Research On Face And Features Location Algorithm

Positioning of the target image is a hot and important issue in image processing, andpositioning of facial features is one of its research orientations. What’s more, in the field ofobjective Chinese traditional medicine facial diagnosis, it is also inevitable to be mentioned.Based on the research of theories which are related to extraction of the area of skin andpositioning of the target image, author puts an emphasis on discussion about the extraction ofthe target area with connected domain algorithm to achieve the purpose of positioning theface image’s features.Main work is composed of two aspects as follows:In the aspect of the color image of the face detection, the author has made a furtheranalysis of the entire process based on skin color model and has created a skin model in theYCbCr space. After analyzing common skin model, the Gaussian model has been chosed,and put forwards specific steps and data.In the aspect of positioning of facial features, a connected domain algorithm has beenpresent to extract the target organ. Operate skin probability image and gray-scale image withthreshold segmentation in the obtained skin area separately. some common methods ofthreshold segmentation and edge detection also have been analyzed and contrasted in detail andchooses a better way to segment for different organs and images. Then according to thegeometric characteristics of the facial features, it sets pertinence extraction conditions toachieve the positioning of target connected component.We can get a complete algorithm combining algorithms of face detection and facialfeatures position described. As the positioning procedure involved in the research doesn’trequire strictly high quality of the input image, and it has a broad applicability, it has laid agood foundation for further study and application.