The human image segmentation algorithm based on face detection and biased normalized cuts

Abstract

Attributed to pose variation (frontal, profile, et.), the color and texture difference of clothes, the presence of noise, low contrast, uneven illumination and complex background. There are enormous difficultly in human image segmentation. In this paper, we propose an automatic human image segmentation method based on the face detection and biased normalized cuts. First, we use face detection algorithm to detect human faces, and get facial contours. Then we establish object seeds estimation model based on the position of the detected face, and get the object seeds. Using these seeds, we use biased normalized cuts algorithm to segment the image. Finally, we perform region merging based on the previous seeds and segmentation results, and the image is divided into two parts (object and background). We implement a large amount of experiments over a public segmentation database of Berkeley etc. Experiments show that our method can segment different types of human image and obtain satisfactory results. Compared with Grabcut method, our propose method can be obtained more accurate results in many images. Qualitative and quantitative experimental results demonstrate our method produces high quality segmentations and effectively improve the segmentation efficiency.