【摘要】 该文对文献［３］中的相似度计算进行了改进，用低维矩阵替代高维矩阵进行计算。先证明没有改变原有图像矩阵的相似度。然后用新方法和原方法解决一组人脸图像识别问题。实验结果表明，新方法大大减少了计算量，又能保持稳定的识别率，具有较高的识别效果。更多还原

【Abstract】 This paper improves the similarity of matrices proposed by reference ,replaced high order matrix by low order matrix, to reduce calculation. First we prove that the similarity of original image matrices is kept. Then the new method is used to solve the recognition of human facial images. Experimental results have shown that compared with reference, the method is effective, and offers high recognizability.更多还原