Semantic Segmentation and Contextual Information Based Image Scene Interpretation: A Review

摘要

The images in the scene consist of several objects that depict different relationships among themselves. Interpretation understands those relationships. Thus, scene interpretation is a scene description in which the scene models are consistent with the evidence, context information, and world knowledge. On one side, images in scene interpretation are useful in extracting the information that is related to the physical world and is meant for human operators. On the other side, it has always constituted a great challenge because of the varieties of complex objects due to which computer vision is not much capable to comprehend the information regarding the images in the scene. It also requires many efforts to extract the deeper meaning of the scene. So in review paper, we are going to summarize the methodology proposed by various researchers in terms of semantic segmentation and contextual information to interpret the images and highlight their contributions and challenges which still persist. Our analysis of the different methods proposed is also provided to draw some conclusions.