The Tissue Segmentation And Feature Description Of Multi Spectral Medical Images

Multispectral medical images are useful to assistant medical diagnosis and treatment.The accurate tissue automatic segmentation and feature description of multi spectral medical images play an important role in marking abnormal tissue.The traditional segmentation method is mainly focus on single target,however,the multi spectral images contain a few kinds of tissues and a variety of features.The significance of different features in each band image is different.Therefore,this research emphasizes at multi tissues segmentation by different features.Firstly,the feature salient graph is used to analyze the features significance of each band images of multi spectral medical images,and the most significant tissue of each band is determined as the target tissue to be segmented.Although by using the GLCM texture description method the target tissue of each band image is segmented for the tissues with high significance,it is difficult to segment the target tissue whose significance is not high.After analyze the correlation of different band images by Sheffield index,a method of enhancing the significance of the target tissue by selecting effective spectral information is proposed.Image maximum between-cluster variance is used to determine the segmentation threshold value of target tissue whose significance is enhanced.The effectiveness of the tissue segmentation method is verified by intestine multi spectral images.Secondly,based on the variety of the tissue vessels and the gray value of the surrounding pixels of vessel,improved LBP is proposed to construct vessel description operator.The vessel is extracted through the matching degree of the pixel neighborhood and the vessel description operator which is calculated by binary coding.Finally,the segmented tissues are described with the features of the region features,the pattern of tissue growth features and the contour features,besides,the gray features,the gray statistical features and the skewness and kurtosis features are also used to describe the tissue features.These features are used to recognize lesion.