CAJ | 학술논문

基于纹理特征的方法被广泛应用于人脸识别。然而纹理特征依赖于图像的高频细节信息，当图像出现模糊时，单纯利用纹理特征的识别方法的识别精度会急剧下降。为了克服纹理特征的在模糊人脸识别中的不足，提出了一种基于色彩特征和纹理特征融合的识别方法。首先参照人类的对立色感知机制提取人脸的色彩特征；然后，将该色彩特征和纹理特征分别用于识别分类；最后，将二者的识别相似度进行融合，得到最终的识别结果。该色彩特征描述了图像的低频信息，其对图像模糊不敏感，并且与描述图像高频信息的纹理特征具有良好的互补性。在FERET和AR人脸库上的实验表明，融合色彩特征和纹理特征有效地提高了模糊人脸的识别精度。
기우문리특정적방법피엄범응용우인검식별。연이문리특정의뢰우도상적고빈세절신식，당도상출현모호시，단순이용문리특정적식별방법적식별정도회급극하강。위료극복문리특정적재모호인검식별중적불족，제출료일충기우색채특정화문리특정융합적식별방법。수선삼조인류적대립색감지궤제제취인검적색채특정；연후，장해색채특정화문리특정분별용우식별분류；최후，장이자적식별상사도진행융합，득도최종적식별결과。해색채특정묘술료도상적저빈신식，기대도상모호불민감，병차여묘술도상고빈신식적문리특정구유량호적호보성。재FERET화AR인검고상적실험표명，융합색채특정화문리특정유효지제고료모호인검적식별정도。
The texture feature based methods are widely used for face recognition. However, as the texture feature relies on the high frequency details of the image, the performance of a method that merely utilizes texture feature deteriorates drastically when the image is blurred. To overcome this defect of the texture feature, a method fusing color and texture features was proposed for blurred face recognition. This method extracted a type of facial color feature using an opponent color model which was in accordance with the human visual mechanism. This type of color feature and a certain type of texture feature were used for recognition separately, and then their similarity scores were fused to make final decision. The color feature is a description of the low frequency components of the image, which is robust to image blur and complementary to the texture feature. Experiments on the FERET and AR face databases show that the recognition performance for blurred face image is effectively improved by fusing the color and texture features.