Denoising And Interpolation Of Open Curves With Applications On Post-processing Pulmonary Fissure Segmentation

In the medical images or natural images,a large number of features can be expressed by the open curves.The detections of open curves are of significance to the segmentation,registration and target recognition in images.Open curves detections have difficulties with ever-changing shapes,difficulty to build models,noises interfering etc.In the detection of lung fissures in the medical image,the pulmonary fissures appear as thin,weak and varying structures,low contrast,ununiform density distribution in the two-dimensional section.This is further complicated by interferences of other tissues.Based on post-processing and enhanced threshold images,a smooth method with denoising and interpolation and a curve denoising algorithm based on the active contour model against the above difficulties are adopted to detect the pulmonary fissures.The main work includes:1.A denoising and interpolation algorithm based on least squares idea and parameter estimation is proposed to solve the problem that the open pulmonary fissures curves contains fractures and few noises in the binary image.Firstly,for the two-dimensional slice,give the noises on the image a low weight,the missing data zero weight,and the size of the missing data are estimated from the entire data set.This paper obtain the residual square sum based on the penalized least squares,and also the minimization criterion of the image.Then the second order differential is used to express the smoothness,and the minimization criterion of deformation is surely obtained.The parameter S in the minimization criterion is estimated by generalized cross validation(GCV).This method consists in constructing weights function by using the current residuals and updating them,from iteration to iteration,until the residuals remain unchanged,and achieve robust smoothing denoising and interpolation.Finally,the experimental results show that this curve de-noising algorithm can effectively reduce the noise and compensate for the missing fissure.2.Aiming at the enhanced threshold image which contains fractures and noises,this paper proposes an improved algorithm to segment pulmonary fissures based on active contour model open curves.Firstly,the image is marked by analyzing the connected domain.We obtain the distance transformation of the marked image,so does the distance field,and then detect the ridge points of the open curves according to positive and negative changes of the gradient value of the image gradient.Taking two ridge points together randomly completes the curve initialization.With the distance field as the image force field,the constant value of the force is larger than the end force of the end point as the tension to form the tensile field.We also add constant stretching forces which are larger than the end force to the ends of the curves to form the tensile field.The image force potential field and the tensile force field form the external force potential of the curve.The curve begins to evolve by minimizing the internal force and external force of the curve.Then set the curve evolution and iteration termination conditions to terminate the curve evolution,thus detecting the curves on the lung fissure image.Finally,we remove the curves whose length is less than the threshold by setting the length of the threshold constraint.and Get the segmentation of the fissures.Finally,the experimental results show that this curve de-noising algorithm can effectively reduce the noise and compensate for the missing fissure.3.In order to verify the validity of the algorithm,we carry out experiments using the typical data in GLUCOLD database,and compare with the accurate lung fissure reference.The correct rate is used to judge the integrity of lung fissure,False Negative Rate and False Detection Rate indicate whether the fissures obtained using the algorithm are more or less detected.The experimental results show that the algorithms in this paper are stable and effective.The detection of pulmonary fissure has certain practical significance for the treatment of pulmonary diseases.