Abstract: This paper focuses on segmenting human from photo images. It has found several applications like album making, photo classification and image retrieval. The result can be further applied to many useful applications like part recognition which can be further applied to gesture analysis as well as in tracking. Segmenting human from photo images is still a challenge because of numerous real world factors like shading, image noise, occlusions, and background clutter and also because of great variability of shapes, poses, clothes etc. Previous works on human segmentation requires shape-matching processes. In this paper, we propose a simple method to automatically recover human bodies from photos. We use some haar cascades to detect human body that is haar cascade_upperbody and haar cascade_lowerbody which helps in performing upperbody and lower body segmentation. We need to perform CT (coarse torso) detection using MCTD algorithm for accurate upper body segmentation. Lower body is then extracted accurately using MOH based graph-cut algorithm. Experimental results show that, the proposed algorithm works well on VOC 2006 and VOC 2010 data set for segmenting person with various poses. Thus achieving high performance compared to conventional methods.