Research On Image Segmentation Methods Based On Superpixel

Image segmentation is the technique of dividing the image into a number of meaningful target areas.It divides the image into several regions with unique properties through grayscale,color and texture,so that the same region has the same properties under certain criteria.There is a clear distinction between adjacent areas,and from which to extract the region of interest.Image segmentation is an important field in digital image processing.At the same time,image segmentation is also an important step in image analysis and understanding.The traditional image segmentation method is difficult to meet the application requirements for the high-resolution image.In 2003,Ren proposed the concept of superpixel.Superpixel segmentation is an image preprocessing technology,superpixel instead of the original image pixel as the basic unit of the subsequent processing,so you can get the image of redundant information,reduce the processing of images The complexity of the implementation of the follow-up algorithm to speed up.At the same time,superpixel is more conducive to the extraction and expression of image features,but also a better suppression of the impact of noise.In this paper,based on the study of image segmentation theory,in view of the limitations of the traditional segmentation algorithm and the superpixel segmentation technology in recent years,this paper proposes a method based on the superpixel segmentation method.1.This paper firstly introduced the traditional image segmentation algorithm,compared the advantages and disadvantages of such methods.Then the basic idea and principle of the main superpixel segmentation algorithm are introduced in detail,and the contrast experiment of the superpixel segmentation algorithm is done.2.In view of the limitation of traditional segmentation algorithm,a segmentation algorithm based on SLIC superpixel and region growth is proposed,and the performance of traditional regional growth algorithm is improved by using superpixel.Experiments on the Berkeley segmentation database show that the method can achieve better results for the segmentation of color images.3.Using the algorithm of SLIC superpixel combined with region growth proposed in Chapter 3,the organs of CT images were divided into two parts.The texture features of the image were incorporated into the growth criterion of regional growth algorithm.Experiments on the cancer image database published in TCIA show that this method can achieve better segmentation of CT images in medical abdomen.