基于位平面和协作表示的人脸识别算法

Face Recognition Using Bit-Plane Images and Collaborative Representation

为提高人脸识别的准确性和高效性,提出了一种基于位平面信息的协作表示人脸识别算法,利用8位灰度图像的8个位平面和协作表示原理进行人脸识别.采用直方图均衡算法增强图像的对比度,并将均衡后的图像分解成8个包含有不同信息的二进制位平面图像.根据各位平面图像包含的识别信息量对它们进行加权求和,构造虚拟人脸图像.最后在虚拟人脸图像数据库上采用协作表示进行识别.实验结果表明,该算法具有较高的识别率和较快的识别速度.

As a new representation approach, collaborative representation was employed to collaboratively represent the query image by all training images from all classes. In this paper, a novel face recognition algorithm using collaborative representation based on the binary bit-plane images had been proposed. The original gray images are firstly equalized and decomposed into 8 bit-plane images. And then the new virtual face images, which contained more discrimination information, were constructed by the 8 bit-plane images and the weight vector that was calculated by the recognition rate and the order of the bit-plane. Finally the collaborative representation was performed on the virtual face images. Experimental results show the proposed combination approach has very competitive classification rate and recognition speed.