Research On Cell Images Segmentation Algorithm Based On Grabcut And SLIC

The complicated background and irregular shape of cell images cause the difficulty of using computer vision technology to process cell images.Therefore,in practical application,the segmentation algorithms of cell image must have high precision and high efficiency.In this paper,two improved algorithms are proposed to resolve the problems which the existing cell image segmentation algorithms are sensitive to noise and can not divide boundary precisely.At present,in the field of cell image segmentation,the algorithm based on energy minimization and the segmentation method based on superpixel are two popular methods.In terms of algorithms based on energy minimization,traditional Grabcut algorithm greatly reduced user interaction,but it still needs to select the rectangle manually,and the degree of automation is low,so it is difficult to meet the practical application requirements.In terms of superpixel segmentation algorithms,the SLIC algorithm has a better evaluation on the boundary preserving degree and the superpixel shape.However,in the region where the edge of the cell image is fuzzy,superpixel classification errors may occur in the iteration.It will produce the wrong superpixel resulting in failure.In this paper,the Grabcut algorithm and the SLIC algorithm are improved respectively to ensure the accuracy of segmentation and the faster speed of computing.The main contents and innovations of this paper include the following aspects:1.A segmentation algorithm of cell images which is based on Grabcut algorithm combined with HFT model is proposed in this paper.First,the saliency map of the cell image is calculated using the visual saliency model HFT,and then the morphological closed operation is performed to obtain the initial contour of the cell image.The rectangle is constructed by the initial contour,and the rectangle is merged with the cell image obtained by the fractal wavelet adaptive denoising algorithm to complete the initialization of the Grabcut algorithm.Finally,the rectangle is passed to the Grabcut algorithm to achieve cell image segmentation.The experimental results show that the proposed algorithm can improve the accuracy of the algorithm and obtain high segmentation precision and fast segmentation speed,which can meet the requirements of cell image analysis.2.A segmentation algorithm based on improved SLIC and region merging of cervical cell images is proposed in this paper.First,the meanshift treatment is used to eliminate noise on the cervical cell image,then the two-dimensional Otsu adaptive threshold processing is conducted to abtain the initial contour,then based on SLIC algorithm the superpixel regions are obtained.The superpixel regions are fused to the original image to complete the initial segmentation.Finally,in the initial segmentation map,automatic marking is completed by combining the grayscale information and gradient information of the cell image,and the maximum similarity criterion is used to merge.It is not necessary to set the threshold in advance.The non-marker background regions were merged with labeled automatically,while the non-marker object regions are identified and avoided from being merged with background.Several experiments of segmenting the cytoplasm are conducted on cervical cells images.The proposed algorithm can extract cytoplasm from a single-cell cervical smear image accurately in a relatively short time.