نوع مقاله : مقاله پژوهشی
گروه مهندسی کامپیوتر دانشگاه اراک
دانشکده مهندسی، دانشگاه اراک
Image segmentation is a basic issue in machine vision. One of the important tasks of machine vision and image processing is to recognize the pattern and one of the most important algorithms is U-Net segmentation. The U-Net algorithm has been identified as a popular algorithm in recent years due to its accurate response, high accuracy, high processing speed and learning, no need for large data sets for learning and no need for complex and expensive hardware. Image components and their fragmentation have become part of medical image processing. In this paper, we explain the U-Net algorithm and its convolutional network, as well as the most appropriate setting for the parameters and super parameters of this algorithm to optimize and achieve maximum accuracy in solving image-processing problems with this algorithm. In other words, a proposed method for segmenting medical images is performed on the DIC-C2DH-HeLa data set, which based on an architecture, the so-called "fully convolutional network", we have modified and expanded this architecture in such a way that with educational images Work very little and provide more segments that are detailed. The results showed that the proposed method has a higher accuracy than the other proposed method.
Keywords: U-Net Algorithm, deep learning, medical image processing, segmentation