Research On Segmentation And Feature Extraction Algorithms Of Cervical Cell Images

With the rapid development of automation,informationization and intellectualization and economy,people pay more and more attention to physical health problems and hope to diagnose the existing problems quickly.Cervical cancer is a gynecological malignant tumor with a high incidence,which has seriously affected women's health.At present,research on cervical cancer has been attached great importance in the field of pathology at home and abroad.This paper focuses on the key technology of cervical cell smear auto-reading system and how to overcome the existing difficulties in cervical cell auto-reading technology,through studying the related principles of image processing and pattern recognition.The related research mainly includes the semantic description of cervical cell image,the segmentation of cervical cell image and the feature extraction of image.It is of great significance to solve the key technology of cervical cell auto-reading.The cell images collected by microscopy are complicated because of the need to collect,produce and dye.Firstly,this paper focuses on the segmentation method of cervical cell images,and then analyses and studies how to extract the accurate features of single cervical cell obtained by segmentation.The specific related research contents are described as follows:1?There are a large number of overlapping cells in cervical cell image,that is,there are intersections between image cells and cells.In the research of cervical cell image segmentation,a hierarchical and efficient cervical cell segmentation method based on Graph Cut algorithm is proposed to segment overlapping cells accurately and extract individual cells.Firstly,the target cervical cell image is obtained,and the image segmentation processing is completed by solving the optimal segmentation threshold of the initial cervical cell image,and then the image segmentation algorithm is used to segment the first segmented image again to obtain a single target cell.After the processing in this paper,the efficiency and accuracy of image segmentation are effectively improved.2? It is necessary to separate the nucleus and cytoplasm of cervical cells after the preparation and segmentation of single cells in cervical cell image.This paper mainly discusses how to improve the accuracy and efficiency of the separation of the nucleus and cytoplasm of cervical cells.A method based on Grab Cut algorithm and Canny operator edge detection is proposed to effectively separate the nucleus and cytoplasm of cervical cells.The gray histogram is obtained by gray level change pretreatment of cervical cell image,and the Canny operator double threshold is determined by histogram,and then the contour of cervical cell nucleus and cytoplasm is obtained.According to the contour obtained,the improved Grab Cut algorithm is used to separate nucleus and cytoplasm,and the efficiency of the Grab Cut algorithm is improved accordingly.3?In the aspect of feature extraction,the basic morphology of cervical cells was studied,and the basic shape and texture features were extracted.At the same time,through the in-depth study and analysis of the basic principles of K-means algorithm,a K-means feature extraction clustering algorithm is proposed for descriptors of nuclear area,nuclear-to-cytoplasmic ratio and RGB color mean value of nuclear cytoplasm,and the algorithm is improved.Then appropriate shape and texture characteristics were selected according to the experimental data,and a particle size characteristic suitable for cervical cancer cells was added according to the characteristics of cervical cancer cells.Finally,the appropriate features are selected according to the experimental data.