基于图像分割和孔洞修复的三维伤口测量方法

伤口测量是临床医学研究中一项重要工作。传统的接触式伤口测量方法存在测量结果稳定性差、易造成伤口二次伤害的问题，基于2D图像处理的非接触式伤口测量方法存在精度差、无法获得伤口深度信息的问题。因此，本文设计了一种基于图像分割和孔洞修复的三维伤口测量方法。采用异源图像对齐算法解决了异源图像由于视差导致的图像像素错位问题；通过基于改进的区域生长法设计了一种交互式伤口区域分割方法，实现伤口区域精准分割；最后，采用基于RBF的三维点云孔洞修复算法进行伤口皮肤复原，从而得到伤口的最大深度、体积等三维参数。实验结果表明，本文方法的测量误差低于3%，其测量精度以及稳定性皆优于接触式和基于2D图像处理的伤口测量方法，满足临床使用以及医学研究的应用需求。

3D Measurement of Wound Dimensions Based on Image Segmentation and holes Inpainting

The accurate measurement of wound dimensions is vital in clinical medical research. The accuracy and stability of traditional contact measurement method are poor, which may cause further injury to the wound. The non-contact measurement approach based on 2D image cannot obtain the depth information of wound. Moreover, its accuracy is also poor. Therefore, an accurate3D measurement method of wound dimensions based on image segmentation and hole inpainting was proposed. The color image and depth image of wound were aligned in pixels via image alignment algorithm. An improved region growing algorithm assisted by interactive region selection was proposed for the accurate segmentation of wound region. Finally, the wound surface was reconstructed in the point cloud image by RBF-based algorithm. The depth information of wound could be achieved from the reconstructed wound surface. So the maximum depth, volume and other three-dimensional parameters of the wound could be calculated. The experimental results show that the measurement error of our approach was less than 3%, which was more accurate and robust than the traditional contact measurement method and the measurement method based on 2D image. The accurate 3D measuring method of wound dimensions proposed by this paper meets the application requirements of clinical and medical research.