhance the image. After that, the images are given to the modified watershed segmentation (MWS) algorithm which separates the ROI part from MRI Image. The testing consequences demonstrate that the proposed system accomplishes to give the good result related to the available strategies. Xilinx Virtex-5 FPGA is used to implement in this paper.

脑图像分割的目标是在反映众多解剖结构的不同非相容同质区域中分离图像。通过磁共振进行的大脑分割对诊断大脑紊乱有很多意义，例如阿尔茨海默氏症、帕金森相关综合征等。然而，自动分割MR图像并不是一项简单的工作。本研究的主要目的是通过解决目前文学作品中存在的问题，为MRI图像中ROI（Region Of Interest）区域的分割提供更好的分割方法。MRI 分割不是一项微不足道的任务，因为获取的 MR 图像不完美，并且经常被噪声和其他图像伪影破坏。图像处理技术的多样性促成了许多图像分割技术的产生。那是因为没有通用的方法，也不是所有的方法都必须适用于适合所有图片的特定形式的图片。例如，其他方法仍然使用灰度直方图，而其他方法则针对渗色情况整合详细的空间图片细节。有些方法使用统计技术，但有些方法确实结合了现有信息以提高分割效率。一些方法利用概率或模糊方法。然而，目前的所有流程都存在某些不便之处。因此，我们打算为ROI区域分割提出一种新的分割方法。拟议的工作包括三个阶段，即预处理、边缘检测和分割。首先，从数据库中提取 MRI 图像，并通过应用高通滤波器对每个输入图像进行增强。完成预处理方法后，使用增强的canny边缘检测（ECED）方法对图像进行增强。之后，将图像提供给改进的分水岭分割 (MWS) 算法，该算法将 ROI 部分与 MRI 图像分开。测试结果表明，所提出的系统能够提供与可用策略相关的良好结果。本文采用 Xilinx Virtex-5 FPGA 来实现。增强精明边缘检测（ECED）方法用于增强图像。之后，将图像提供给改进的分水岭分割 (MWS) 算法，该算法将 ROI 部分与 MRI 图像分开。测试结果表明，所提出的系统能够提供与可用策略相关的良好结果。本文采用 Xilinx Virtex-5 FPGA 来实现。增强精明边缘检测（ECED）方法用于增强图像。之后，将图像提供给改进的分水岭分割 (MWS) 算法，该算法将 ROI 部分与 MRI 图像分开。测试结果表明，所提出的系统能够提供与可用策略相关的良好结果。本文采用 Xilinx Virtex-5 FPGA 来实现。