Three Dimensional Reconstruction And Analysis On Images Of Skin Scars

In the field of no-invasive diagnosis for human skin scars using their image information,two-dimensional image processing is the most popular method.However,this method has its intrinsic defects like loss of integral information.Therefore,three-dimensional image processing based on two-dimensional image can solve these problems.This thesis studied on 3-dimensional reconstruction and quantitatively analysis about the elastin second harmonic generation images to observe the difference of different types and durations of human skin scars.On the stage of 3-dimensional reconstruction,firstly,pre-processed the image and extracted the destination area including elastic fibers,secondly,reconstructed the 3-dimensional image using different methods and compared the differences of each other.After that we adopted 3-d thinning algorithm which can perfectly maintain the topology of the object to extract the skeleton of the 3-d image.On the stage of quantitatively analysis,we calculated a series of quantitatively features like branch length,branch node number,branch number,branch volume ratio,branch cross-sectional area and direction homogeneity based on the above processing steps.According to these features,we can quantitatively analyze the difference of different types and durations of human skin scars.Finally,based on the features,this thesis compared the difference between one-dimensional regression model and random forest regression model in prediction of human scar duration and the result showed random forest regression model had better performance.