Segmentation in multiple levels for extracting salient regions of human body from single images

Importance of the digital image processing has increased in a massive way in recent days. Digital image processing has its own way in high end applications. Extracting the human pose from the images is a difficult task. In present days there are several researches have done for estimating the exact pose of the human. In several applications we require human body segmentation it is a difficult task. For example we required to do the segmentation of the image for understanding of the given picture and recognition of the activity in that picture. To manage with the extremely high difficult pose like human hand is behind him who is not visible in the image, complexity of the picture, and appearances of various human bodies in the picture. There are various existing works which need the computationally difficult training and some existing works requires the template matching processes. The main aim of this paper is to see the salient regions of human body to estimate the pose in the picture by using bottom to up style which in turn uses multiple level of segmentation. The approach is divide into five steps namely detection of the face, second one is segmentation in multiple level, third is detection of skin, fourth is segmentation of the upper body and finally the segmentation of the lower part of the body. Finally we can prove that the simulation results of our method achieve better results and high efficiency over the traditional existing methods. In this paper we improved the results with spline regression.