Intelligent Image Segmentation Methods Using Deep Convolutional Neural Network

Image segmentation is conceded as one of the significant procedures in the digital image processing and computer vision tasks involving object detection, object recognition, anatomical structure analysis, etc. The majority of these computer-aided analyses require an intelligent segmentation approach for easy understanding of each image component separately, automatically, and spontaneously. Consequently, AI becomes the primary guiding medium towards achieving image segmentation intelligently. Several AI-based algorithms such as Deep Convolutional Neural Network (DCNN) and adversarial networks can successfully provide solutions to these intelligent segmentation approaches. However, the proficiency of these recent approaches is primarily data-dependent and problem domain-specific. Thus, it is essential to understand the existing model architecture to solve any specific task or develop customized segmentation models. Therefore, the main objective of this chapter is to discuss the concepts related to the image segmentation elaborately. As a result, in this chapter, different available dataset domains such as natural scenes, aerial imaging, autonomous driving and the purpose of image segmentation are discussed. The underlying general mathematical operations combined with the currently used handy performance metrics are reviewed. Besides, different segmentation types along with intelligent segmentation algorithms under each category are explained. Finally, the prevailing research challenges that still demand researchers’ attention are highlighted.