表情比例图的鲁棒性增强算法  

Enhancement Algorithm of Robustness for Expression Ratio Image

作　　者：刘剑毅[1] 

机构地区：[1]西安交通大学人工智能与机器人研究所,西安710049

出　　处：《西安交通大学学报》2006年第8期919-922,927,共5页Journal of Xi'an Jiaotong University

基　　金：国家自然科学基金资助项目(60205001)

摘　　要：针对表情比例图(ERI)方法的光照条件及对准误差的鲁棒性较差的问题,提出了一种新的比例图滤波方法,该方法利用形态学膨胀算子提取对准误差所致的噪声区域及其过渡带,然后利用线性插值恢复出平滑的比例图,同时还提出了基于商图理论的光照校准方法,并应用于照明条件不一致下的表情绘制,即通过求解空间3个独立点光源的组合系数,实现了样本图从任意光照向期望光照的校准.实验结果表明,所提方法能够探测并消除对准误差带来的比例图噪声,能够实现不同照明条件下的样本图像光照校准,从而提高了ERI算法应用于表情合成时的鲁棒性.Aiming at the problems of lacking robustness on illumination and image alignment in expression ratio image（ERI） algorithm, a novel ERI filtering method was proposed, which extracts the noise region and its margin induced by un-precise image alignment using dilation opera- tor, and retrieves a smooth ratio image by linear interpolation. A quotient image theory based illumination calibration method for facial expression rendering under un-uniform lighting conditions was also proposed. This method calibrates the sample images from various illuminations to the expectant same illumination by finding the combination coefficients of three independent light sources. Experiment results show that the proposed methods can detect and eliminate the ERI noise induced by un-precise image alignment, and they can implement the illumination calibration for sample images coming from various lighting conditions and so that enhance the robustness of ERI algorithm applied to facial expression synthesis.

关 键 词：表情比例图 形态学膨胀 光照校准 表情合成 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]