图像栈的特征提取以及在线虫分类中的应用

Feature extraction of image stack and its application in nematode classification

三维多焦距图像栈中的有效信息分布在不同的图像层上,其特征提取及分类跟普通二维图像有很大区别。针对图像栈的分类问题,提出了一种基于多方向图像融合的多线性特征提取和分类方法。首先,通过图像融合获取三维图像栈沿多个正交方向的融合图像,并从中提取特征;然后,通过典型相关分析(canonical correlation analysis,CCA)方法,将不同方向融合图像提取的特征进行融合并抽取组合的典型相关特征用于图像分类;其次,由于图像栈数据包含了样本、类别和方向等多个影响分类的因素,因此将多方向图像融合方法嵌入到多线性分析中,综合考虑多个因素交互作用时对图像栈分类的影响。本文提出的方法在线虫图像栈数据上进行了实验,识别率达到了97.0%,实验结果表明该方法具有较高的准确性。

The effective information of 3D multi-focal image stack is distributed on different image layers, so the feature extraction and classification of image stacks are significantly different from that of 2D images. In this paper, an image fusion based muhilinear analysis approach is presented to use for classification of muhi-focal image stacks. First, the image fusion techniques are used to combine the relevant information of multi-focal images within a given image stack into a single image. Besides, multi-focal images within a stack are fused along 3 orthogonal directions, and multiple features extracted from the fused images along different directions are combined by using canonical correlation analysis （ CCA ）. Furthermore, because multi-focal image stacks represent the effect of different factors- texture, shape, different instances within the same class and different classes of objects, the image fusion method within a muhilinear framework is embeded to propose an image fusion based muhilinear classifier. The experimental results demonstrate that the multi- direction image fusion based multilinear classifier can reach a higher classification rate （97%） than other classification methods.