基于彩色纹理分析和神经元网络的结肠镜图像异常检测

Abnormality Detection in Colonoscopic Images Based on Colour Texture Analysis and Neural Network

提出了一种新的基于彩色纹理分析和多层前馈型神经元网络的方法,对结肠镜图像进行异常检测。通过对两种常用的灰度图像纹理分析方法(同现矩阵方法和离散小波变换方法)进行扩展,提出相应的彩色纹理分析方法。彩色纹理分析提取出的特征参数构成作为识别器的多层前馈型神经元网络的输入。实验结果表明该方法能有效检测结肠镜图像中的各种异常。

This paper presents a new scheme for abnormality detection in colonoscopic images based on colour texture analysis and neural network. Two colour texture analysis methods are proposed by extending two grey scale methods respectively, i.e. cooccurrence matrix and discrete wavelet transform. Experimental results indicate that this system is able to detect various abnormality in colonoscopic images.