基于压缩感知理论的HEp-2细胞识别

摘要

In view of the problems of the low quality of characterization and low recognition rate in HEp-2 cells recognition,the six category HEp-2 (human epithelial type 2)recognition method was presented based on the compressed sensing theory.The HEp-2 cell image was transformed into two-dimensional discrete wavelet domain.In the wavelet transform domain,the SVD (singular value decomposition)method was used to extract the characteristic value of HE-2 cells image.Decisions for the HEp-2 cell category were made using compressed sensing classifier.Computer simulation results show that the HEp-2 cell recognition method based on compressed sensing theory has not only high recognition rate,but also high quality of characterization,low computational complexity and easiness to implement.%针对 HEp-2细胞识别中低特征描述质量及低识别率等问题,提出一种基于压缩感知理论的识别6类 HEp-2(hu-man epithelial type 2)细胞的方法。对HEp-2细胞图像进行二维离散小波变换(two-dimensional discrete wavelet transform,2D-DWT)；在小波变换域中,采用奇异值分解(singular value decomposition,SVD)方法对 HEp-2细胞图像进行提取特征值；采用压缩感知分类器对 HEp-2细胞做出类别判断。仿真结果表明,基于压缩感知理论的 HEp-2细胞识别方法识别率高,具有高质量特征描述、低计算复杂度及易于实现等特性。