中文摘要：

在肺癌图像精细分类中，进一步区分小细胞肺癌、鳞肺癌、腺肺癌、细支气管肺泡癌还不够成熟，为此，在改进现有精细图像分类研究工作的基础上，利用无须码本与释文的快速模板匹配框架，融合了LBP（Local Binary Pattern）纹理特征和小波矩形状特征，提出了适合肺癌数据的精细图像分类新方法．将纹理特征与形状特征融合，通过分配两种特征的权重，用融合特征进行模板匹配．匹配结果表示成特征响应图的形式，再通过改进的均值空间金字塔模型，从特征响应图中抽取有用特征，进行分类训练．实验结果表明，该方法在肺部影像数据库联盟（LIDC）数据库上达到了91．75％的平均正确率，证明了肺癌图像精细分类方法的有效性．

英文摘要：

Lung cancer fine image is further divided into small cell lung cancer, squamous lung cancer, gland cancer, bronchioloalveolar carcinoma, yet can not be achieved. To this end, with improvement based on the existing fine image classification work, this paper takes advantage of fast template matching framework and the wavelet moment features fused with LBP（Local Binary Pattern）texture, presenting data for lung cancer fine-grained image classification method. This paper will feature the texture and shape of the feature fusion, by assigning two feature weights, characterized by the fusion template matching. Matching results expressed as characteristic response in the form of graphs, and through improved mean spatial pyramid model, extract useful features characteristic response from the figure for classification training. Experimental results show that our method on LIDC （Lung Image Database Consortium）database reached an average accuracy rate of 91.75%, which is the basic proof of the effectiveness of our lung images fine classification method.