一种基于高动态范围(HDR)人脸图片的室外光照估计算法

An Algorithm for Recovering Outdoor Illumination Information from High Dynamic Range(HDR)Face Images

从人脸图片中进行室外光照估计在增强现实、人脸识别、人脸图片防伪等多个方面的应用都有着重要意义。由于低动态范围(LDR)人脸图片的动态范围小,会丢失一部分信息,用于光照估计会不准确,而HDR图片则有更大的动态范围,包含完整的信息。因此不同与以往用LDR图片进行光照估计的算法,提出一种基于HDR人脸图片进行光照估计的算法。该算法首先通过对人脸图片进行三维建模,得到人脸几何;然后利用朗伯体反射模型以及普利兹天空模型建立人脸光照模型,得到参数模型下人脸的着色;最后用参数模型下的人脸着色信息和真实人脸图片中的着色信息形成能量方程,并对能量方程进行优化求解得出光照参数。实验结果证明该光照估计算法的准确性。

The outdoor illumination estimation from the face images has important significance in many aspects such as augmented reality,face recog⁃nition,and face image anti-counterfeiting.Because the dynamic range of Low Dynamic Range(LDR)face images is small,a part of infor⁃mation will be lost and used for lighting estimation will be inaccurate,while the HDR images have a larger dynamic range and contains complete information.Therefore,different from the previous algorithms for illumination estimation using LDR pictures,proposes an algo⁃rithm based on HDR face images for illumination estimation.The algorithm firstly obtains the face geometry by 3D modeling the face im⁃age.Then,using the Lambert Reflection Model and the Perez Sky Model to build a face illumination model,the face coloring under the pa⁃rameter model is obtained.Finally,the energy equation is formed by the face coloring information under the parameter model and the color⁃ing information in the real face image,and the energy equation is optimized to obtain the illumination parameters.The experimental results prove the accuracy of this illumination estimation algorithm.