分形编码在人脸识别中的应用

利用人脸固有的对称特性,提出了基于对称性与方差的快速分形人脸图象压缩算法.对于一值域块,与其匹配的定义域块被限定在其对称区域中,从而可加快编码速度.本文分析了分形码的特点,定义了图象间分形码距离,进而提出了基于分形码的人脸识别算法,并对该算法的时间复杂性作了分析.通过在ORL和YALE人脸库上的实验结果表明,本文提出的识别算法的识别率分别可达到88%和92%,与本征脸方法基本持平,比直接采用基于方差的分形编码方法要高,且本文算法在人脸有装饰物情况下具有一定的稳定性,实验还对识别算法中参数的取值作了分析.

Application of Fractal Encoding in the Face Recognition

The fast fractal face image compression algorithm based on symmetry and variance is proposed by using the inherent feature of facial symmetry. The matched domain blocks are restricted in the symmetrical region for a range block that can speed up the encoding speed. After analyzing the features of fractal codes, this paper defines the distance of fractal codes between images, furthermore, this paper proposes the face recognition algorithm based on fractal codes and analyzes the time complexity of the algorithm. The experimental results on the ORL and YALE face databases shows that the recognition rate of the proposed algorithm can achieve to 88% and 92%, respectively, that remains the same with the eigenface method and is higher than that of the variance based fractal encoding method. The proposed algorithm has certain robust with the face changing in accessories. The value of the parameter in the algorithm is also analyzed.