中文摘要：

针对人脸训练集中的样本存在不同程度的变换（比如平移、旋转、缩放等），导致传统的基于稀疏表示的分类算法（SRC）、基于协同表示的分类算法（CRC—RLS）在分类时精度降低，提出了一种基于一阶和二阶信息的图像分类表示算法（SRC—FSD）。这种方法通过测试图像的像素值是由对应训练图像的邻近像素值图像的重建而成的，利用泰勒公式，考虑了一阶和二阶信息，目的是消除变换对图像造成的影响，从而提高测试样本的分类精度。最后在extended Yale B和AR人脸数据集上进行了不同维度下样本的对比实验，实验结果表明该方法具有很好的鲁棒性，有效地提高了人脸识别率。

英文摘要：

Considering different level transformations in training samples such as translation, rotation, scaling and so on, it may reduce the classification precision in traditional algorithms such as sparse representation based classification （SRC） and collaborative representation based classification with regnlarized least square （CRC_RLS）. To alleviate these problems, this paper developed a new method that an image representation classification algorithm combining the first and second derivative information （SRC_FSD） of image. This method used relative training image pixels to reconstruct the testing image. It adopted the first and second derivative information from Taylor formula to eliminate the effects of transformation of image, thereby improved the classification accuracy. Finally, experimental results on the extended Yale B and AR face databases in different dimensions show that the proposed method has excellent robust and improves face recognition rates efficiently.