Image segmentation forms an important concept in the computer vision technology. Image segmentation breaks the image into boundaries that differentiate meaningful components. For computer vision to realize its full potential it is essential that the image segmentation algorithms give accurate results in a fast and efficient way. In hierarchical architecture based IoT networks set up to "see" the world, methods of computer vision need more analysis. The need for low cost setup for IoT networks in terms of memory and their computational capabilities demands research for developing methods that are resource sensitive and can be successfully integrated into such networks of low end IoT servers. Addressing this need, a refined graph cut segmentation technique for low to medium resolution images and for constrained devices is presented in the paper. Implementation and analysis of the refined graph cut segmenter for linux based IoT servers is discussed. A comparison with the contemporary segmentation methods under similar constraints is also presented. © 2019 Procedia Computer Science. All rights reserved.