Font Size: a A A

Research On Face Detection And Facial Features Extraction

Posted on:2005-06-22Degree:MasterType:Thesis
Country:ChinaCandidate:X C XiaoFull Text:PDF
GTID:2168360125958877Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
In recent years, human face has already become an important research object in the field of computer graphics and image processing. Today, the study on human face mainly covers five aspects, 1) face detection, 2) human face features extraction, 3) the recognition of human face, 4) the reconstruction of three-dimensional human face, and 5) the human facial expression animation with 3D. This thesis mainly focuses on the face detection and human face features extraction.Firstly we explore the clustering characteristics on xanthous face based on all kinds of color models, and we also think the color models such as HSI model and YCbCr model can be well used to express the clustering characteristics of xanthous faces. Then, we study on the distribution character of complexion of human face in HSI and YCbCr color models, and set up an appropriate complexion model suitable for xanthous face. Based on the model, we propose a good detection algorithm on the human face featuresWe use the model called Active Contour Model (the Snake method) to detect shapes of the five sense organs on a human face. Although the Snake method is very common, it is very useful after we improved it in many aspects. Firstly, the Snake initial curve can be automatically gotten in our improved Snake method in both simple and complex image background. Secondly, we have also gotten a better precision in which the Snake curve can fit shapes of the five sense organs on a human face better after we reconstructed the image energy item in the Snake's energy function according to the complexion characteristics of human face. Thirdly, a new energy item based on the shape characteristics, i.e. the shape restriction energy item, is added to the Snake function to restrict the shape of Snake during the Snake's iteration computation to prevent the Snake curve from converging mistakenly.Finally, we developed a software program based on our algorithm. Results from our experiments show that the Snake curve produced by our improved Snake method can fit automatically shapes of the five sense organs on a human face more quickly and better. When we have the lasted Snake curves we have had all shape characters of the five sense organs on a human face.
Keywords/Search Tags:Human face, Feature detection, Feature extraction, Snake, Active contour model, Energy function
PDF Full Text Request
Related items