| Summary: | 碩士 === 國立臺南大學 === 數位學習科技學系碩士班 === 100 === Hair is not only an important feature for human appearance but also became a part of nowadays fashion. Automatic hair region labeling plays an important role in the studies of human identification, hair styles classification, hair sketch and modeling. Since its solid property, variety colors, and different environmental illumination conditions, it is a challenging issue for automatic hair segmentation. In this thesis, we propose an effective algorithm to segment hair in an image. We get the seed regions of hair using mouth corners and cheek’s skin color information, then we take advantage of maximum likelihood and iterative convex hull region growing method to discover the hair region. The experimental results show that the proposed method can conquer the problems coming from illuminations and dyeing, and the hair regions can be successfully segmented.
|
|---|
