Design And Implementation Of Interactive Segmentation System On Virtual Human Image

Nowadays Digitized Virtual Human (VH) Project is one of the research focusesin the crossing field of information technology, biotechnology and othertechnologies, aiming at visualizing human body structure and simulating humanphysiological function to deepen the understanding of human itself and stimulatethe development of medical science and other sciences in related fields. Toimplement this project, it’s a critical step to segment VH images precisely, for it canprovides segmented data for modeling and visualizing.Image segmentation is a challenging task and yet there isn’t an universalmethod after years of research. Although many good methods are proposed oneafter another, methods for VH image segmentation are still short and haveshortcomings. The research purpose of this article is to design an appropriatesegmentation framework for VH image segmentation so as to provide valid andaccurate data for following image processing.Firstly, this article does some preparation work on the original dataset of VHimages, including converting file format, decreasing image resolution and croppingimage and so on, in order to read data more conveniently and reduce theinterference of invalid data to increase the efficiency of image processing.Secondly, this article segments the trunk and lungs of VH images in aninteractive way with the help of GrabCut method, and proposes an idea for sequencesegmentation to accelerate the segmentation work of trunk and lungs according tothe similarity and continuity of VH images. As for a heart with complex anatomicalstructures, this article uses color labeling method based on FloodFill to label theheart to reach the segmentation goal.Thirdly, this article does gray processing on those segmented data, fixes theimage problems of noise, unsmooth edges and inner holes, and extracts edges oftrunks and lungs to provide data for simulation of electrophysiological model.At last, this article merges all the segmented data into one single image withtrunk, lungs and heart data, in order to reduce the storage space and providedirectly-used data for following visualization.What’s more, this article designs and implements an interactive segmentationsystem for VH images, providing a support platform for segmentation research ofVH images.