Analysis and Synthesis of Human Lips for the Application in Video Communication

• Main Authors: Houng, Loung-Choung, 洪榮聰
• Other Authors: Wang Jia-Shung
• Format: Others
• Language: zh-TW
• Published: 1998
• Online Access: http://ndltd.ncl.edu.tw/handle/94941897511184064871

碩士 === 國立清華大學 === 資訊工程學系 === 86 === In this thesis we describe a system which contributes to the analysis andsynthesis of specific images containing a human head with facial actions during speech. Among all actions we only concerns the motion and deformation of the mouth of human head within the images. At first, we apply several techniques including edge enhancement, and profileextraction in HSV color space, to extract the feature points on the speaker's mouth in each frame of these images. According to these feature points, we might analyze the motion and deformation of mouth of the speaker during speech. We also could build up the lip shape model for each frame of images, and might determine parameters for controlling the shape of mouth. After the shape of lips has been determined, we could adjust the generic 2-D lip mesh model to the 2-D personalized lip mesh model within each image by the triangle warping technique. In the step of image synthesis, we take the first frame of the images containing a human head as background, obtain the texture data of lip mesh model from this frame, and stitch the personalized lip mesh model into the background to generate final video sequence. The new video displays the simulation that the human articulates syllables with motion anddeformation of mouth. Finally, we can estimate the control parameters for the lip shape model in a sequence of images by a set of functions. The work can reduce large amount of storage and require less bandwidth in network transmission.