Abstract: In order to improve robust performance of illumination and greatly reduce the dimension of original face images，this paper proposes a face recognition algorithm based on bit-planes in compressed domain. The method can take the advantages of face feature directly extracted in compressed domain and bit-planes of image，and the reconstructed expression of face image.The experi-mental results indicate that the proposed approach not only increases the recognition rate，but also needs fewer projection vectors.

Key words: Discrete Cosine Transform（DCT）, bit-plane, Principal Components Analysis（PCA）

摘要： 针对光照等干扰因素对人脸识别的不利影响，以及人脸图像的维度过高等问题，提出了一种新的基于位平面的压缩域人脸识别算法。该算法综合利用了图像丰富的位平面信息和压缩域内图像处理技术，提出新的人脸特征表达方法。实验结果表明,该方法不仅能够有效地降低图片维度，并且具有较好的鲁棒性。

关键词: 离散余弦变换, 位平面, 主成份分析（PCA）